Patent Grant
12018404
The present disclosure relates generally to optimizing manufacturing processes. More particularly, the present disclosure relates to system, controller, and method for optimizing decortication processing of biomass products.
Decortication is a process that separates and may also grind bales of whole straw, also sometimes known as the stalk, of certain dicotyledonous plants into various fibrous and non-fibrous fractions for internal use or sale to others. In a particular case, the straw can be hemp straw.
The fibres from hemp stalk typically provide certain benefits; for example, the stalk typically has a high strength to weight ratio. As another example, the stalk fiber contents can typically have properties such as being anti-bacterial and resistant to water.
From an agronomical perspective, hemp typically grows rapidly in a variety of soils and climates. It is typically resistant to crop pests, reducing the requirement for herbicides and pesticides. It is a trio “product” crop, capable of producing seed, stalk, and valuable biochemicals, at above average volume and dollar yields, compared to some other crops.
Once processed, the typical components of hemp straw, also known as the stalk, are bast fibre, hurd, short fibre and dust. Each of these components can become valuable raw materials for use in a variety of consumer, commercial and industrial products. These components can be utilized in many industries, for example, consumer goods, automotive, construction materials, industrial absorbents, and animal and pet care.
Typically, decortication is used to fractionate whole hemp straw into various components, such as clean long and short bast fibre, clean hurd, and hemp dust. The bast fibre fraction is typically approximately 20-25% by weight of a decortication system's product, depending on, for example, the variety being grown, agronomic practices, retting extent, moisture content at time of processing, decortication method, and desired product cleanliness (i.e. extent of contamination with hurd or dust). Examples of applications for “clean” bast fibre (i.e. >95% purity) include air-laid nonwoven mats and blankets, specialty pulp and paper, cellulose chemicals, bio-composite feedstock for thermoplastic products, textiles and geotextiles. Short fibre is typically shorter and may be of lower purity than clean bast, accounting for approximately 0-5% by weight of the decortication system's product. The hurd fraction typically accounts for approximately 40-50% by weight of the decortication system's product, depending on, for example, the variety being grown, agronomic practices, retting extent, moisture content at time of processing, decortication method and power input to process, and desired hurd properties such as cleanliness and particle size distribution. Examples of applications for clean hurd are construction, industrial and consumer absorbents (for example, kitty litter), animal bedding, and as a feedstock for bio-composite products, bioenergy industries, and biochar (for example, horticultural grade). Hemp dust is typically mechanically detached during the decortication process and collected. Hemp dust typically accounts for approximately 25-40% by weight of the decortication system's product, depending on, for example, the variety being grown, agronomic practices, retting extent, moisture content at time of processing, decortication method and power input to process. As an example, hemp dust can be used as feedstock for biochar production (horticultural grade), as a filler for bio-composite thermoplastics, and energy pellets.
Exemplary objectives of the decortication process are to maximize the commercial value of the sum of the processed fractions while minimizing the processing costs, which are mostly related to power consumption, labour, operational breakdowns and problems which increase production downtime, as well as maintenance and repair costs, and feedstock price.
Typically, the decortication process begins with a preliminary evaluation of raw material grading and selection before it is processed. This can be done either in the field (i.e. before the material ships) or at the plant site (i.e. as the material is received). Conventionally, selection and purchase of straw suitable for decortication is left to experienced and knowledgeable straw specialists who travel from field to field, physically inspecting crops and performing rudimentary manual tests (both qualitative and quantitative) related to the apparent fibre quality, retting extent and moisture content. This specialist would then make a purchasing decision based on these tests along with knowledge of the processing facility's volume needs, the grower's price expectations, and the availability and quality of materials in other locations. Typically, the decortication process concludes when various processed commodity straw fractions have been processed from the hemp stalk and are placed either into bulk bins for future use (for example, compounding and extrusion into composite trim) or packaged for sale to others (for example, baled bast fibre for sale to nonwoven lines, hurd in 1 m3 totes for sale to insulated concrete manufacturers).
In an aspect, there is provided a method for optimizing flexible material handling for decorticated agricultural biomass, the method comprising: receiving decorticated components of the decorticated agricultural biomass, the decorticated components comprising bast fibre and hurd; analyzing one or more mechanical characteristics of the bast fibre; determining if the bast fibre is above a predetermined quality threshold based on the analyzed mechanical characteristics; where the bast fibre is above the predetermined threshold, directing the bast fibre for cutting, and cutting the bast fibre into one or more lengths; and where the bast fibre is below the predetermined threshold, directing the bast fibre for blending, and blending the bast fibre and hurd in determined proportions.
In a particular case of the method, the decorticated components further comprise dust, and wherein the dust in blended with the bast fibre and hurd where the bast fibre is below the predetermined threshold.
In another case of the method, the bast fibre is ground by milling the bast fibre prior to blending the bast fibre and hurd in determined proportions where the bast fibre is below the predetermined threshold.
In yet another case of the method, the quantity of hurd blended with the bast fibre, where the bast fibre is below the predetermined threshold, is based on the determined mechanical characteristics of the bast fibre to achieve desired mechanical characteristics of the combination.
In yet another case of the method, the bast fibre that is combined with the hurd comprises lowest quality of the bast fibre and comprises increasing quality of the bast fibre until the desired mechanical characteristics of the combination are achieved.
In yet another case of the method, the desired mechanical characteristics of the combination comprises desired values for at least one of elongation, flexural modulus, impact resistance, and hardness.
In yet another case of the method, the one or more mechanical characteristics of the bast fibre are determined using one or more of optical scanning techniques, Vis-NIR scanning, a hygrometer, and a densitometer.
In yet another case of the method, cutting the bast fibre into one or more lengths comprises cutting the bast fibre into one of a plurality of specified lengths based on the mechanical characteristics of the bast fibre.
In yet another case of the method, the predetermined quality threshold is determined using a trained machine learning model by minimizing a weighted cost function.
In yet another case of the method, the weighted cost function comprises weights associated with one or more of reducing inventory, reducing waste, maximizing strength, maximizing adhesion and maximizing throughput rate.
In another aspect, there is provided a system for optimizing flexible material handling for decorticated agricultural biomass, the system comprising: a material handler to receive decorticated components of the decorticated agricultural biomass, the decorticated components comprising bast fibre and hurd; an analyzer to analyze one or more mechanical characteristics of the bast fibre; and a material handler module, executed on one or more processors, to: determine if the bast fibre is above a predetermined quality threshold based on the analyzed mechanical characteristics; where the bast fibre is above the predetermined threshold, to direct the bast fibre for cutting, and cutting the bast fibre into one or more lengths; and where the bast fibre is below the predetermined threshold, directing the bast fibre for blending, and blending the bast fibre and hurd in determined proportions.
In a particular case of the system, the decorticated components further comprise dust, and wherein the dust in blended with the bast fibre and hurd where the bast fibre is below the predetermined threshold.
In another case of the system, the bast fibre is ground by milling the bast fibre prior to blending the bast fibre and hurd in determined proportions where the bast fibre is below the predetermined threshold.
In yet another case of the system, the quantity of hurd blended with the bast fibre, where the bast fibre is below the predetermined threshold, is based on the determined mechanical characteristics of the bast fibre to achieve desired mechanical characteristics of the combination.
In yet another case of the system, the bast fibre that is combined with the hurd comprises lowest quality of the bast fibre and comprises increasing quality of the bast fibre until the desired mechanical characteristics of the combination are achieved.
In yet another case of the system, the desired mechanical characteristics of the combination comprises desired values for at least one of elongation, flexural modulus, impact resistance, and hardness.
In yet another case of the system, the one or more mechanical characteristics of the bast fibre are determined using one or more of optical scanning techniques, Vis-NIR scanning, a hygrometer, and a densitometer.
In yet another case of the system, cutting the bast fibre into one or more lengths comprises cutting the bast fibre into one of a plurality of specified lengths based on the mechanical characteristics of the bast fibre.
In yet another case of the system, the system further comprising a data science module, executed on the one or more processors, and to determine the predetermined quality threshold using a trained machine learning model by minimizing a weighted cost function.
In yet another case of the system, the weighted cost function comprises weights associated with one or more of reducing inventory, reducing waste, maximizing strength, maximizing adhesion, and maximizing throughput rate.
These and other aspects are contemplated and described herein. It will be appreciated that the foregoing summary sets out representative aspects of systems and methods to assist skilled readers in understanding the following detailed description.
Preferred embodiments of the present disclosure will now be described, by way of example only, with reference to the attached Figures, wherein:
Before the subject matter of the present disclosure is described in further detail, it is to be understood that the invention is not limited to the particular embodiments described, but only by the scope of the claims appended hereto. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only, and is not intended to be limiting, since the scope of the present disclosure will be limited only by the appended claims.
For simplicity and clarity of illustration, where considered appropriate, reference numerals may be repeated among the Figures to indicate corresponding or analogous elements. In addition, numerous specific details are set forth in order to provide a thorough understanding of the embodiments described herein. However, it will be understood by those of ordinary skill in the art that the embodiments described herein may be practiced without these specific details. In other instances, well-known methods, procedures and components have not been described in detail so as not to obscure the embodiments herein. Also, the description is not to be considered as limiting the scope of the embodiments described herein.
Various terms used throughout the present disclosure may be read and understood as follows, unless the context indicates otherwise: “or” as used throughout is inclusive, as though written and/or; singular articles and pronouns as used throughout include their plural forms, and vice versa; similarly, gendered pronouns include their counterpart pronouns so that pronouns should not be understood as limiting anything described herein to use, implementation, performance, etc. by a single gender; “exemplary” should be understood as “illustrative” and “exemplifying” and not necessarily as “preferred” over other embodiments. Further definitions for terms may be set out herein; these may apply to prior and subsequent instances of those terms, as will be understood from a reading of the present disclosure/description.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Although any methods and materials similar or equivalent to those described herein can also be used in the practice or testing of the present invention, a limited number of the exemplary methods and materials are described herein.
It must be noted that as used herein and in the appended claims, the singular forms “a”, “an”, and “the” include plural referents unless the context clearly dictates otherwise.
Any module, unit, component, server, computer, terminal, engine, or device exemplified herein that executes instructions may include or otherwise have access to computer readable media such as storage media, computer storage media, or data storage devices (removable and non-removable) such as, for example, magnetic discs, optical disks, or tape. Computer storage media may include volatile and non-volatile, removable and non-removable media implemented in any method or technology for storage of information, such as computer readable instructions, data structures, program modules, or other data. Examples of computer storage media include RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile discs (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, network-accessible storage (such as cloud storage), or any other medium which can be used to store the information and which can be accessed by an application, module, or both. Any such computer storage media may be part of the device or accessible or connectable thereto. Further, unless the context clearly indicates otherwise, any processor or controller set out herein may be implemented as a singular processor or as a plurality of processors. The plurality of processors may be arrayed or distributed, and any processing function referred to herein may be carried out by one or by a plurality of processors, even though a single processor may be exemplified. Any method, application or module herein described may be implemented using computer readable/executable instructions that may be stored or otherwise held by such computer readable media and executed by the one or more processors.
The present disclosure relates generally to economic processing and optimization of biomass products, including biomass agricultural products. More particularly, the present disclosure relates to system, controller, and method for decortication processing.
Advantageously, the embodiments described herein can provide a low-cost, high-capacity, easily scalable, and flexible processing of biomass products, including biomass agricultural products. The embodiments described herein also advantageously allow for decortication processing with ‘Just-In-Time’ production of ordered goods, such that products can be primarily produced for order and not necessarily for inventory. In a particular case, the biomass agricultural products can be dicotyledonous plant products; for example, hemp, flax, kenaf, and jute. Further, advantageously, the embodiments described herein substantially reduce material handling and reduce production costs in a broad range of agricultural waste processing. Further, advantageously, the embodiments described herein allow for the optimization enabled by the vertical integration of hemp stalk processing to automatically direct raw hemp stalk components, on demand, towards production of higher value feedstock or finished products.
Advantageously, the embodiments described herein allow for spectrometry analysis, for example using visible and near-infrared spectrometer (Vis-NIR), of samples received from incoming bales to determine at least one of retting extent, chemical composition, and moisture content to enable a controller to predict a potential yield of at least one of bast, hurd, and dust when processing these bales. The spectral analysis may accordingly be used as a basis for material selection within a computer-integrated manufacturing (CIM) environment that is simultaneously optimizing production of each different product for different customers based on a pre-defined production schedule and delivery dates for open orders. As described herein the use of a spectrometry sensor may replace the subjective quality assessments of such materials at a stacking yard. The spectrometry sensors can be advantageously used by an artificial intelligence (AI) system (e.g., a machine learning paradigm) to select received bales to process into various fractions, and which can then be scanned by other “intelligent sensors” (including another spectrometry sensor at the bast fibre baling line). In most cases, as the sensed bales are moved throughout the decortication process, the bales can be at least virtually tracked, and their output characteristics and financial yields compared to input characteristics can be stored in a database.
In some cases, decortication can be performed in a stationary facility. In other cases, decortication can be performed in a mobile facility that can be moved into regions with substantial hemp crop production.
Decortication can be performed on properly-retted hemp straw bales having a moisture content in the range of 10-20% by weight; although some decorticators can be designed and capable of decorticating “green” material, shortly after processed in the field and without any retting. Material can be new or weathered, with older material likely being drier with higher spoilage, depending on how it was stored. The valuation of straw can depend on a variety of factors, for example, variety, maturity, crop year, moisture content, retting extent, spoilage, dockage (i.e. weeds, seeds and dirt). Higher prices are typically given to clean, weed-free, properly-retted straw bales from fibre hemp varieties at 12-15% moisture.
In some cases, a pulverizing line is fed clean hurd with a pre-determined amount of just enough bast added back in (preferably, short bast first, then more long bast if required) to achieve pre-determined mechanical property targets in the compounded pellets or extruded profiles.
At a breakage stage 54, bales are broken apart into a loose straw form, such that the straw can be fed to a decorticator. This stage also allows the straw pieces to be cut into lengths of a predetermined size, or utilizing the system, pieces can be cut to fulfill demand after the stalk has been decorticated. In an example, a cutting unit (for example, a digitally-controlled, self-adjusting guillotine mechanism) can be used to cut the bales (or portions thereof), followed by “opening” by an opening unit to loosen the straw and allow for pneumatic conveyance and continuous metering to the decorticator. In a particular case, round bales first need to be cut into similar lengths with a guillotine cutter. In another case, square bales may be entered directly into the opening unit or cut into similar lengths with the guillotine cutter. The opening unit opens the bale of stalk and ensures a relatively even flow of this stalk material towards the decorticating section. In some cases, at discharge of the bale opener, after a process sensor, a hygrometer, has been used to do a digital moisture analysis of the fibre being assessed, if the moisture level is too low, moisture can be added to the raw material based on historical or learned knowledge of the optimum moisture content parameters, and feedback from the process sensors; for example, the hygrometer, the decorticator motor load, relative humidity probe located in the dust collection system, and a Vis-NIR analyser located in the bast fibre baler. In some cases, moisture addition can be by automated steam or misting spray nozzles. In some cases, a separate “live-bottom” retention vessel or bin may be required for moisture content to fully equilibrate before decortication or cleaning.
In some cases, the system selects suitable bales from the available bale inventory based on the sample selected from each bale and the subsequent data collected by a Vis-NIR analysis and stored in its material database; for example, the system can utilize only one data source, such as the data retrieved from the preliminary Vis-NIR scan. In other cases, the system can make the selection based on multiple data sources, for example, based on one or more of the following criteria: preliminary Vis-NIR scan results, open orders, delivery dates, projected demand, product inventory levels, raw material inventory levels, raw material age, market prices for different products, and input costs (e.g. power price). Bales selected by the system are retrieved from a stacking yard (for example, each bale having a unique identifier code and the system automatically tracks where it is inventoried using global positioning systems (GPS)) and brought to the plant site for processing. In some cases, the bales can first have weather stabilization in the winter or after heavy rains. This selection process may be exclusively based on sensor data from when the bales are received. The sensor data may be obtained from a Vis-NIR, but any suitable sensor or combination of sensors can be used to provide information on chemical composition, retting extent and moisture content.
Bales can be selected for specific orders based on predictions made by the data science module 122 of fraction yields and product attributes, with the data collected when the bale is first received at the stacking yard as input features (for example, weight, moisture content, Vis-NIR scanning data) to a machine learning model. In this way, the scanned characteristics of the bale can be matched with optimal applications for the resulting decortication fractions. In this way, bales can be declined automatically if their moisture content is in or near the extremes, either too high or too low (for example, due to spoilage and increased fire risk during storage).
In some cases, the technical fibre length can be cut and thus reduced as specified for specific end users. Such end-user applications can include, for example, custom-cut technical fibres for cottonizing blending and spinning with other natural or synthetic fibres for woven fabrics; reinforcement fibres for specialty papers; compounding with thermoplastics and extrusion or pelletizing or injection molding as bio-composite materials; or the like. In a particular case, technical fibres can be deposited on an infeed belt of a Pierret cutter, or similar device, operating at a speed predetermined to deliver a specific cut length. Once cut in the first direction, fibres can then fall onto the infeed belt of the second cutter oriented perpendicular relative to the first cutter. In this way, fibres can be cut in both directions to a high level of uniformity as required by specific end-users. The frequency of the cutting head in relation to the speed of the infeed belt and hold-down rollers can be used to determine the cut length in each stage.
In some embodiments, as described herein, machine learning (ML) techniques can be used, with automated knowledge discovery in data (KDD), on sensor feedback data from a continuous flexural analyser on the extrusion line to adjust fibre cut length to control flexural strength. The output from the continuous flexural analyser can be used to adjust bast fibre content for mechanical property control.
At a decortication stage 56, the decorticator mechanically processes the opened straw to detach bast fibre bundles from hurd particles. Once mechanically detached, the components may be separated into bast and hurd by screening. The mechanical processing and separation can be repeated in order to produce fibres that are clean enough for further refining, or for specific, ordered end products. Dust can be removed via the “condensing units” that separate air from bast and hurd before screening. Preferably, bast fibres are processed that are pure enough to feed a refining line, and to produce clean hurd for sale. At block 68, the hurd can be stored at a storage unit for output or for further processing as described herein.
In an example of the decortication stage 56, moisture-equilibrated straw can be aspirated from the straw opening unit, or live bottom bin system, and fed into the decorticator to detach the hurd from the fibre, and produce a mixture of fibre, hurd and dust. At the exit of the decorticator, material is aspirated into an air separator that removes air and dust from the detached fibre and hurd. The fibre and hurd can then fall into a fibre separator (for example, a screen) where a large part of the hurd can be removed.
In some cases, the bast fibre exiting the fibre separator can then be re-processed in a second decorticator, air separator and fibre separator to produce cleaner bast fibre at approximately 80% to 85% purity relative to the hurd which may be left attached to the bast after the first round of decortication. This level of purity is likely sufficiently clean to feed a refining line. In some cases, the bast fibre can be directed to a baler, or can be further processed in the refining line to a higher level of purity. In some cases, at block 58, dust is extracted from the air handling system via a dust cleaner. Although the level of purity achieved may reach 80-85%, it is contemplated that a level of purity as low as 40% to 50% may be achieved for such reasons as cost-reduction, or higher than 85% if desired or required by downstream customers.
In some cases, the hurd stream can be further classified by particle size for certain markets or industries; for example, insulated concrete, large animal bedding, small animal bedding, or the like. In some cases, short bast fibres (for example, under 2″ in length) may also be recovered from the hurd stream for reintroduction into the bast product, separate packaging for sale, or further mechanical processing into reinforcement fibres for compounding and extrusion as bio-composite materials. At block 66, the short fibre can be stored at a storage unit for output or for further processing as described herein. The reintroduction of bast fibres to the hurd may result in increased mechanical performance of composites (such as bio-composite materials).
In some cases, the decorticator can be fitted with an inverter duty motor and variable frequency drive (VFD). Information, such as power consumption, from the VFD can be used in conjunction with relative humidity measurements taken in the air handling system and bast purity measurements taken at the fibre baler using the spectrometry sensor to adjust operating conditions such as decorticator RPM, moisture addition rates, and straw feed rates.
In some cases, one or more refining stages (for example, each consisting of a fine opener and a screening device such as a step-cleaner) can be used for further purification of decorticated bast fibre from, for example, 80-85% purity to, for example, 95-98% purity; with associated hurd cleaning and short fibre recovery. Advantageously, this can improve bast fibre purity, minimize fibre damage or shortening, clean hurd, and recover short fibre. In this case, bast fibre containing 15-20% hurd is opened using two fine openers. The fibre can then be aspirated, fed directly to two step cleaners, and fed to a finishing cleaner. At the end of the refiner, the bast fibre can contain less than 5% hurd. This can be referred to as “technical fibre”, and can be used in various applications; for example, air-laid nonwovens like insulation batt, geotextiles, fibre mats for resin transfer molding, super-capacitors, or the like. These fibres can also be cross-cut to specific lengths using a Pierret cutter, or similar device, for use in, for example, compounding and extrusion of bio-composite materials, reinforcement fibres in specialty papers, or the like. Hurd removed from the bast can be cleaned of short fibre and blended with clean hurd from the decortication line or packaged separately for sale. In some cases, the particle size of this hurd may be smaller than that from decortication due to the extra mechanical processing it can receive. In some cases, at block 60, dust is extracted from the refiner via a dust collection unit.
At a bast baling stage 62, a baler can be used to densify technical fibres into bales of, preferably, consistent weight and moisture content. In some cases, an analysis of the purity of the fibres can also be undertaken with, for example, a spectroscopy sensor, such as a Vis-NIR unit, associated with the baler. The spectroscopy sensor can measure moisture content for direct feedback control to a humidification step. During the humidification step, moisture can be added to a discharge of a unit feeding a decortication stage, with provision to add more moisture at a feed to a refining stage. Generally, most of the moisture can be added prior to decortication.
The spectroscopy sensor (for example, Vis-NIR sensor) can measure bast fibre purity with every stroke of a bale press hydraulic unit; the bale press hydraulic unit generally comprising a hydraulic ram that compresses bast fibre in the baler. The spectroscopy sensor can provide multiple analyses throughout each bale as it is being made; this can provide enhanced quality control as compared to scanning completed bales. It should be appreciated that in other cases, another sensor or combination of sensors that is capable of determining moisture content (for feedback moisture control) and/or measuring purity (for control of decortication) can be used. Fibre baling is particularly used to package and brand the fibre product, and in some cases, perform quality control and process control feedback.
As shown in the exemplary embodiment of a bale press hydraulic unit 1100 of
In some cases, a “fineness” sensor can be integrated into the bale face 1104 that would measure compressed air flow from a nozzle into the fibre at a specific hydraulic ram pressure. Finer fibers restrict air flow more than coarse fibres, thus giving feedback regarding the value of fibres in either textiles or air-laid nonwovens.
At the bast baling stage 62, the weight of each bale can be measured, along with its moisture content and bast fibre purity (for example, by a spectroscopy sensor such as a Vis-NIR analyzer). In some cases, bale weight can be corrected for moisture content to ensure a consistent mass of product. In some cases, product sensor information can be fed back into a control system for process tuning, and also compiled with the corresponding raw material properties and plant operating conditions for real-time product costing as well as process and/or value optimization. In general, dry weight of the bale is determined as ‘total bale weight’*(100%−% moisture). In this way, fibre can be determined (and in some cases, sold) on a dry basis or at a standard moisture content.
In some cases, using ML and/or KDD, bale selection and processing setup can be automated for specific production runs based on the yields of different fractions, product attributes, or overall economic performance. Advantageously, this can minimize production of “off-grade” materials between different production runs and support new product development through extrapolation of large quantities of operating data to “expand” the range of possible properties and operating conditions.
As the system learns from gathering data from product testing and process operation, it can use KDD to link product attributes and processing conditions to raw material attributes, such as from scans received when the material was received. In this way, a machine learning model can be used to automatically receive orders and select input materials (for example, from the yard) that are optimized to meet the specifications of the orders with respect to, for example, quantity, quality, and the like, as well as set-up the processing conditions of the decortication system. At a bast cutting stage 64, the bast coming from the decortication stage 64 can be cut into a desired or specified length by a cutting unit. In some cases, the cut bast can be sent to a blending stage 70. In other cases, the cut bast can be sent to the bast baling stage 62.
At a hurd pulverizing stage 72, hurd and chopped bast can be ground and sieved prior to compounding with thermoplastics and extrusion into either profiles (for example, for sale to end-users) or with the end-users own master-batch polymer pellets (for example, for sale to other extrusion facilities). Pulverizing can be done to generate consistent fine particles for compounding. In some cases, pulverizing can be undertaken with a hammermill. In some cases, a cyclone can be used for separation of air from pulverized hurd and bast. Hurd particles and a metered amount of bast fibre can be blended together before being introduced to the hammermill with a slotted or perforated screen; the size can be determined by processing extent in a final product. The proportions of hurd particles and bast fibre to be blended together can be chosen to meet mechanical property targets for composites, as described herein. Particles and fibres can be mechanically ground until they pass through the screen, then they can be aspirated to a cyclone and rotary airlock for air removal. The pulverized material can then be dropped into a metering bin and volumetric (or mass-based) metering screw for consistent delivery to a compounder. In some cases, aspiration can be provided by the decortication plant's air handling system, which is preferably sized appropriately for this additional air volume; this approach can be beneficial because any dust carryover can be captured by a baghouse.
At a compounding stage 74, pulverized materials can be combined with liquid thermoplastic polymers or dispersing agents in a pellet or extruded shape of uniform appearance and composition. In some cases, hemp dust can be introduced into the extrusion with the other components. Compounding with varying amounts of polymer pellets can produce compounded pellets for any kind of thermoplastic production, including extrusion, compression molding, pultrusion, or injection molding or can produce extruded profiles (e.g. trim) for sale to others. Pulverized hurd and short fibre can be fed into a compounding extruder, along with the appropriate amount and type of thermoplastic polymers and additives (such as a thermoplastic dispersing agent, if needed), where they are heated and blended together to produce a uniform mixture. Examples of suitable thermoplastic polymers include PVC, polypropylene, polyethylene, polyhydroxybutyric acid, polyvinyl alcohol, and polylactic acid. If master batch pellets are the target product, the mixture can contain an exact amount of polymer required to achieve the desired proportionate mix with the amount of added fibre needed to achieve the required mechanical characteristics of the finished product. If the target compounded product is to achieve a very dense fibre pellet which will be mixed with the minimum amount of polymer needed to just keep the pellet together during shipping, and enough to stop moisture from re-entering the compounded pellet, these pellets will be shipped to the end-user where they will be compounded with the specific polymer, being utilized for the finished end-product, by the end-user. In either case, the pellet can be extruded through a pelleting die and onto a cooling conveyor before bagging. If extruded profiles are the target product, the mixture can be transferred in its molten state to multiple profile extruders for subsequent formation of the finished cross-section. This can be followed by cooling and cutting to length. In some cases, automated testing of extruded profiles for chemical composition (via for example a Vis-NIR analyzer) and flexural strength (via for example mechanical deflection testing) can provide quality control information and feedback to the decortication system to adjust the quantity and/or length of chopped bast fibres being introduced to the compounding system. In some cases, control conditions can be determined by inputting such test data, in some cases coupled with operating conditions and raw material attributes, to ML and KDD techniques.
In some embodiments, the target product of compounding stage 62 may be a durable, easily dispersable, fibre composite pellet having a high fibre content. A durable, easily dispersable pellet at 95% fibre content or higher can then be blended with natural polymers. Where master batch pellets are the target product, it may be desirable to have fibre content of 95% to 100%; however, it will be appreciated that downstream processors of these master batch pellets may be satisfied with any other percentage of fibre content, such as 50% (or higher).
Referring now to
The controller 100 can be executed on a suitable computing device; for example, a desktop computer, a laptop computer, a microcontroller, a programmable logic controller, a field-programmable gate array, a server, or the like.
Additional stored data can be stored in a database 116. In an embodiment, the database 116 can store various data related to decortication processing. As an example, such data can include any one or more of: operating data, customer related data, and operations management data.
Operating data can include, for example:
Customer related data can include, for example:
Operations management data can include, for example:
In an embodiment, as described in more detail in the following, the controller 100 includes various modules 120; including a data science module 122 and a control module 124. In some cases, some or all of the various modules 120 can be combined, be executed remotely on a server-side device, or be executed on other components of system 200 (as described below). In some cases, some or all of the various modules 120 can be executed remotely on a server-side CIM device, such as CIM server 640 (as described in
In some cases, the analyzers 202 can incorporate one or more sensors. In further embodiments, some or all of the components of the system 200 may communicate directly with each other.
Referring now to
At block 302, the input interface 106 is used to receive associative information relating to one or more input units of hemp. In some cases, the input unit can be input bales, input lots, input truckloads, or the like. For the purposes of this disclosure, input units will be described as input bales. These are typically units of hemp brought to a processing facility. The associative information can include, for example, load weights, field location, crop variety/year, names of those involved in growing, date of seed planting, date of harvest, if crop was irrigated or dryland, how much it was watered and how often, fertilizers used (how much and how often), swathing used or a combine for harvesting, baling, loading, or trucking of the bales, where the bales are stored, a bale number identifying each specific bale and the supplier of the bale (directing the system to all required and appropriate data about that farmer and their crop), or the like.
At block 304, one or more of the analyzers 202 analyze the input bales. In a particular case, such analysis includes weighing, sampling and/or scanning. In some cases, the analyzer 202 can be a Vis-NIR spectrometer. In some cases, the analyzer 202 can be a densitometer, for example, to locate internal rocks or metal pieces. In some cases, the analyzer 202 can be a hygrometer to measure moisture content. In some cases, the analyzer 202 can be an optical scanner to measure, for example, the diameter of the stalk being processed. The analyzer 202 can communicate the results of the analysis to the controller 100 to be linked with each particular bale in the database 116. The analysis data can be communicated to the data science module 122 as described herein.
In some cases, the analysis of input bales includes determining chemical composition of the bales; for example, determining a percentage of bast, a percentage of hurd, and/or a percentage of other components. In some cases, the analysis of input bales includes determining retting extent; for example, determining power consumption and/or potential bast purity. In some cases, the analysis of input bales includes determining moisture content; for example, determining potential power consumption and/or potential fibre length. In some cases, the bale may be sorted and segregated based on its qualities determined above for later selection according to a particular use.
For example, when the control module 124 calls for textile grade bast fibres at 98% purity and 1.25″ fibre length, the database 116 is accessed by the control module 124, or another element of the controller 100, to see what materials, for example those in the stacking yard, are best suited to make this product and what hurd product(s) are best made at the same time to maximize system profitability, and how best to configure all processing systems to minimize off-spec material produced in grade transitions. In an example, textile-grade bast fibres should generally be consistently retted, which may preclude some applications for the corresponding hurd (for example, horse bedding). In this way, the control module 124 cannot produce such other applications while running, for example, textiles. In another example, if a large textile order is received that needs to be filled quickly, consistently-retted bales having a high cellulose content (signifying high bast content) may be preferred for throughput and efficiency as well as quality control reasons.
At block 306, the cutting unit 204 may cut the bales, for example, as described in more detail herein with respect to the breakage stage 54.
At block 308, the opening unit 206 opens the bales, for example, as described in more detail herein with respect to the breakage stage 54.
At block 310, the decorticator 208 performs decortication on the opened bales, for example, as described in more detail herein with respect to the decortication stage 56.
At block 311, in some cases, bulk products can be outputted by the controller 100. For example, one or more of hurd, decortication dust, baled refining dust, or baled short fibre can be outputted.
At block 330, the baler 210 densifies technical fibres into bales, for example, as described in more detail herein with respect to the fibre baling stage 58. In some cases, one or more of the analyzers 202 can analyze the retting extent, chemical composition and/or moisture of the bales; for example, using a Vis-NIR analyzer. Such analysis can be fed into the data science module 122 for feedback process control and/or straw selection for product-specific use. In some cases, the collection of all bale data is carried out so that the data may be used to control the selection and processing of individual bales to achieve specific product outcomes. The baled blast can then be cleaned as desired at block 332.
In some cases, the blending unit 216 can blend hurd, short fibre and cut bast into proportions as determined by the control module 124 as described herein.
At block 312, bast outputted from the decortication at 310 can be cut to a desired or specified length, as described herein. The cut bast, along with the other outputs of the decortication at block 310, hurd, refining dust, and short fibre, are delivered to the pulverizer 212 at block 314.
At block 314, the pulverizer 212 grounds and sieves hurd and/or bast, for example, as described in more detail herein with respect to the pulverizing stage 60. In some cases, one or more of the analyzers 202 can analyze the particle size distribution of the hurd; for example, using an image analyzer. Such analysis can be fed into the data science module 122 for feedback process control and/or quality assurance/quality control. In some cases, one or more of the analyzers 202 can analyze the chemical composition of the pulverized hurd and/or bast; for example, using a Vis-NIR analyzer for the hurd or a continuous flexural analyser for the bast. Such analysis can be fed into the data science module 122 for feedback control of the decortication process to reach target purity, for feed-forward control of cutting unit 204 to reach target flexural strength, and/or for straw selection for a product-specific use.
Generally, flexural strength will be proportional to bast length and bast content. Lowest possible strength will be when only pulverized dust output, created in the decortication process, and higher strength than the dust, but lower then the bast occurs when hurd is present. Full product specifications must be known in order to optimize hurd geometry, which impacts bond area and bulk density.
At block 316, the compounder 214 combines the pulverized materials with thermoplastic polymers into a resultant product comprising a pellet or extruded shape of uniform appearance and composition, for example, as described in more detail herein with respect to the compounding stage 62. In some cases, one or more of the analyzers 202 can analyze the flexural strength of extruded material; for example, using a flexural tester. Such analysis can be fed into the data science module 122 for feedback control of cutting unit 204 to reach target flexural strength and/or quality assurance/quality control. In an embodiment, reaching target flexural strength or quality control can achieved by increasing bast content or increasing bast length.
In some cases, at block 318, the data science module 122 receives log information from the input interface 106 inputted by an operator or user.
At block 320, the data science module 122 builds a machine learning model using training data comprising data from the one or more of the analyzers 202 and/or input interface 106. The machine learning model builds interrelationships between characteristics of the resultant product and characteristics of the input bale and/or operating conditions. At block 322, once the machine learning model is sufficiently mature, having received sufficient training data, the control module 124 can use the machine learning model to predict characteristics of the components or of the resultant product based on characteristics of the corresponding input bale and operating conditions. In some cases, such characteristics can include the expected output volume of the resultant product, quality of the resultant product, and time required to produce the resultant product.
In a particular case, the machine learning model can be developed using supervised learning techniques. The supervised learning techniques generate a model by learning relationships and dependencies between the outputs and input features from example data sets. The datasets used for training a supervised machine learning model comprise of labeled examples where input and desired outputs are known in advance. In some cases, supervised learning techniques that be used include, for example, Nearest Neighbor, Naive Bayes, Decision Trees, Linear Regression, Support Vector Machines (SVM), Neural Networks.
In another case, the machine learning model can be developed using unsupervised learning techniques. The unsupervised models can be trained with unlabeled data. These techniques detect patterns to compute a measure of similarity (or dissimilarity) with other data points and summarize or group the data points that may provide meaningful insight into the data. In some cases, unsupervised learning techniques that be used, for example, include k-means clustering, hierarchical clustering, generative adversarial networks, and autoencoders.
In other cases, the machine learning model can be developed using semi-supervised learning techniques. The semi-supervised techniques fall in between the supervised and unsupervised techniques. The input data for semi-supervised learning is mainly unlabeled but comprises of a small exemplar labeled set which serves as “seed” for guiding and growing clusters. In some cases, semi-supervised learning techniques that be used, for example, include generative models, transductive SVM, and graph-based methods.
In some cases, the machine learning model can be a developed using reinforcement learning techniques. Reinforcement learning techniques are generally neither supervised or unsupervised. Reinforcement learning techniques attempt to determine an ideal response within a specific context in order to maximise its performance by retro-feeding the so-far-learned model in order to further improve by learning from its mistakes. Reinforcement learning does not rely strictly on set of labeled data set for learning. Rather, it relies on being able to monitor the response of the actions taken, and measure against a definition of a reward. In that regard, it may be treated as learning via exploring. In some cases, semi-supervised learning techniques that be used, for example, include dynamic programming, Q-learning, temporal difference and deep adversarial networks.
In a particular case, the machine learning model can be a regression-type model, for example, Linear Regression, Logistic Regression, Polynomial Regression, Stepwise Regression, Ridge Regression, Lasso Regression, ElasticNet Regression, Multi-layer perceptron based Regression, Partial least square Regression, Regression Trees, and Support vector machine Regression. The regression model selected may depend on the number of independent variables, shape of the regression line and the type of dependent variable. Initial training data for the model can be comprised of laboratory tests done during an engineering phase of development of the controller 100. In this case, training data can be generated over the start-up/commissioning period, and in some cases, over an initial period (for example, 3 months) of operation by concurrent field sensor measurements and laboratory analysis for output attributes described herein. In this way, time-stamped lab data from manual samples obtained at the sensor location can be manually entered into the database as a “reference”, “ground-truth”, or “calibration” point.
In some cases, an ensemble of machine learning models may be used, such as multiple regression techniques or other techniques running simultaneously. The ensemble may be homogeneous or heterogeneous. A homogeneous ensemble comprises a plurality of classifiers of the same machine learning type (for example, multiple support vector machines). Each model in a homogeneous ensemble may have different parameter values and may be trained using a distinct subset of the samples in the training set. A heterogeneous ensemble comprises a plurality of models belonging to a variety of machine learning techniques; for example, a regression model, a K-means clustering, and an SVM. Models in heterogeneous ensembles may be trained on the same training data or on distinct subsets of the training data. If a multiplicity of a machine learning techniques exists in a heterogeneous ensemble, each instance of the multiplicity may be trained on some samples unique only to that instance.
At block 324, the control module 124 can use the machine learning model to adjust one or more aspects of the other components of the system 200 to achieve a desired resultant product. As an example, adjusting a chop length of the cutting unit 204 or adjusting decorticator speed of the decorticator 208. In a particular case, if flexural strength must increase but bast inventory is not sufficient (for example, because materials have lower bast content), then chop length will be increased. If baler spectrometry sensor indicates low purity but moisture content is on target, decorticator speed can be increased to improve cleanliness.
In this exemplary method, the hurd cleaner 421 receives hurd of various lengths and cleanliness. The hurd cleaner 421 cleans the hurd to remove short bast fibres. The hurd cleaner 421 sends the clean hurd to the pulverizer 431 and any short bast fibres to a short bast fibre bin 422. The short bast fibre bin 422 sends a first portion, for example most of the short bast fibres, to a baler 424 and a second portion, for example the remainder of the short bast fibres, to be re-blended 423 back into the hurd to improve mechanical performance. In some cases, the baler 424 bales the short bast fibres and sends them to the warehouse.
In this exemplary method, the pulverizer 431 receives the clean hurd, the re-blended long bast fibres, and the re-blended short bast fibres (together “three materials”). The pulverizer 431 pulverizes the three materials together into a powder-like form. The pulverizer 431 pelletizes the powders using a dispersing agent 441. The dispersing agent 441 may be liquid and compatible with the natural polymer that is used in subsequent compounding. The pulverizer 431 sends the pellets to a pellet mill 451. The pellet mill 451 sends the pellets to pellet bulk storage/bagging.
During the decortication process, there may be compounding and profile extrusion of trim, baseboards, casings, moldings, and other such products. There may also be continuous mechanical testing of extruded profiles, with feedback that controls the addition of short and long bast fibres.
In an example of mechanical testing, a continuous dynamic flexural tester 1200, as shown in
Quality information is tagged onto the ID number of the bale for tracking/traceability throughout the inventory and production process. Tagging can be, for example, a bar code, an RFID chip, or the like. In an example of the control module 124, a CIM server 540 can be used to determine when certain products are to be made. The CIM server 540 can then scan the inventory system to “find” which bales are “predicted” to be most apt to provide the most appropriate quality of stalk for the expected outputs. One or more artificial intelligence (AI) modules 550, each part of the data science module 122, can be used to generate the prediction. In some cases, the CIM server 540 can be its own module or component of the controller 100. Decortication sensors 571 monitor operating metrics of decortication equipment 561 (for example, pumps, motors, drives, or the like) to monitor production and material handling equipment in order to interrupt any problems in the decortication equipment 561 which could cause a bottleneck, jam the line, or cause a production line shutdown. The decortication sensors 571 may be connected to the CIM server 540 via a Bluetooth (or other suitable) connection. The decortication equipment 561 can produce bast 580 and hurd 590.
Turning to
In some cases, each of the above stages of the decorticator 208 can include its own analyzer 202. In some cases, along with each stage, each decorticated component can also include a dedicated analyzer 202.
At block 1504, one of the analyzers 202, located after the decorticator 208, analyzes at least one of the decorticated components. For example, the analysis can include using a Vis-NIR and/or other optical scanning techniques to determine the quality of the bast fibres. In further cases, other analyzer or sensor can be used, as described herein; for example, a hygrometer, a densitometer, or the like. In an example, the analyzer 202 can have a “moving” scanning capability which can provide improved material handling. In some cases, the grading of quality can be based on one or more criteria of analysis; for example, one or more of color of the bast fibre, shape of the bast fibre, cellulose content of the bast fibre, and the like. In some cases, only a sample of a batch of bast fibre coming from the decorticator needs to be sampled to determine the quality of that batch. For example, the bast fibre can be balled or bunched together; and in such cases, the analyzer 202 can sample quality of a representative quantity of bast fibre from the bunch.
At block 1506, the material handler module 130 directs a specified portion of the bast to further processing based on the quality of the bast fibre. In a particular case, the bast can have binary grading of either above a predetermined quality or less-than the predetermined quality. In an example, the grading can be based on grading standards from an international organization such as ASTM International (for example, Active Standard ASTM D7879, Active Standard ASTM D5867, or other standards determined by ASTM Subcommittee D13.11 on Cotton Fibers). The predetermined quality threshold can also be based on the requirements of a desired use of the end-product. In some cases, the bast fibre quality for the predetermined quality threshold can be multi-factorial, and/or exist on a continuum of many different grades, each of which having its own predetermined quality. For example, the predetermined quality can include acceptable ranges for fibre length, fibre purity, fineness, and the like. In this way, if one of the factors falls below the acceptable range for a given grade, one or more of the other factors can still be in the acceptable range and thus meet the predetermined quality threshold.
At block 1508, if the bast receives a premium quality, the bast can be processed; in some cases, such that the higher quality bast fibre can itself be outputted; for example to be cut and/or baled. As described herein, at block 1522, the bast fibre can be cut at specified lengths according to the quality of the fibre and baled as described herein. In some non-limiting examples, the bast fibres can be cut to 5-7 mm for pulp and paper applications, 20-30 mm for spunlace applications, 35-50 mm for applications involving blending with cotton (knits and wovens), 75-150 mm for applications involving airlaid nonwovens (erosion control mats, automotive panels), and 150 mm or longer for applications involving linen-like wovens, In some cases, the premium quality bast fibre can be diverted for pulverizing to produce bio-pellets with very high composite strength. Particularly, where the analyzer 202 is sensing that the pellets are not meeting the quality requirements for the pellets, as described herein; and thus, require the greater strength of the higher-quality bast fibres.
In some cases, as described herein, the input material can be selected based on its anticipated product quality, which in some cases can be based on previous decortication outputs. In some cases, this anticipated product quality can be associated with the predetermined quality requirement. If the decorticated bast fibres do not meet the anticipated quality predicted by the system (i.e., determined that the bast fibre is below the predetermined quality), then it can downgraded and processed as described herein. In some cases, the system can use this missed prediction as feedback to make changes to processing conditions in an effort to get on-grade, or to tune its models for predicting quality of future decorticated products.
At block 1510, if the material handler module 130 determines that the quality of the bast fibre is below the predetermined quality, the bast fibre can be directed for bio-pellet processing, unless as described herein. In an embodiment, the bio-pellet processing can include combining the lower-quality bast fibre with hurd from the decortication. As described herein, the bio-pellet processing can also include combining dust and/or polymer with the bast fibre and hurd. In a particular case, at block 1512, the hurd can be processed prior to combination with the bast fibres; for example, ground into a powder-like consistency by milling. In some cases, at block 1514, the bast fibre can be further processed before combination; for example, ground into a powder-like consistency by milling. In some cases, at block 1516, the dust can be further processed before combination; for example, purified or cleansed by a purifier. In some cases, blended materials can be further processed after combination in block 1518; for example, ground into a powder-like consistency by milling.
In some cases, only a selected quantity of the lower-quality bast fibres can be combined with the hurd. As hurd fibres are generally shorter and weaker than bast fibres, there is generally a limit to their technical performance in composites. In this way, the bast fibres, even the lower quality bast fibres, can be used to reinforce the composite pellets. In some embodiments, the material handler module 130 directs the quantity of hurd based on the determined quality of the bast fibre detected by the analyzer 202 and the quantity of overall quantity required. For example, where the bast fibre directed for combination has a low quality, more bast is added until certain mechanical characteristics are reached with hurd added to achieve an overall quantity of the pellet. The mechanical characteristics can be, for example, to increase the structural properties such as flexural strength of the pellet. As bast fibres have higher strength characteristics than hurd, combining a higher ratio of bast in the bio-pellet combination can increase the relative strength of the bio-pellet where a higher strength pellet is desired; however, accordingly the bast is more financially valuable than the other decortication by-products, so an increase in strength unnecessarily can increase the cost of the bio-pellet unnecessarily. The strength characteristics can include, for example, elongation, flexural modulus, impact resistance, hardness, and the like. Generally, decortication dust from the decorticator 208 can be included in the combination as a filler; but, in some cases, dust can be beneficial for interaction and/or dispersion of the polymer in the bio-pellet.
At block 1518, the blending unit 216 can combine the hurd and bast fibres (in some cases with the dust and/or polymer) to create the bio-pellets. In a particular case, the blending can be based on an order of increasing value. Where there is a composite quality specification in accordance with the metrics described herein, the bast fibre products can be added in order of increasing value until such quality specification is reached. Whereby refining dust from the decorticator is the least valuable bast fibre product, the short bast fibre is next least valuable, and the long bast fibres are the most valuable. In this way, the more valuable bast fibre products are not unnecessarily used for producing the composites. In an example, where the bast fibre quality is low, the ratio in the bio-pellets can be 55% hurd, 20% dust, and 25% bast fibre. Generally, bio-pellet performance can be a function of fibre content and fibre length distribution. In this way, generally, bio-pellets with longer bast are tougher, stiffer, and require less fibre for technical performance; bio-pellets with more bast are tougher, stiffer, and more expensive; bio-pellets with more hurd are lighter, weaker, and less expensive; bio-pellets with more polymer are heavier, stronger, and more expensive. Polymer can be a thermoplastic binding agent to keep pellets intact, and a lubricating agent for the fibres in the pellet mill. In this way, polymer can produce lower density pellets that are more easily dispersed in downstream extruders and/or injection molders. The polymers can be, for example, (1) thermosetting or thermoplastic, (2) water-borne or solvent-borne, (3) bioderived or petroleum-derived, and (4) durable or biodegradable. Examples of polymers include polyethylene, polyesters, polyurethanes, starches, and the like.
At block 1520, where a substantial quantity of the bast is determined to be high quality, the material handler module can direct the unused hurd for further processing not destined for bio-pellets. For example, such hurd can be compounded and outputted. In some cases, the compounded hurd can have advantageous applications, such as organic kitty litter.
In some cases, the predetermined bast fibre quality threshold can be manually provided by a user or by an administrator or system designer. In other cases, the predetermined bast fibre quality threshold can be determined by the data science module 122 using a machine learning technique (such as those described herein). In some cases, the data science module 122 can employ a trained machine learning model to determine the predetermined threshold by minimizing a weighted cost function. The model can be trained using a plurality of simulations or using a variety of incoming bales with actual training events on the system. In an example, the weighted cost function can weigh various parameters including one or more of:
As described herein, in some cases, the system can take into account raw material data (agronomic, harvesting, and) that is linked to each bale or shipment of input bales such that end products of such input bales can be tracked. In some cases, operating data can be routinely obtained by a distributed control system (DCS) in order to link raw material data with material properties and operating conditions, thereby enabling examination of such correlations; such as inputs to the machine learning model.
In some cases, the extremes of the predetermined threshold can be used; for example, all bast fibre to be processed into bio-pellets or all bast fibre to be processed by the cutting unit 204.
In some cases, the weighting of the parameters can be at least partially based on input received from a user or by an administrator; for example, based on meeting targets for one or more specific customers of the bast fibres and/or bio-pellets. In some cases, parameters for the weighted cost function can also include scheduling parameters. For example, the input interface 106 can receive a desired output and desired time for output of the produced decortication outputs of bast fibres and/or bio-pellets. The data science module 122 can input the scheduling parameters into the machine learning model such that the data science module 122 can use the machine learning model to weigh the other weighted parameters to meet the scheduling parameters. In this way, advantageously, the system 1400 can deliver continuously amenable just-in-time delivery of the decorticated products.
In some cases, the data science module 122 can also use the machine learning model to determine the ratio of decorticated components in the bio-pellets by setting the ratio as another output parameter of the machine learning model. In this way, the material handler 220 can receive outputs of the machine learning model from the data science module 122 and direct bast fibres and specific quantities of hurd, dust, and/or polymer to the blending unit 216 for combination into bio-pellets. In further cases, the data science module 122 can use a separate machine learning model to determine the ratio of decorticated components in the bio-pellets using at least some of the weighted cost function parameters as inputs.
In some cases, the data science module 122 can also use the machine learning model to determine the length of bast fibres cut by the cutting unit 204, by setting the cutting length as another output parameter of the machine learning model. In this way, the length of the bast fibres can incorporate the weighted cost parameters.
Advantageously, the system 1400 can provide continuous production of both bast fibre and other products of the decorticated products, such as bio-pellets. This in contrast to other approaches to decortication systems that typically have fixed production and typically do not process the non-bast decortication products. In this way, the present embodiments provide technical improvements that can generate greater value by using fibres that would otherwise create a lesser quality product or would be discarded.
A bast scanning and sampling system 532 can scan the bast 580 before being delivered to the automated cutter 562. Cutter sensors 572 monitor operating metrics of the automated cutter 562. The cutter sensors 572 may be connected to the CIM server 540 via a Bluetooth (or other suitable) connection. The second sensor 572 can be, for example, weight sensors, level indicators, motor loads, flexural tester (E/IM), temperature sensors, pressure sensors, or any process sensor. The automated cutter 562 delivers the cut bast to an automated baler 563, which may be, more generally, a bast bundling unit. The baler 562 can be, for example, a cardboard baler, a vertical or horizontal bagging line, or the like. A bale scanning system 533 can scan the contents of a bale of bast. The bale scanning system 533 may utilize a Vis-NIR scanner embedded in the compression head of the automated baler 563: every time it pushes down to compress the bundle, the Vis-NIR unit in the head is triggered and scans the top row of bast on the bundle. In further cases, instead of the VIS-NIR unit in the head, other scanning approaches can be used; for example, a Vis-NIR on an infeed chute before compression, image capture and analysis, fluorescence analysis, or the like. The bast scanning and sampling system 532 and the bale scanning system 533 each (or both) can include an AI module of its own or be connected to an external AI module, such as the data science module 122. The CIM server 540 communicates with the decortication equipment 561 and receives scanning data from the bast scanning and sampling system 532 and/or the baling scanning system 533.
In some embodiments, the scanning systems 531, 532, 533 can be, for example, a 3D scanning system (e.g., to determine internal fibre characteristics); an OTC scanner; or an optical scanner (e.g., to determine shape and size). Other suitable scanning systems that perform analogous or complementary functions may also be used, for example, those that determine pellet durability. In some embodiments, the sensors can be a densitometer (e.g., to measure density through a bale or a stalk); a hygrometer (e.g., to measure moisture content in a bale or a stalk); or a photon sensor. Other suitable sensors that perform analogous or complementary functions may also be used.
In some cases, the densitometer can be used to scan for solid contaminants, e.g., stones or other contaminants such as broken pieces of metal or plastic. These contaminants could damage or destroy any production equipment and thus such scan can avoid equipment repairs and downtime. Generally, plastic can be a significant contaminant of hemp fiber being used in textiles. In this way, individual bales can be tracked per farmer to identify problem suppliers.
In some cases, a hygrometer can be used to monitor and check moisture content of the bast fibre in order to make adjustments during processing; for example, adding water via a sprinkler where the moisture is too low to improve quality and output value. If the moisture is too high, the particular bale can be side-lined or the system can use historical data to determine adjustments to throughput rate to accommodate the surplus moisture or adjust processing speed/throughput. In some cases, hygrometers can be used at other locations of the system to check the moisture level.
The sensor data can include sensors associated with motors, pumps, circuits, and valves in the system. This allows the CIM server 640 to monitor for any momentary load stress(es) anywhere in the system in real-time. For decortication, these stresses often take place when the stalk is broken down into its fibres. However, when the fibres are emerging from the decorticator line, it is not unusual that some loosened bast fibre(s) wrap(s) around a conveyor or inside a piece of equipment. The fibre itself is so strong that it generally will not break; thus, it is usually the equipment that breaks. The sensors, monitored in real-time, can react to changes in their readings because sudden changes can be an indicator of a potential production problem looming, including bottlenecks or machine jamming. In some cases, the sensor data can be used by the machine learning model, as described herein, to learn over time to identify and predict which changes to sensor readings are potentially problematic.
The CIM server 640 has one or more files—these files are processed to automate functions of the decortication system. The files may include, for example, a farmer file 641, an equipment file 642, a client file 643, a production file 644, an open orders file 645, and a trends/statistics file 646. The farmer file 641 keeps a running control on quality of stalk by tracking each bale processed for yield and contaminants; and provides farmer supply traceability. The equipment file 642 monitors motors, pumps, and valves to anticipate operational problems; automatically controls production; automates maintenance scheduling; and stores data to achieve AI-based machine learning. The client file 643 monitors each order to ensure that it is fulfilled according to the commitments made to a client, relative to specific, objective quality criteria, such as lead times, order quantities, and promised delivery dates. The production file 644 automatically plans production scheduling based on, for example: (a) matching outputs from scanned, input bales to match quality needs; and (b) fulfilling orders with 100% quality control, and on time. The open orders file 645 monitors production in real time and checks individual client orders against their master file criteria to ensure the requested results and delivery date are fulfilled. The trends/statistics file 646 monitors market trends and pricing to process the more desired valuable and preferred products for short-term inventory production based on historical trends and statistical opportunity. The AI modules 650 may use a “top-down, bottom-up” problem-solving process; they record and analyze all past production variables and their resultant “values” in order to “predict” future production outcomes. As part of the data science module 122, the AI modules 650 can build and use machine learning models to arrive at the predictions.
In some embodiments, the CIM server 640 uses some or all of its files at different stages of production, or even within each stage. When input stalk is conveyed to the decortication equipment 561, these files include, for example: the farmer file 641 (for quality and yield), the equipment file 642 (for maintenance and productivity), the client file 643 (for specifications, terms, and quantities), the production file 644 (for planned output), the open orders file 645 (for adjustments based on changes), and the trends/statistics file 646. When cut bast is delivered to the automated baler 563, these files include, for example: the farmer file 641 (for quality and yield), the client file 643 (for specifications, terms, and quantities), the production file 644 (for production), and the trends/statistics file 646. When the compounded/pelletized hurd is delivered to the bio-pellet storage silos 565, these files include, for example: the farmer file 641 (for quality monitoring and stalk yield), the equipment file 642 (for production and equipment monitoring/control), the client file 643 (for specifications and orders), the production file 644 (for production planning and scheduling), and the trends/statistics file 646. When the bio-pellets are delivered to the extruders/injection molders 566, these files include, for example: the equipment file 642 (for maintenance and productivity), the client file 643 (for specifications, terms, and quantities), the production file 644 (for planned output), the open orders file 645 (for adjustments based on changes), and the trends/statistics file 646.
Advantageously, as an example, the system 200 can preferentially select input bales based on desired product specifications of the fibre application or order volumes. As another example, the input bales can be selected based on desired bast purity requirements and/or end-use applications. In an exemplary case, input bales can be selected by receiving one or more orders, selecting raw materials that according to the machine learning models have produced the best product for the lowest cost, predict the yield of all fractions, look for other orders that can be filled with non-targeted fractions, and begin the decortication process.
As an example, the one or more of the analyzers 202 can be used to analyze the pulverized hurd to confirm hurd purity through chemical composition analysis. Such analysis can be used to close the component mass balance (in conjunction with Vis-NIR analysis on straw bales and clean bast product), functioning as calibration cross-check with other analyzers 202.
As an example, using the analysis of the data science module 122, the system 200 can determine how much chopped bast fibre is required in extruded product to achieve flexural strength targets. Whereby, for a defined extruded cross-section, flexural strength is typically a function of bast content, bast fibre length, pulverized particle size, and resin content in the extruded part.
As another example, based on continuous deflection testing (deflection distance and load) on extruded parts by one ore more of the analyzers 202, using continuous flexural testing, the controller 100 can control bast content. In general, if flexural strength fluctuates outside of control limits, the bast content is adjusted.
As another example, based on continuous deflection testing (deflection distance and load) on extruded parts by one or more of the analyzers 202, using the analysis of the data science module 122, the system 200 can adjust chop length of the cutting unit 204 to control flexural strength.
As another example, the data science module 122 can predict that the input stalk is relatively high in moisture for a desired characteristic of output fibre. This type of prediction may be part of standard model-predictive control, possibly informed by KDD. Thus, the control module 124 can instruct the system 200 components, such as the decorticator 208, to slow their respective throughput rate down, such as slowing the feed rate and/or saw speed. This can ensure that the fibre being processed from the stalk is not “damaged”, for example by ripping the fibre, because it was processed too fast.
As another example, the data science module 122 can use the machine learning model to extrapolate correlations between data from the one or more analyzers 202 at different stages of the decortication process to conceptualize the characteristics of the resultant products; for example, having certain weight, strength, or the like. In a particular case, operating the system will generate data that will form the basis for machine learning models (for example, regression-based models). For the regression-based models, the regression curves can be extrapolated to predict output of the decortication process based on the input characteristics of the input material. In a particular case, operating into the extrapolated region can thus generate new data that is no longer extrapolated. Repeating this approach multiple times can lead to the discovery of new material capabilities and possibly new products.
In some embodiments, various data science or machine learning techniques can be used. The data science module 122, either supervised or unsupervised, aims to use the data received from the one or more analyzers 202 to optimize particular aspects of the decortication process. In this case, the received data represents a training set, upon which the data science module 122 uses to refine the machine learning model, which the control module 124 uses to control the components of the decortication process. A number of machine learning approaches can be used by the data science module 122. For example, the Long Short Term Memory (LSTM) neural network, GPNet, or a suitable alternative thereto, can be used by the data science module 122 to efficiently improve particular aspects of the decortication process. When the database 116 contains no or little data, the data science module 122 can start with a limited supervision model and then move to an unsupervised model as the database 116 grows. The training set can be generated during start-up and can be corroborated by lab analysis to back-up advanced sensors (e.g., Vis-NIR).
In some embodiments, the one or more analyzers 202 can include a humidity sensor probe in the air handling system to allow the controller 100 to control moisture before decortication. In some embodiments, the one or more analyzers 202 can include a sensor to determine decorticator 208 motor loads to allow the controller 100 to control moisture before decortication. In some embodiments, the one or more analyzers 202 can include a load sensor in product bins to allow the controller 100 to control component yields. In some embodiments, the one or more analyzers 202 can include a sensor to determine pressure and/or position of a hydraulic ram of the baler 210 to allow the controller 100 to monitor and adjust for fibre fineness. In some embodiments, the one or more analyzers 202 can include a sensor to determine bale weights of the baler 210 to allow the controller 100 to monitor and adjust for fibre fineness. In some embodiments, the one or more analyzers 202 can include a sensor to determine air backpressure and/or flow rate on a face of the bale to allow the controller 100 to monitor fibre fineness. The data science module 122 can link the data obtained by the one or more analyzers 202 to raw material attributes, which can help the CIM to better select materials based on those parameters, using, for example, KDD and auto-tuning model-predictive control (MPC). In further embodiments, other sensors at various stages of the decortication process can be included as the one or more analyzers 202.
Advantageously, the present embodiments provide an approach for pre-qualifying the bales because with crops like hemp, one cannot assume it be of consistent quality from bale to bale. Advantageously, the optical or RFID scanners as the sensors of embodiments of the present disclosure can be used to track a farmer who supplied a particular bale or set of bales. In this way, quality of bales determined using other sensors (as described herein) can be associated with that particular farmer to generate a statistical profile of the quality of the bales from such farmer. Further, such tracking provides an opportunity for tracking and evaluating yield derived from each farmer's supply (for example, financial yield); thus allowing a comparative analysis of the supply provided by every farmer. In some cases, each farmer can supply agronomic data to be used as input into the machine learning model, described herein. In this way, the system can evaluate, compare, and automatically determine which combination of growing variables (e.g., varietal, irrigation, fertilizer, planting date, harvest date, etc.) resulted in the best yield for the past season. In some cases, this can be compared to the quality of fibre determined by the system and/or the farmer's volume yield of stalk per acre.
Advantageously, the system can use multiple instances of Vis-NIR spectrophotometry to optimize the output of the decortication process. Vis-NIR spectrophotometer measures a material's reflection or transmission by measuring light intensity/absorbency at a specific wavelength, within wavelength ranges that can span 135-3300 nm. In this way, different characteristics of the bast fibre can be determined by the reflected light; allowing for a “profile” of each natural fibre's inherent characteristics and quality factors. In an example, a first Vis-NIR scanner can be used for core sample analysis of each incoming bale. At arrival, a core sample of each bale can be taken. This sample of incoming bales can be used to optimize output value of each bale after the decortication process, as described herein. A second Vis-NIR scanner can be used for “dynamic” evaluation that takes place after the decortication process in order to evaluate the resultant individual fibres, or batch of fibres, to determine their qualities and characteristics. This allows the system to segregate these fibres for subsequent processing, as described herein. This dynamic evaluation can lead to optimizing the output value and provides further input data on the quality of bales provided by each farmer. A third Vis-NIR scanner can be used for quality control and for feedback to the system. This scanning can take place as the finished bast fibres are being baled for shipment. In some cases, the third Vis NIR scanner can be recessed into a head of the automatic baler. This scanning station not only enables the system to identify any “wrong” or inferior quality fibres, it also serves as a feedback mechanism for the entire facility operations to monitor quality of the output. This input data can be used by the machine learning model to match quality of the output fibre against an expected or desired quality, in real-time, in order to improve predictions and make adjustments during processing.
Advantageously, the embodiments described herein can be used for automated production planning and scheduling. Using the predicted output, the system 200 can perform real-time statistical analysis based on open orders, committed delivery dates, availability of raw materials and other production inputs, market prices for inputs and products, historical and contracted customer needs and expectations. As an example, a customer can provide desired specifications for the resultant product, such as through an online portal, and receive predicted pricing and completion date. Advantageously, the system 200 can allow for “just in time” production planning and execution. Thus, reducing costs associated with carrying inventory, which ties up capital and production time, significantly impacting gross margins, virtually eliminating old or damaged goods, handling costs and the need to do clearance sales. Also advantageously, the system 200 allows for material traceability from reception through to finished products. Further, the system 200 allows for full quality data sets that include raw material, in-process and finished product testing along with full process operating data. In this way, the system 200 provides the ability to generate Certificates of Analysis (C of A) with every order.
By applying machine learning techniques to analyzer 202 data generated by the system 200, decortication processes can be increasingly automated, thereby reducing the requirement for human intervention, and ultimately improving efficiency and accuracy of the resultant product.
The above described embodiments of the invention are intended to be examples of the present disclosure and alterations and modifications may be effected thereto, by those of skill in the art, without departing from the scope of the present disclosure, which is defined solely by the claims appended hereto. For example, embodiments discussed can be varied and combined, in full or in part.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/CA2020/050989 | 7/16/2020 | WO |
| Publishing Document | Publishing Date | Country | Kind |
|---|---|---|---|
| WO2021/007675 | 1/21/2021 | WO | A |
| Number | Name | Date | Kind |
|---|---|---|---|
| 805452 | Deegan et al. | Nov 1905 | A |
| 2926391 | Haas | Mar 1960 | A |
| 4805679 | Czinner | Feb 1989 | A |
| 4936206 | Miles et al. | Jun 1990 | A |
| 5101672 | Anthony et al. | Apr 1992 | A |
| 5262022 | Tench et al. | Nov 1993 | A |
| 5447276 | Aldridge | Sep 1995 | A |
| 5466188 | Schaal | Nov 1995 | A |
| 5507073 | Aldridge | Apr 1996 | A |
| 5507074 | Chen | Apr 1996 | A |
| 5805452 | Anthony et al. | Sep 1998 | A |
| 5892157 | Syre | Apr 1999 | A |
| 6079647 | Leduc | Jun 2000 | A |
| 6098454 | Ghorashi et al. | Aug 2000 | A |
| 6112131 | Ghorashi et al. | Aug 2000 | A |
| 6133348 | Kolla et al. | Oct 2000 | A |
| 6612258 | Isman | Sep 2003 | B2 |
| 6719225 | Hesch | Apr 2004 | B1 |
| 6841231 | Liang et al. | Jan 2005 | B1 |
| 7669292 | Chute | Mar 2010 | B2 |
| 8475628 | Abbott | Jul 2013 | B1 |
| 9777128 | Henry | Oct 2017 | B2 |
| 11535954 | Czinner | Dec 2022 | B2 |
| 11668022 | Henry | Jun 2023 | B2 |
| 20030050728 | Sarabi et al. | Mar 2003 | A1 |
| 20030181395 | Handa | Sep 2003 | A1 |
| 20080289149 | Chute | Nov 2008 | A1 |
| 20110105650 | Yan | May 2011 | A1 |
| 20140099497 | Panigrahi et al. | Apr 2014 | A1 |
| 20150166745 | Henry | Jun 2015 | A1 |
| 20150240058 | Panigrahi et al. | Aug 2015 | A1 |
| 20160325288 | Bates | Nov 2016 | A1 |
| 20170101512 | Henry et al. | Apr 2017 | A1 |
| 20180000307 | Henry | Jan 2018 | A1 |
| 20180000308 | Henry | Jan 2018 | A1 |
| 20180304274 | Bates | Oct 2018 | A1 |
| 20190390368 | Hefner | Dec 2019 | A1 |
| 20200398285 | Pildysh | Dec 2020 | A1 |
| Number | Date | Country |
|---|---|---|
| 617689 | Apr 2017 | AT |
| 101688328 | Mar 2010 | CN |
| 101916519 | Dec 2010 | CN |
| 102559541 | Jul 2012 | CN |
| 105579634 | May 2016 | CN |
| 105637127 | Jun 2016 | CN |
| 106435760 | Feb 2017 | CN |
| 106567138 | Apr 2017 | CN |
| 106795658 | May 2017 | CN |
| 107492251 | Dec 2017 | CN |
| S61160407 | Jul 1986 | JP |
| 2004167730 | Jun 2004 | JP |
| 2422821 | Jun 2011 | RU |
| 2621552 | Jun 2017 | RU |
| 1078273 | Mar 1984 | SU |
| 9116480 | Oct 1991 | WO |
| 0120065 | Mar 2001 | WO |
| 2017080999 | May 2017 | WO |
| 2019071361 | Apr 2019 | WO |
| Entry |
|---|
| EP Extended European Search Report for EP 18866822.2, European Patent Office, completed: May 19, 2021, dated May 28, 2021. |
| First Office Action for 201880073222.2, China National Intellectual Property Administration, dated Nov. 9, 2021. |
| International Preliminary Report on Patentability for PCT/CA2018/051295, The International Bureau of WIPO, International filing: Oct. 15, 2018, dated Apr. 14, 2020. |
| International Search Report for application No. PCT/CA2020/050989 dated Jul. 16, 2020. |
| International Search Report for PCT/CA2018/051295, ISA/CA, international filing: Oct. 15, 2018, search completed: Jan. 3, 2019, dated Feb. 5, 2019. |
| Office Action for U.S. Appl. No. 16/755,184, USPTO, application filed: Apr. 10, 2020, dated Apr. 22, 2022. |
| Office Action for 2020115475, Russian Federal Service for Intellectual Property, application filed: Oct. 15, 2018, completed: Mar. 11, 2022, dated Mar. 31, 2022. |
| Search Report for 2019101974297, China National Intellectual Property Administration, dated Mar. 15, 2019. |
| Second Office Action for 201880073222.2, China National Intellectual Property Administration, dated Mar. 7, 2022. |
| Written Opinion of ISA for PCT/CA2018/051295, Canadian Intellectual Property Office, international filing: Oct. 15, 2018, completed: Jan. 21, 2019, dated Feb. 5, 2019. |
| Written Opinion of the International Searching Authority for application No. PCT/CA2020/050989 dated Jul. 16, 2020. |
| Written Opinion of the International Searching Authority for application No. PCT/CA2020/050989, The International Bureau of WIPO, international filing: Jul. 16, 2020, dated Jan. 18, 2022. |
| Notification to Grant Patent Right for Invention for Chinese application No. 201880073222.2, China National Intellectual Property Administration, dated Nov. 22, 2022. |
| Office Action for Chinese application No. 202080064686.4, China National Intellectual Property Administration, dated Feb. 8, 2023. |
| Search Report for Brazilian application No. BR112020007217-6, BRPTO, dated May 26, 2022. |
| Search Report for Chinese application No. 202080064686.4, China National Intellectual Property Administration, dated Feb. 6, 2023. |
| Third Office Action for Chinese application No. 201880073222.2, China National Intellectual Property Administration, dated Aug. 8, 2022. |
| Extended European Search Report for EU application No. 20839703.4, EPO, dated Jul. 13, 2023. |
| Office Action for Chinese application No. 202080064686.4, China National Intellectual Property Administration, dated Aug. 26, 2023. |
| Search Report for Chinese application No. 202080064686.4, China National Intellectual Property Administration, dated Aug. 25, 2023. |
| Number | Date | Country | |
|---|---|---|---|
| 20220259769 A1 | Aug 2022 | US |
| Number | Date | Country | |
|---|---|---|---|
| 62874546 | Jul 2019 | US |
