Patent Application
20240367464
The invention relates to a computer-implemented method for forecasting the current wear state of an aeroplane tyre on the basis of parameters influencing its lifetime.
The tyres installed on aeroplanes (or “aeroplane tyres” or simply “tyres”) are, as is the case for other vehicles, wearing parts that are replaced when they achieve their removal thresholds (tread-pattern height below the limit of use of the tyre) or when the structure is too degraded to continue using it. The term “lifetime”, such as will be used below, refers to the longevity of an aeroplane tyre, i.e. the time to when it is removed from service, having reached the wear removal threshold, or, in other words, the time for which the aeroplane tyre is able to provide the expected service, without encountering endurance problems, which would require it to be removed earlier. In the aeronautics industry, it is known to measure this lifetime by counting the number of landings that a tyre performs between when it is first installed on an aeroplane and when it is uninstalled from the aeroplane (this also being called “landings per carcass” or “landings per tread” or LPT).
There are prior-art solutions for forecasting when maintenance of tyres is required depending on their expected lifetime. For example, U.S. Pat. No. 10,295,333 discloses a method for determining the state of a tread of a tyre on the basis of images of the tread. Furthermore, U.S. Pat. No. 10,179,487 discloses a method allowing a display representative of the potential wear level of a tyre, or of the state of its tread, to be generated. This display is partially based on measurements of the depth of the tread, which measurements are acquired beforehand. However, neither of these two solutions is applicable to aeroplane tyres. They therefore do not pertain to the field of forecasting a lifetime of an aeroplane tyre in number of landings. Publication WO2020/169833 discloses a method for detecting degradation of an aeroplane tyre. In the disclosed method, the determination of the position of a central point of an actual tyre comprises a first step of defining, for each captured point, a vector normal to a surface of a first three-dimensional object that passes through the captured point. The method comprises a second step of estimating a position of the central point of the actual tyre on the basis of the normal vectors (the estimation of the position of the central point is an iterative process). The method also includes a third step during which the first three-dimensional object is matched to a theoretical tyre of known dimensions and of known orientation with a view to obtaining a second three-dimensional object, forming a matched tyre. The second three-dimensional object is converted to obtain one or more two-dimensional objects, which are analysed to detect degradation of the actual tyre.
Publication WO2019116782A1 discloses a device for measuring the remaining groove height of at least a first tyre installed on an aeroplane, and predicting the wear levels of the other tyres installed on the aeroplane, including the first tyre. Wear computations are carried out depending on many parameters, including wear energy and the internal pressure of the aeroplane tyre.
However, the wear exhibits significant performance deviations (counted in number of landings) for a given tyre model that may easily exceed 100% between the observed minimum and maximum values. At the present time, these deviations are an issue with respect to maintenance operations because they prevent airlines from applying predictive-maintenance strategies that would allow them to optimize their operations and costs. At the present time, whether or not an aeroplane tyre will have to be removed because of damage to its structure is unforecastable because this is associated with a probabilistic event (for example, harm to the tyre caused by an external object present on the runway or taxiway). This damage is often referred to as “foreign object damage” or “FOD” and occurs when the aircraft passes over a runway or taxiway littered with rigid objects (the terms “aeroplane” and “aircraft” are used interchangeably). The objects that cause FOD comprise any type of object that is able to damage tyres (including, nonlimitingly, loose bits of metal, fragments of pavement, catering supplies, building materials, rocks, sand, items of luggage and wild animals). These objects are found at terminal gates, on runways and taxiways and on other surfaces over which aircraft move on the ground.
The current inability to forecast the moment of removal of a tyre at the end of its wear life (when the remaining tread-pattern height will become lower than or equal to 1.2 mm±0.2 mm) places on customers burdensome constraints in terms of inspection, of management of maintenance teams and of management of stocks of tyres, of brakes and of wheels, with the associated economic consequences. In addition, since aeroplane tyres are the component with the shortest lifetime of the tyre-brake-wheel-landing gear assembly, the entire maintenance chain of the tyre-brake-wheel-landing gear system of the aeroplane (for example, a system such as the ATA32) is dependent thereon.
On the basis of a physical model, it would in theory be possible to estimate the remaining tread-pattern height of an aeroplane tyre and to deduce therefrom, subsequently, a lifetime remaining for this tyre (either in number of landings or in remaining tread-pattern height). However, this model would require many physical parameters (including, nonlimitingly, the temperature of the tread, the load on the tyre, its pressure, its speed, etc.) in order to be accurate enough. The aforementioned physical parameters are difficult to obtain without placing many sensors in the tyre.
The recent improvement in machine-learning and data-analysis techniques, combined with platforms for computing and storing data, has opened the way to development of new predictive approaches to managing aeroplane tyres that overcome the drawbacks of the prior art. In the field of artificial intelligence, machine-learning techniques are known, and their essence is to be “trained” on a high number of situations. By virtue of adjustment of weighting coefficients in a training phase, machine learning may predict the result of a new situation presented thereto. It will be noted that a plurality of different machine-learning methods are possible, including supervised learning (in which the algorithm is trained on a set of labelled data and learns until it is capable of obtaining the desired result), unsupervised or semi-supervised learning (in which the data are not labelled so that the network may learn so as to increase the accuracy of the algorithm), reinforcement learning (in which the algorithm is rewarded for positive results and punished for negative results) and active learning (as it learns, the algorithm requests examples and labels to refine its prediction) (see https://www.lebigdata.fr/reseau-de-neurones-artificiels-definition).
It would therefore be advantageous to explore the relationship between the use of aeroplane tyres (i.e. parameters influencing the lifetime of an aeroplane tyre) and the lifetime performance of tyres. This performance knowledge comprises a comprehension of wear profiles observed as a result of parameters of the use of the aeroplane.
Thus, the disclosed invention relates to a method for forecasting the remaining lifetime of an aeroplane tyre, wherein the remaining lifetime is estimated indirectly via the wear state of the tyre. This remaining lifetime corresponds to a potential residual wear, which serves as a predictive-maintenance tool allowing all the activities that are associated with the maintenance schedule of the aeroplane tyres to be better scheduled.
The invention relates to a computer-implemented forecasting method for forecasting a number of residual landings corresponding to achievement of a removal threshold of an identified tyre that is installed on an identified aeroplane, the forecasting method comprising the following steps:
In certain embodiments of the method, the influencing parameters comprise:
In certain embodiments of the method, the maintenance schedule created by the system during the comparing step comprises:
In certain embodiments of the method, the method further comprises a supervised-learning method that receives as input the obtained influencing parameters and the data of the training database so that the processor may acquire known wear states corresponding to the number of landings carried out by the identified tyre to construct the forecasting model.
In certain embodiments of the method, the supervised-learning method comprises a supervised-learning method of GBR (gradient boosting regressor) type.
In certain embodiments of the method, the training database includes images of wear profiles corresponding to known wear states and corresponding numbers of landings carried out by the identified tyre.
In certain embodiments of the method, during the forecasting step, the number of residual landings of the identified tyres is computed on the basis of data corresponding to the influencing parameters of future landings.
In certain embodiments of the method, the method further comprises a step of simulating destination airports of the identified aeroplane on the basis of historic flight data.
In certain embodiments of the method, the simulating step is performed by way of a Markov chain in which each state represents one airport and each inter-airport link represents a probability of landing in one airport starting from another.
In certain embodiments of the method, the simulating step is repeated a plurality of times via a Monte-Carlo loop with a view to forecasting an end-of-life of the identified tyre.
Further aspects of the invention will become obvious from the following detailed description.
The nature and various advantages of the invention will become more obvious from reading the following detailed description, in conjunction with the attached drawings, throughout which the same reference numerals denote parts that are identical, and in which:
The solution proposed by the invention aims to estimate a lifetime of an aeroplane tyre without adding new sensors and solely using historic information (comprising, for example, data corresponding to dates of departure, dates of arrival, airports visited and climatic data) of flights that the tyre in question has made and general information on this tyre (comprising, for example, data corresponding to its retread level, its position of installation on the aeroplane, the theoretical average number of landings possible in this position and the performance of similar tyres excluding removal as a result of FOD). These data, which describe a large part of the use of a tyre, will be used in a machine-learning model in order to forecast the current wear state of the tyre on the basis of parameters influencing its lifetime.
Regarding the characteristics of an aeroplane tyre concerned by the invention, its geometry must be taken into account.
With reference to
With reference to
With reference now to
As used here, the “remaining lifetime” of an identified tyre refers to its potential residual wear as a function of the values of the parameters influencing its longevity. The remaining lifetime may be determined continuously or at regular, predefined or sporadic intervals by collecting data corresponding to the parameters influencing the lifetime of the identified tyre. On the basis of the collected data, the method of the invention may define a current wear state of the tyre with a view to determining the remaining lifetime of the identified tyre, which lifetime is computed in number of landings accessible.
The term “historic information” (in the singular or plural) is used here to refer to data corresponding to the historic flights of the identified aeroplane having the identified tyre installed. These data may comprise, nonlimitingly, data corresponding to the dates of departure and/or arrival of the identified aeroplane, the airports visited by the identified aeroplane (these data being capable of being collected from various sources, including the data of historic flights attributed to an airline 104 to which the identified aeroplane belongs), data on the aeroplane (including, nonlimitingly, the manufacturer, the version of the aeroplane model, its identification code, etc.) and meteorological and/or climatic conditions 106 during the historic flights.
The term “general information” (in the singular or plural) is used here to refer to data corresponding to the identified tyre. These data may comprise, nonlimitingly, its size (which may be represented by the type of tyre and/or its nomenclature), its construction code (for example, “-” for bias-ply and “R” for radial), its provenance as regards production 108 (for example, the name and/or trademark of the producer of the identified tyre, its date of manufacture and its site of manufacture, distribution and/or storage), its serial number, its load rating, its speed rating, its expected landings per tread or LPT and/or its unique identification code. By way of example, for a tyre of a size 52×21.0R22, the number “52” represents the diameter of the tyre in inches, the number “21.0” represents the cross-sectional width at the widest point of an inflated new tyre, the letter “R” represents a radial tyre, and the number “22” represents the diameter of the rim in inches.
The general information may also comprise the installation position 110 (including the installation history) of the identified tyre. It will be noted that the installation position shown in
The general information may also comprise the retreading level (if applicable) of the identified tyre. The term “retreading” is used here to refer to the method used to return a worn tyre to a state fit for operation by renewing the rubber of the tread and the one or more plies. During the retreading process, old tread products are removed and replaced by new materials. Retreading is a method that is known and regulated in the aerospace industry. Data corresponding to the retreading level of an identified tyre are managed by the airline 104 and/or by the producer (represented by the number 108).
In one embodiment of the invention, the historic information and/or the general information may be generated and/or managed, at least in part, by one or more airports (or by one or more networks of airports including the specific airport). In this embodiment, the system 100 aims to simulate destination airports of the identified aeroplane on the basis of the historic information. It will be noted that airlines 104 manage their aeroplanes differently, and their modes of management are therefore capable of modification. Thus, in this embodiment, the general information may also be used to specify the relationship between the historic information and the properties of the identified tyre by performing a personalized simulation with respect to the destination airports.
The communication network 102 of the system 100 comprises one or more communication devices (not shown) that capture and transmit the collected data to the server 102a. The one or more communication devices comprise one or more portable devices such as a mobile network device (for example, a mobile telephone, a laptop computer, one or more portable devices connected to the network, including “augmented reality” and/or “virtual reality” devices, and/or any combinations and/or any equivalents). In all cases, the communication devices may comprise clothes and/or wearables connected to the network, and worn by one or more operators (where each operator is a human or a known apparatus such as a robot). By way of example, a monitoring device worn by an airline pilot may monitor via video the conditions of a flight and send the corresponding data (being the historic information of the identified tyre) to the server 102a of the communication network 102.
The one or more communication devices may also comprise one or more remote computers able to transfer data via the communication network 102. By way of example, a portable device of the system 100 may transmit the historic information and/or the general information of the identified tyre to a remote computer of the system 100. On the basis of the transmitted data, the remote computer may transmit to the portable device a statement regarding maintenance of the identified tyre, indicating a plan regarding expected undertakings (for example, advice to remove from service the identified tyre because a forecasting model of the invention indicates that its tread depth is too low).
The communication network 102 may include wired or wireless links and may employ any data-transfer protocol known to those skilled in the art. Examples of wireless links may include, nonlimitingly, radiofrequency (RF) links, satellite links, (analogue or digital) cell or mobile telephone links, Bluetooth® links, Wi-Fi links, infrared links, ZigBee links, local-area-network (LAN) links, wireless-local-area-network (WLAN) links, wide-area-network (WAN) links, near-field-communication (NFC) links, links according to other wireless-communication standards and configurations, their equivalents, and a combination of these elements.
It will be understood that the communication network 102a implies the use of one or more processors as will be understood by anyone skilled in the art. The term “processor” (or, alternatively, the term “programmable logic circuit”) (in the singular or in the plural) refers to one or more devices capable of processing and analysing data and comprising one or more software packages for processing same (for example one or more integrated circuits known by those skilled in the art as being included in a computer, one or more controllers, one or more microcontrollers, one or more microcomputers, one or more programmable logic controllers (or PLCs), one or more application-specific integrated circuits, one or more machine-learning algorithms and/or one or more other known equivalent programmable circuits).
The invention therefore capitalizes upon methods and tools based on artificial intelligence (or “AI”) to complete partial information that is delivered (on the basis of the historic information and general information received). An analysis of the machine learning is carried out using a machine-learning model such as an artificial neural network comprising a plurality of layers. The machine-learning analysis receives as input processed data from the communication network 102, i.e. the historic information and the general information of the identified tyre. In the embodiments, data are received by way of inputs into various layers of the machine-learning model. The algorithm allows continuous improvement across all of the aeroplane tyres, ensuring self-improvement of the system 100 from the experience it acquires, particularly as regards the choice between a plan to service and a plan to inspect the identified tyre.
With reference once again to
On starting the forecasting method 200 shown in
The introducing step comprises a step of creating a training database (or “database”) of the wear states, which is introduced into the forecasting model. The created training database may comprise a reference database of the aeroplane tyres that are intended to be installed on the identified aeroplane. The database may comprise an already created reference (for example, a table of remaining lifetimes of the aeroplane tyres at a plurality of wear thresholds). The database may include parameters corresponding to a plurality of commercially available tyres (including parameters that form part of the general information discussed above). The specific source of the wear states of a plurality of wear levels is not essential to the method described here, which would function equally well using data obtained exclusively from tyres having reached their wear threshold. By way of example, a system may implement the method of the invention using data obtained from measurements performed on tyres received by a factory after deinstallation of the installed assembly.
This system implements the method of the invention to assist an airline with transmission of information on the progress of wear of the tread of an identified tyre to personnel of an airport served by the airline, perhaps during an inspection of the aeroplane. This system implements the method of the invention using data from measurement of the wear threshold of the “stocked” tread to prepare personalized reminders to be sent to an operator of the aeroplane (i.e. the airline or an airport served by the airline) to forecast servicing of the identified tyres. The created database may include images of wear profiles corresponding to known wear states and the corresponding number of landings carried out. The wear profile of an identified tyre is defined by the outline of the external surface of its tread during use. Thus, a new tread having not yet been employed is considered to be an envelope worn to 0%, and a tread that meets the manufacturer's advisory conditions for removal is considered to be an envelope worn to 100%. The profile described here may consist of one or more (3D) surface measurements (carried out beforehand) that reproduce all or some of the external surface of a tread. Furthermore, the profile described here may consist of one or more (2D) linear measurements (carried out beforehand) in one or more planes that contain the axis of symmetry of the tread. The training database therefore comprises expected images (and data therefore) corresponding to (3D, 2D, 1D) profiles of worn tyres, and the landings of the aeroplane.
The forecasting method of the invention 200, which method is shown in
In one embodiment, the machine-learning method employed during the training step 204 comprises a supervised-learning method. The supervised-learning method may comprise one or more neural networks (for example, autoencoders, ANNs, CNNs, RNNs, perceptrons, long short-term memory (LSTM), Hopfield networks, Boltzmann machines, deep belief networks, deconvolutional neural networks, generative adversarial networks (GANs), etc.) and their complements and equivalents. The one or more CNNs may be trained with ground-truth data that are generated using data representative of the influencing parameters (for example, the data incorporated into the training database described above).
In one embodiment in which the training step 204 comprises a supervised-learning method, this step comprises a supervised-learning method of GBR (gradient boosting regressor) type. The GBR learning method receives as input data corresponding to the historic information and general information of the identified tyre (including, nonlimitingly, the position of the identified tyre on the identified aeroplane, the retreading level of this identified tyre and the proportion of landings in each of the airports visited by this identified tyre). The GBR learning method aims to predict the number of residual landings (remaining LPT) corresponding to achievement of the removal threshold of the identified tyre. Thus, the output to be predicted of the forecasting model will be the forecast of the wear state of the identified tyre (characterized, for example, by a date range in which the removal threshold of the identified tyre will be reached). This for example allows tyres with an excessively pronounced wear state (i.e. a wear state beyond a predetermined wear threshold that ensures correct operation of the tyre) to be removed preventively. By way of example, in certain use cases, the rate of wear of the shoulder of the identified tyre induces appearance of working plies before the skid limit is reached. Providing a wear profile of a worn aeroplane tyre allows immediate and unforecast removals to be avoided.
The forecasting method 200 of the invention further comprises a step 206 of forecasting the number of residual landings (or “remaining LPT”) before achievement of the removal threshold of the identified tyre. The number of residual landings of the identified tyres may be computed on the basis of data corresponding to the influencing parameters of future landings. It will be noted that the future-landing data may be hypothetical or real depending on the type of management applied by an airline to its fleet of aeroplanes and its ability to determine, within a given timeframe, the routes to which an aeroplane will be assigned. An offset between the true wear states and the number of forecast landings is denoted by a computed error, such an error indicating a variation in the tread of the identified tyre. The computed errors may be input into the training database (described above) to improve the predictive capacity of the forecasting model.
The forecasting method 200 further comprises a comparing step 208 during which the remaining LPT before achievement of the removal threshold of the identified tyre, which is output by the forecasting model, is compared to a value of the removal threshold. This removal threshold is defined by the user (for example, an airline) and/or by the manufacturer of the identified tyre with a view to organizing maintenance operations.
During the comparing step, the system 100 indicates a plan 210 to service the identified tyre if the number of residual landings output by the forecasting model is higher than the removal-threshold value defined for the identified tyre. Similarly, the system 100 indicates a plan 212 to inspect the identified tyre if the number of residual landings output by the forecasting model is equal to or lower than this removal-threshold value. In the case where the tread of the identified tyre has reached or is close to reaching the wear removal threshold, the system 100 takes into account the wear states of worn tyres obtained from measurements of wear states performed on aeroplane tyres received by a factory after use. In the case where the identified tyre must be replaced by another tyre of the same type, the forecasting model is updated during the replacement of the worn tyre by an identified tyre having a tread considered new (see the number 214 of
With reference again to
In embodiments of the method of the invention aiming to simulate destination airports of the identified aeroplane, the simulation of a number of destinations allows the proportion of landings of the identified tyre at each of a plurality of airports identified in the simulation to be computed (therefore, using “real” and simulated data). This step 400 of simulating destination with a view to completing the life of the identified tyre is repeated a plurality of times via a Monte-Carlo loop (see number 402 of
In the hypothetical case where the airline employs more scheduled management and knows the destinations of its aircraft a few weeks in advance, it could replace the simulation with introduction of true destinations (or even choose to employ an intermediate semi-deterministic management if it does not exactly know how its aircraft will be used but is able to define relatively probable destinations). Before implementing the Markov chain to recognize the long-term time-dependent models of the forecasting process, which reveal the behaviours of the relevant characteristics necessary to the uncertainty model, it is essential to study and understand the nature of the dataset (historic information and general information) that will be used to evaluate the models.
In all the embodiments of the forecasting method of the invention, one or more steps may be carried out iteratively.
Although the embodiments of the forecasting method of the invention are described here in the context of use of neural networks by way of machine-learning model, other types of machine-learning models may be used. The latter include, nonlimitingly, models using linear regression, logistic regression, decision trees, support vector machines, naive Bayes methods, k-nearest neighbour (k-NN) algorithms, k indicating a group, random forests, algorithms for reducing dimensionality, and gradient descent.
The invention therefore predicts a longevity (and therefore forecasts the maintenance schedule) of an aeroplane tyre using easily accessible data, allowing a reliable forecasting model to be created. Even though human specialists exhibit flexibility when making decisions regarding the plan to service or the plan to inspect an aeroplane tyre, they are not capable of analysing the large amounts of data required to make real-time decisions with a view to determining whether the identified aeroplane must be serviced. Personnel are incapable of analysing the large amounts of data required to make this decision. To this end, it is necessary to adopt an approach for amending airline operations that uses data on the various empirical rules employed by human specialists, in order to forecast maintenance statements.
The system 100 may include preprogrammed management information. For example, a forecasting-method adjustment may be associated with the parameters of the typical physical environments in which the system 100 functions (for example, the parameters of the visited airports). In some embodiments, the system 100 (and/or an installation incorporating the system 100) may receive voice commands or other audio data representing, for example, start or stoppage of capture of the data corresponding to the historic information and/or general information of the identified tyres, or start or stoppage of movement of the communication device. A request made to the system 100 may include a request for the current state of an automatic-forecasting-method cycle. A generated response may be represented audibly, visually, in a tactile manner (for example by way of a haptic interface) and/or in a virtual and/or augmented manner. This response, together with the corresponding data, may be recorded in a neural network.
The terms “at least one” and “one or more” are used interchangeably. The ranges given as lying “between a and b” encompass the values “a” and “b”.
Although particular embodiments of the disclosed apparatus have been illustrated and described, it will be understood that various changes, additions and modifications can be made without departing from the spirit or the scope of the present description. Therefore, no limitation should be imposed on the scope of the invention described, apart from those set out in the appended claims.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2103344 | Mar 2021 | FR | national |
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/EP2022/057268 | 3/21/2022 | WO |
