Patent Application
20200319626
Embodiments of the present invention generally relate to the technical field of device monitoring in a process industry, and in particular relate to a method and apparatus for monitoring the state of a device in a process industry and a computer storage medium.
Equipment maintenance is the most important in the manufacturing industry. Improper equipment maintenance and any equipment fault will lead to a low efficiency and a high safety risk and especially cannot be borne in a process industry.
Once formed, a product is difficult to recover in a process industry. That is to say, the product cannot be disassembled once assembled. Physical or chemical methods such as mixing, separation, pulverization and heating are mainly used for raw materials in a process industry to increase the value of raw materials. Typical process industries include the pharmaceutical industry, the chemical industry, the petrochemical industry, the power industry, the steelmaking industry, the energy industry, the cement industry, the food industry, the beverage industry, etc. Enterprises in the process industries mainly adopt the make-to-stock, batch and continuous production methods.
The equipment of an enterprise in the process industries is a fixed production line with a large investment and a fixed process flow. The equipment has a limited production capacity. The maintenance of equipment in the production line is especially important and no fault is allowed. An equipment fault in the process industries will lead to a forced stop of the production process which involves many chemical reactions, and most intermediate products produced from incomplete reactions caused by the production stop will be discarded only as rejects. The incurred losses of raw materials and energy are enormous. In addition, in most cases, the equipment needs to be restarted after a production stop, and it will take the equipment some time to get stabilized and the efficiency and the safety will be influenced. Therefore, it is necessary to online monitor the state of the equipment so as to reduce unplanned stops by detecting the “health” problem of the equipment as early as possible.
However, it is very difficult to determine the health condition of the online equipment or obtain the fixed rules for monitoring the lifecycle state of the equipment, especially in a process industry where complex and highly sensitive correlations between a large number of variables are involved, because of a lack of an accurate definition of the boundary between the normal and abnormal equipment states.
The inventors have discovered that most manufacturers still perform planned routine maintenances, and in this case, the equipment runs until a scheduled maintenance time. This policy is non-real-time and highly dependent on experience. As a result, an excess of maintenances is caused because the maintenance time tends to be selected before any potential fault or a lack of maintenances is caused because it is almost impossible to provide all different fault modes in advance.
At least one embodiment of the present invention provides a method for monitoring the state of a device in a process industry to realize the online monitoring of the state of the device in the process industry.
At least one embodiment of the present invention provides an apparatus for monitoring the state of a device in a process industry to realize the online monitoring of the state of the device in the process industry.
At least one embodiment of the present invention provides a computer storage medium to realize the online monitoring of the state of a device in a process industry.
At least one embodiment of the present invention is directed to a method for monitoring the state of a device in a process industry, comprising:
collecting and saving data of the multi-dimensional states of a device in a process industry;
adopting a preset correlation analysis algorithm to perform a correlation analysis of the collected history data of the multi-dimensional states of the device to obtain the correlated state dimensions;
acquiring history data of correlated dimensional states according to the obtained correlated state dimensions and adopting a preset algorithm for establishing a device state reference model to model the acquired history data of correlated dimensional states to obtain the device state reference model; and
acquiring real-time data of the corresponding correlated dimensional states of the device state reference model in real time and determining whether the current state of the device is normal according to a preset device state determination condition, the real-time data of the correlated dimensional states and the device state reference model.
An apparatus of at least one embodiment, for monitoring the state of a device in a process industry, comprises:
a data acquisition module, collecting and saving data of the multi-dimensional states of a device in a process industry,
a correlation calculation module, adopting a preset correlation analysis algorithm to perform a correlation analysis of the history data collected by the data acquisition module of the multi-dimensional states of the device to obtain the correlated state dimensions,
a state reference model establishment module, acquiring history data of correlated dimensional states from the data acquisition module according to the correlated state dimensions obtained by the correlation calculation module and adopting a preset algorithm for establishing a device state reference model to model the acquired history data of correlated dimensional states to obtain the device state reference model, and
an online monitoring module, acquiring real-time data of the corresponding correlated dimensional states of the device state reference model in real time from the data acquisition module according to the device state reference model established by the state reference model establishment module and determining whether the current state of the device is normal according to a preset device state determination condition, the real-time data of the correlated dimensional states and the device state reference model.
A computer readable storage medium, in at least one embodiment, stores a computer program and the computer program performs the steps of the above-mentioned method for monitoring the state of a device in a process industry when executed by a processor.
An apparatus, in at least one embodiment, for monitoring the state of a device in a process industry comprises a processor and a memory, and
applications, which can be executed by the processor to enable the processor to perform the steps of the above-mentioned method for monitoring the state of a device in a process industry, are stored in the memory.
The following detailed description in combination with the drawings will make more obvious other characteristics, features and advantages of embodiments of the present invention.
At least one embodiment of the present invention is directed to a method for monitoring the state of a device in a process industry, comprising:
collecting and saving data of the multi-dimensional states of a device in a process industry;
adopting a preset correlation analysis algorithm to perform a correlation analysis of the collected history data of the multi-dimensional states of the device to obtain the correlated state dimensions;
acquiring history data of correlated dimensional states according to the obtained correlated state dimensions and adopting a preset algorithm for establishing a device state reference model to model the acquired history data of correlated dimensional states to obtain the device state reference model; and
acquiring real-time data of the corresponding correlated dimensional states of the device state reference model in real time and determining whether the current state of the device is normal according to a preset device state determination condition, the real-time data of the correlated dimensional states and the device state reference model.
The step of collecting and saving data of the multi-dimensional states of a device in a process industry comprises:
collecting data of the multi-dimensional states of the device in the process industry from one or any combination of the manufacturing data acquisition system, the supervisory control and data acquisition system and the simulation system and saving the data,
or collecting data of the multi-dimensional states of the device in the process industry from one or any combination of the manufacturing data acquisition system, the supervisory control and data acquisition system and the simulation system, and the device design system and saving the data.
After obtaining the correlated state dimensions, the method further comprises:
outputting the obtained correlated state dimensions to the user, and if receiving modification information input by the user about the correlated state dimensions, modifying the correlated state dimensions according to the modification information.
After obtaining the device state reference model, the method further comprises:
outputting the obtained device state reference model to the user, and if receiving modification information input by the user about the device state reference model, modifying the device state reference model according to the modification information.
The correlation analysis algorithm is the Pearson correlation algorithm and/or the Kendall correlation algorithm, and
the algorithm for establishing a device state reference model is the Dirichlet process algorithm based on the Gaussian mixture model.
An apparatus of at least one embodiment, for monitoring the state of a device in a process industry, comprises:
a data acquisition module, collecting and saving data of the multi-dimensional states of a device in a process industry,
a correlation calculation module, adopting a preset correlation analysis algorithm to perform a correlation analysis of the history data collected by the data acquisition module of the multi-dimensional states of the device to obtain the correlated state dimensions,
a state reference model establishment module, acquiring history data of correlated dimensional states from the data acquisition module according to the correlated state dimensions obtained by the correlation calculation module and adopting a preset algorithm for establishing a device state reference model to model the acquired history data of correlated dimensional states to obtain the device state reference model, and
an online monitoring module, acquiring real-time data of the corresponding correlated dimensional states of the device state reference model in real time from the data acquisition module according to the device state reference model established by the state reference model establishment module and determining whether the current state of the device is normal according to a preset device state determination condition, the real-time data of the correlated dimensional states and the device state reference model.
The data acquisition module, in at least one embodiment, collecting and saving data of the multi-dimensional states of a device in a process industry, is specifically used to
collect data of the multi-dimensional states of the device in the process industry from one or any combination of the manufacturing data acquisition system, the supervisory control and data acquisition system and the simulation system and save the data,
or collect data of the multi-dimensional states of the device in the process industry from one or any combination of the manufacturing data acquisition system, the supervisory control and data acquisition system and the simulation system, and the device design system and save the data.
The apparatus, in at least one embodiment, further comprises a man-machine interaction module, and
the correlation calculation module is further used to output the obtained correlated state dimensions to the user through the man-machine interaction module, and if receiving modification information input by the user about the correlated state dimensions from the man-machine interaction module, modify the correlated state dimensions according to the modification information.
The apparatus, in at least one embodiment, further comprises a man-machine interaction module, and
the state reference establishment module is further used to output the obtained device state reference model to the user through the man-machine interaction module, and if receiving modification information input by the user about the device state reference model from the man-machine interaction module, modify the device state reference model according to the modification information.
A computer readable storage medium, in at least one embodiment, stores a computer program and the computer program performs the steps of the above-mentioned method for monitoring the state of a device in a process industry when executed by a processor.
An apparatus, in at least one embodiment, for monitoring the state of a device in a process industry comprises a processor and a memory, and
applications, which can be executed by the processor to enable the processor to perform the steps of the above-mentioned method for monitoring the state of a device in a process industry, are stored in the memory.
From the above-mentioned technical solutions, it can be seen that in the present invention, the correlated state dimensions are obtained by performing a correlation analysis of history data of the multi-dimensional states of a device, then a device state reference model is established by using the history data of the correlated dimensional states, and finally the device state reference model is used to determine the real-time data of the correlated dimensional states to determine whether the state of the device is normal, thus realizing the online monitoring of the state of the device.
To make clearer the objects, technical solutions and advantages of the present invention, the following further describes in detail the present invention in combination with the drawings and embodiments.
“One” or “the” in the singular form in the description or claims of the present invention is also intended to cover the plural form, unless otherwise explicitly specified in the document.
The method for monitoring the state of a device in a process industry provided by the present invention comprises: collecting and saving data of the multi-dimensional states of a device in a process industry; adopting a preset correlation analysis algorithm to perform a correlation analysis of the collected history data of the multi-dimensional states of the device to obtain the correlated state dimensions; acquiring history data of correlated dimensional states according to the obtained correlated state dimensions and adopting a preset algorithm for establishing a device state reference model to model the acquired history data of correlated dimensional states to obtain the device state reference model; acquiring real-time data of the corresponding correlated dimensional states of the device state reference model in real time and determining whether the current state of the device is normal according to a preset device state determination condition, the real-time data of the correlated dimensional states and the device state reference model.
The step of collecting and saving data of the multi-dimensional states of a device in a process industry can comprise collecting data of the multi-dimensional states of the device in the process industry from one or any combination of the manufacturing data acquisition system, the supervisory control and data acquisition system, and the simulation system and saving the data, or collecting data of the multi-dimensional states of the device in the process industry from one or any combination of the manufacturing data acquisition system, the supervisory control and data acquisition system and the simulation system, and the device design system and saving the data.
In order to obtain the more accurate result of the correlated state dimensions, after obtaining the correlated state dimensions, the method further comprises outputting the obtained correlated state dimensions to the user, and if receiving modification information input by the user about the correlated state dimensions, modifying the correlated state dimensions according to the modification information.
In order to obtain a more accurate device state reference model, after obtaining the device state reference model, the method further comprises outputting the obtained device state reference model to the user, and if receiving modification information input by the user about the device state reference model, modifying the device state reference model according to the modification information.
Wherein, the correlation analysis algorithm can be the Pearson correlation algorithm and/or the Kendall correlation algorithm, and the algorithm for establishing a device state reference model can be the Dirichlet process algorithm based on the Gaussian mixture model.
The apparatus for monitoring the state of a device in a process industry provided by at least one embodiment of the present invention comprises: a data acquisition module, collecting and saving data of the multi-dimensional states of a device in a process industry, a correlation calculation module, adopting a preset correlation analysis algorithm to perform a correlation analysis of the history data collected by the data acquisition module of the multi-dimensional states of the device to obtain the correlated state dimensions, a state reference model establishment module, acquiring history data of correlated dimensional states from the data acquisition module according to the correlated state dimensions obtained by the correlation calculation module and adopting a preset algorithm for establishing a device state reference model to model the acquired history data of correlated dimensional states to obtain the device state reference model, and an online monitoring module, acquiring real-time data of the corresponding correlated dimensional states of the device state reference model in real time from the data acquisition module according to the device state reference model established by the state reference model establishment module and determining whether the current state of the device is normal according to a preset device state determination condition, the real-time data of the correlated dimensional states and the device state reference model.
Wherein, the data acquisition module collecting and saving data of the multi-dimensional states of a device in a process industry is specifically used to collect data of the multi-dimensional states of the device in the process industry from one or any combination of the manufacturing data acquisition system, the supervisory control and data acquisition system and the simulation system and save the data, or collect data of the multi-dimensional states of the device in the process industry from one or any combination of the manufacturing data acquisition system, the supervisory control and data acquisition system and the simulation system, and the device design system and save the data.
The apparatus further comprises a man-machine interaction module, and the correlation calculation module is further used to output the obtained correlated state dimensions to the user through the man-machine interaction module, and if receiving modification information input by the user about the correlated state dimensions from the man-machine interaction module, modify the correlated state dimensions according to the modification information.
The apparatus further comprises a man-machine interaction module and the state reference establishment module is further used to output the obtained device state reference model to the user through the man-machine interaction module, and if receiving modification information input by the user about the device state reference model from the man-machine interaction module, modify the device state reference model according to the modification information.
The computer readable storage medium provided by at least one embodiment of the present invention stores a computer program and the computer program performs the steps of the above-mentioned method for monitoring the state of a device in a process industry when executed by a processor.
Another device for monitoring the state of a device in a process industry provided by at least one embodiment of the present invention comprises a processor and a memory, and applications, which can be executed by the processor to enable the processor to perform the steps of the above-mentioned method for monitoring the state of a device in a process industry, are stored in the memory.
An embodiment of the present invention is described in detail below.
Step 101: Collect and save data of the multi-dimensional states of a device in a process industry.
Specifically, data of the multi-dimensional states of the device in the process industry can be collected from one or any combination of the manufacturing data acquisition system, the supervisory control and data acquisition system and the simulation system and can be saved, or data of the multi-dimensional states of the device in the process industry can be collected from one or any combination of the manufacturing data acquisition system, the supervisory control and data acquisition system and the simulation system, and the device design system and can be saved.
Here, the data of each dimensional state corresponds to a time sequence data flow for describing a state of a device. For example, if n sensors (for example, a temperature sensor, a vibration sensor, a pressure sensor, etc.) are deployed on a device to detect a plurality of states of a plurality of components of the device, data of n dimensional states of the device will be collected.
Step 102: Adopt a preset correlation analysis algorithm to perform a correlation analysis of the collected history data of the multi-dimensional states of the device to obtain the correlated state dimensions.
The correlation analysis algorithm is, for example, the Pearson correlation algorithm or the Kendall correlation algorithm.
For example, data of n dimensional states of a device is collected, it is learned that data of m (m≤n) dimensional states is correlated after a correlation analysis, and the dimensional values of the m dimensional states are 1, 3, 4, 8 . . . , respectively.
In practical applications, one of the dimensional states can be used as a target state, and then a correlation analysis algorithm is used to analyze which dimensional states are correlated to the target state.
Step 103: Acquire history data of correlated dimensional states according to the obtained correlated state dimensions and adopt a preset algorithm for establishing a device state reference model to model the acquired history data of correlated dimensional states to obtain the device state reference model.
The algorithm for establishing a device state reference model is, for example, the Dirichlet process algorithm based on the Gaussian mixture model.
For example, if it is learned from the analysis by use of a correlation analysis algorithm that a plurality of state dimensions correlated to a target state dimension (for example, state dimension p) are dimensions q, u and v, respectively, then an attempt is made to establish a mapping when the device state reference model is established. In the mapping, history data of state dimensions q, u and v is used as input parameters and history data of state dimension p is used as output parameter. The mapping is the device state reference model, that is to say, the device state reference model depicts the relationship between state dimension p and state dimensions q, u and v.
A device state reference model can be established for each target state dimension of different target state dimensions of a device.
Step 104: Acquire real-time data of the corresponding correlated dimensional states of the device state reference model in real time according to the device state reference model and determine whether the current state of the device is normal according to a preset device state determination condition, the real-time data of the correlated dimensional states and the device state reference model.
For example, real-time data of state dimensions p, q, u and v is acquired in real time according to the device state reference model established to depict the relationship between state dimension p and state dimensions q, u and v, then real-time data of state dimensions q, u and v is input into the device state reference model for calculations to obtain an expected data value of state dimension p, the difference between the expected data value of state dimension p and the real-time acquired actual data value of state dimension p is calculated, it is determined whether the difference is less than a preset threshold, if yes, the state of dimension p is considered normal, and otherwise, the state of dimension p is considered abnormal.
The device state determination condition can be: if the difference between the expected data value of the device target state obtained through calculations according to the device state reference model and the actual data value of the device target state is less than a preset threshold, the device target state is considered normal, and otherwise, the device target state is considered abnormal. The specific device state determination condition can be set by the user according to the application scenario.
Specifically, after the current state of a device is determined to be normal or abnormal, the determined current state of the device can be output to the user, and the user can modify the current state of the device according to experience. If the user modifies the current state of the device according to experience, the corresponding device state reference model can be considered inaccurate, and the real-time data acquired in Step 104 of corresponding correlated dimensional states of the device state reference model is used as history data to re-establish a device state reference model.
The data acquisition module 21 collects and saves data of the multi-dimensional states of a device in a process industry.
In practical applications, the data acquisition module 21 can collect data of the multi-dimensional states of the device in the process industry from one or any combination of the manufacturing data acquisition system, the supervisory control and data acquisition system and the simulation system and save the data, or collect data of the multi-dimensional states of the device in the process industry from one or any combination of the manufacturing data acquisition system, the supervisory control and data acquisition system and the simulation system, and the device design system and save the data.
The correlation calculation module 22 adopts a preset correlation analysis algorithm to perform a correlation analysis of the history data collected by the data acquisition module 21 of the multi-dimensional states of the device to obtain the correlated state dimensions.
In practical applications, the apparatus 20 can further comprise a man-machine interaction module 25, and the correlation calculation module 22 is further used to output the obtained correlated state dimensions to the user through the man-machine interaction module 25, after that, if receiving modification information input by the user about the correlated state dimensions, the man-machine interaction module 25 sends the modification information to the correlation calculation module 22, and the correlation calculation module 22 modifies the correlated state dimensions according to the modification information.
The state reference model establishment module 23 acquires history data of correlated dimensional states from the data acquisition module 21 according to the correlated state dimensions obtained by the correlation calculation module 22 and uses a preset algorithm for establishing a device state reference model to model the acquired history data of correlated dimensional states to obtain the device state reference model.
In practical applications, the apparatus 20 can further comprise a man-machine interaction module 25, and the state reference model establishment module 23 is further used to output the obtained device state reference model to the user through the man-machine interaction module 25, after that, if receiving modification information input by the user about the device state reference model, the man-machine interaction module 25 sends the modification information to the state reference model establishment module 23, and the state reference model establishment module 23 modifies the device state reference model according to the modification information.
The online monitoring module 24 acquires real-time data of the corresponding correlated dimensional states of the device state reference model in real time from the data acquisition module 21 according to the device state reference model established by the state reference model establishment module 23 and determines whether the current state of the device is normal according to a preset device state determination condition, the real-time data of the correlated dimensional states and the device state reference model.
Step 301: The data acquisition module 21 periodically collects and saves data of the multi-dimensional states of a device in a process industry.
Specifically, the data acquisition module 21 can collect data of the multi-dimensional states of a device in a process industry from the following systems:
1. Manufacturing Data Acquisition System, for Example, Distributed Control System (DCS).
The manufacturing data acquisition system collects data of the multi-dimensional states of the device from various sensors and/or various measuring instruments on the device in real time, and then only needs to send the collected data of the multi-dimensional states of the device to the data acquisition module 21 in real time, wherein the time sequence data flow collected from each sensor or measuring instrument is used as data of a dimensional state.
For example, for the DCS in the petrochemical industry, the collected data of the multi-dimensional states of a device can contain:
(11) Temperature data collected from temperature sensors on the device, wherein the specific DCS needs to record the collection time, the IDs/Nos. of the temperature sensors and the corresponding temperature values
(12) Vibration data collected from vibration sensors on the device, wherein the specific DCS needs to record the collection time, the IDs/Nos. of the vibration sensors and the corresponding vibration values (for example, flags indicating whether vibrations occur)
(13) Pressure values collected from pressure sensors on the device, wherein the specific DCS needs to record the collection time, the IDs/Nos. of the pressure sensors and the corresponding pressure values
(14) Valve opening values collected from valve opening sensors on the device, wherein the specific DCS needs to record the collection time, the IDs/Nos. of the valve opening sensors and the corresponding valve opening values.
2. Supervisory Control and Data Acquisition (SCADA) System
The SCADA system collects data of the multi-dimensional states of the device from various sensors and/or various measuring instruments on the device in real time, and then only needs to send the collected data of the multi-dimensional states of the device to the data acquisition module 21 in real time.
For example, for the SCADA system in the hydropower industry, the collected data of the multi-dimensional states of a device can contain:
(21) Current values collected from ammeters on the device, wherein the specific SCADA system needs to record the collection time, the IDs/Nos. of the ammeters and the corresponding current values.
(22) Voltage values collected from voltmeters on the device, wherein the specific SCADA system needs to record the collection time, the IDs/Nos. of the voltmeters and the corresponding voltage values.
3. Device Design System
For the device design system, the data collected by the data acquisition module 21 of the multi-dimensional states of a device can contain:
(31) Distance between sensors on the device
(32) Length of pipelines on the device
(33) Design data of pipelines on the device, for example, shapes and dimensions of components of the pipelines, connections between components and distances between components.
4. Simulation System
Sometimes, the simulation system can be utilized to simulate the actual production process, and in this case, the data acquisition module 21 can directly collect the required data of the multi-dimensional states of a device.
Step 302: The correlation analysis module 22 adopts a preset correlation analysis algorithm to perform a correlation analysis of the collected history data of the multi-dimensional states of the device to obtain a correlation indicator vector, wherein the correlation indicator vector indicates the correlated state dimensions and the description of the correlation between the correlated dimensional states.
It should be noted that when a correlation analysis is performed for the collected history data of the multi-dimensional states of the device in this step, the state dimensions can be based on the collected object of the state data, for example, the state data coming from each collected object such as sensor or measuring instrument is used as data of a dimensional state.
The time range of history data can be set by the user and can be a recent month or a recent year, for example.
For example, when the correlation calculation module 22 performs a correlation analysis of history data of n (n is an integer greater than 1) dimensional states of the device, the obtained correlation indicator vector indicates that m (m is an integer greater than 1 but less than or equal to n) dimensional states are correlated and the correlation indicator vector contains a description of the correlation between m dimensional states.
The description of the correlation between correlated dimensional states can be as follows: A change of a dimensional state or multi-dimensional states will cause a change to another dimensional state or other multi-dimensional states. For example, when a first change happens to the state of dimension a in a first time, a second change will be caused to the state of dimension b in a second time, a third change will be caused to the state of dimension c in a third time, . . . , wherein the first time, the second time, the third time, . . . , can be a time point or a time range.
For example, if the state dimensions related to the vibrations in a position of a pump need be analyzed, the correlation calculation module 22 performs a correlation analysis of the history data of vibrational state of dimension a collected by the data acquisition module 21 from the vibration sensor (sensor A) in the position and the history data of the states collected from other sensors on the pump, and finally learns that the states of corresponding dimensions b, d and f of sensors B, D and F are correlated to the vibrational state of dimension a of sensor A and obtains the description (for example, how long and what change will happen to the data of the states of dimensions b, d and f output from sensors B, D and F after sensor A detects that the vibrational state of dimension a is intensified: e.g. the temperature output from sensor B will rise in 0 to 10 ms after sensor A detects that the vibration is intensified) about the correlation between the states of dimensions b, d and f and the state of dimension a.
In practical applications, the user can set a target state dimension for a correlation analysis on the correlation calculation module 22, and the correlation calculation module 22 will calculate the dimensions related to the target state dimension according to the collected history data of all state dimensions of the device.
Wherein, the correlation analysis algorithm is, for example, the Pearson correlation algorithm or the Kendall correlation algorithm. The correlation analysis algorithm is not restricted in the present invention and can be preset in advance according to experience and the scenario.
In addition, the user further needs to configure some configuration parameters necessary for a correlation analysis on the correlation calculation module 22, for example, configure the time range (for example, in a recent month or a recent year) of history data of a state, and a correlation determination threshold (for example, when the correlation between two state dimensions is greater than a first threshold, the two state dimensions are determined to be correlated, and otherwise, the two state dimensions are not correlated).
After the correlation calculation module 22 obtains a correlation indicator vector, the correlation indicator vector can be output through the man-machine interaction module 25, and
the user can modify the correlation indicator vector (for example, delete a part of state dimensions from the correlation indicator vector or add other state dimensions to the correlation indicator vector) according to his/her own judgment, or the user can change the values of a part or all of the configuration parameters of the correlation analysis algorithm through the man-machine interaction module 25 and trigger the correlation calculation module 22 to perform a correlation analysis again until the user is satisfied with the output correlation indicator vector.
Step 303: The state reference model establishment module 23 determines the correlated state dimensions according to the correlation indicator vector generated by the correlation calculation module 22, acquires history data of correlated dimensional states from the data acquisition module 21, and uses a preset algorithm for establishing a device state reference model to model the acquired history data of correlated dimensional states to obtain the device state reference model.
For example, if the correlation indicator vector generated by the correlation calculation module 22 indicates that m dimensional states of the device are correlated, the state reference model establishment module 23 acquires history data of m dimensional states from the data acquisition module 21. The time range of history data can be set by the user.
The algorithm for establishing a device state reference model is, for example, the Dirichlet process algorithm based on the Gaussian mixture model. The algorithm for establishing a device state reference model is not restricted in the present invention and can be preset in advance according to experience and the scenario.
In addition, the user further needs to configure some configuration parameters necessary for the algorithm for establishing a device state reference model on the state reference model establishment module 23, for example, configure the time range (for example, in a recent month or a recent year) of history data of a state.
Likewise, after the state reference model establishment module 23 obtains the device state reference model, the device state reference model can be output to the user through the man-machine interaction module 25, and the user can modify the device state reference model through the man-machine interaction module 25, or the user can change the values of a part or all of the configuration parameters of the algorithm for establishing a device state reference model through the man-machine interaction module 25 and trigger the state reference model establishment module 23 to re-establish a device state reference model until the user is satisfied with the output device state reference model.
Step 304: The online monitoring module 24 acquires real-time data of the corresponding correlated dimensional states of the model from the data acquisition module 21 in real time according to the established device state reference model and determines whether the current state of the device is normal according to a preset device state determination condition, the real-time data of the correlated dimensional states and the device state reference model.
Likewise, after the online monitoring module 24 obtains the device state monitoring result (normal or abnormal), the device state monitoring result can be output to the user through the man-machine interaction module 25, and the user can modify the device state monitoring result through the man-machine interaction module 25 according to experience. If the user modifies the current state of the device according to experience, the corresponding device state reference model can be considered inaccurate, and the real-time data acquired in Step 104 of corresponding correlated dimensional states of the device state reference model is used as history data to re-establish a device state reference model.
Considering that the aging of a device will cause the production capacity of the device to change, steps 301 to 303 can be performed at intervals, for example, once a month, to continuously update the device state reference model, or can be performed at any time according to the requirements of the user.
collect and save data of the multi-dimensional states of a device in a process industry;
adopt a preset correlation analysis algorithm to perform a correlation analysis of the collected history data of the multi-dimensional states of the device to obtain the correlated state dimensions;
acquire history data of correlated dimensional states according to the obtained correlated state dimensions and adopt a preset algorithm for establishing a device state reference model to model the acquired history data of correlated dimensional states to obtain the device state reference model;
acquire real-time data of the corresponding correlated dimensional states of the device state reference model in real time and determine whether the current state of the device is normal according to a preset device state determination condition, the real-time data of the correlated dimensional states and the device state reference model.
In practical applications, the step performed by the processor 41 of collecting and saving data of the multi-dimensional states of a device in a process industry in real time specifically comprises:
collecting data of the multi-dimensional states of the device in the process industry from one or any combination of the manufacturing data acquisition system, the supervisory control and data acquisition system and the simulation system and saving the data, or collecting data of the multi-dimensional states of the device in the process industry in real time from one or any combination of the manufacturing data acquisition system, the supervisory control and data acquisition system and the simulation system, and the device design system and saving the data.
In practical applications, after obtaining the correlated state dimensions, the processor 41 further performs the following step:
output the obtained correlated state dimensions to the user, and if receiving modification information input by the user about the correlated state dimensions, modify the correlated state dimensions according to the modification information.
In practical applications, after obtaining the device state reference model, the processor 41 further performs the following step:
output the obtained device state reference model to the user, and if receiving modification information input by the user about the device state reference model, modify the device state reference model according to the modification information.
In practical applications, the correlation analysis algorithm is the Pearson correlation algorithm and/or the Kendall correlation algorithm, and the algorithm for establishing a device state reference model is the Dirichlet process algorithm based on the Gaussian mixture model.
At least one embodiment of the present invention further provides a computer readable storage medium, the computer readable storage medium stores a computer program and the computer program performs the following steps of the method for monitoring the state of a device in a process industry when executed by a processor:
collect and save data of the multi-dimensional states of a device in a process industry;
adopt a preset correlation analysis algorithm to perform a correlation analysis of the collected history data of the multi-dimensional states of the device to obtain the correlated state dimensions;
acquire history data of correlated dimensional states according to the obtained correlated state dimensions and adopt a preset algorithm for establishing a device state reference model to model the acquired history data of correlated dimensional states to obtain the device state reference model;
acquire real-time data of the corresponding correlated dimensional states of the device state reference model in real time and determine whether the current state of the device is normal according to a preset device state determination condition, the real-time data of the correlated dimensional states and the device state reference model.
In practical applications, the step performed by the computer program of collecting and saving data of the multi-dimensional states of a device in a process industry in real time specifically comprises:
collecting data of the multi-dimensional states of the device in the process industry from one or any combination of the manufacturing data acquisition system, the supervisory control and data acquisition system and the simulation system and saving the data, or collecting data of the multi-dimensional states of the device in the process industry in real time from one or any combination of the manufacturing data acquisition system, the supervisory control and data acquisition system and the simulation system, and the device design system and saving the data.
In practical applications, when executed by a processor, the computer program further performs the following step:
output the obtained correlated state dimensions to the user, and if receiving modification information input by the user about the correlated state dimensions, modify the correlated state dimensions according to the modification information.
In practical applications, when executed by a processor, the computer program further performs the following step:
output the obtained device state reference model to the user, and if receiving modification information input by the user about the device state reference model, modify the device state reference model according to the modification information.
In practical applications, the correlation analysis algorithm is the Pearson correlation algorithm and/or the Kendall correlation algorithm, and the algorithm for establishing a device state reference model is the Dirichlet process algorithm based on the Gaussian mixture model.
The advantages of the prevent invention are as follows:
Online monitoring of the state of a device is realized;
semi-supervision is adopted and the visualization aided interactive knowledge discovery is supported so that users, for example, device operators/engineers, can combine their experience/domain knowledge to participate in the whole device state monitoring process and construct adaptive and robust device state reference models under different environments.
The above-mentioned embodiments are only preferred embodiments of the present invention, but are not used to restrict the present invention. Without departing from the spirit and principle of the present invention, modifications, equivalent replacements, and improvements should all fall within the scope of protection of the present invention.
This application is the national phase under 35 U.S.C. § 371 of PCT International Application No. PCT/CN2017/105572 which has an International filing date of Oct. 10, 2017, which designated the United States of America, the entire contents of which are hereby incorporated herein by reference.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/CN2017/105572 | 10/10/2017 | WO | 00 |
