An industrial equipment state identification and health management method and system

Abstract

The application provides an industrial equipment state recognition and health management method and system, the method comprising: acquiring comprehensive data of an industrial equipment at each historical time in a preset first historical period; calculating feature information according to operation data and environment data at each historical time; screening all feature information to obtain target feature information; searching for comprehensive data corresponding to the target feature information, training a model based on the comprehensive data corresponding to the target feature information to obtain an industrial equipment state information recognition model; inputting comprehensive data of the industrial equipment at a current time into the industrial equipment state recognition model to obtain state information corresponding to the industrial equipment at the current time, and generating a health management method according to the state information. The application automatically mines key boundary samples through an intelligent screening mechanism, reduces labeling costs while ensuring label quality by using a differentiated labeling strategy, and improves the accuracy of the final model recognition.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial equipment, in particular to an industrial equipment state recognition and health management method and system. BACKGROUND

[0002] In the field of industrial manufacturing, data-driven equipment state recognition is the core of health management. Current methods usually collect equipment operation data and environmental data, and use machine learning models for state warning. However, this method faces two major challenges in practical application: first, the amount of industrial data is huge but the value density is uneven, and some high-value data has a very low proportion, making it difficult to be effectively mined. Second, the cost of providing labels for massive data is too high, which further reduces the recognition efficiency of the model. SUMMARY

[0003] The purpose of the present application is to provide an industrial equipment state recognition and health management method and system to improve the above problems.

[0004] In order to achieve the above purpose, the embodiments of the present application provide the following technical solutions:

[0005] On the one hand, the embodiments of the present application provide an industrial equipment state recognition and health management method, which comprises:

[0006] Obtaining comprehensive data of the industrial equipment at each historical time in a preset first historical period, the comprehensive data including operation data and environmental data;

[0007] Calculating feature information according to the operation data and environmental data at each historical time, screening all feature information to obtain target feature information, finding comprehensive data corresponding to the target feature information, training a model based on the comprehensive data corresponding to the target feature information to obtain an industrial equipment state information recognition model;

[0008] Inputting the comprehensive data of the industrial equipment at the current time into the industrial equipment state recognition model to obtain the state information corresponding to the industrial equipment at the current time, and generating a health management method according to the state information.

[0009] Secondly, the embodiments of the present application provide an industrial equipment state recognition and health management system, which comprises:

[0010] The acquisition module is configured to acquire comprehensive data of the industrial equipment at each historical time in a preset first historical period, the comprehensive data including operation data and environmental data;

[0011] The training module is configured to calculate feature information according to operation data and environment data at each historical moment, filter all the feature information to obtain target feature information, find comprehensive data corresponding to the target feature information, train a model based on the comprehensive data corresponding to the target feature information, and obtain an industrial equipment state information recognition model.

[0012] The management module is configured to input comprehensive data of the industrial equipment at the current moment into the industrial equipment state recognition model to obtain state information corresponding to the industrial equipment at the current moment, and generate a health management method according to the state information.

[0013] In a third aspect, an embodiment of the present application provides an industrial equipment state recognition and health management device, which comprises a memory and a processor. The memory is configured to store a computer program, and the processor is configured to execute the computer program to implement the steps of the industrial equipment state recognition and health management method.

[0014] In a fourth aspect, an embodiment of the present application provides a readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of the industrial equipment state recognition and health management method are implemented.

[0015] The present application has the following beneficial effects:

[0016] The present application realizes automatic and accurate positioning of key samples from massive historical data by using a dynamic interval division based on an order of magnitude and a double-threshold screening mechanism. The method first identifies a target cluster representing the mainstream through scale grouping, and then finds back samples highly similar to the target cluster but not classified into the target cluster, i.e., some boundary samples and difficult example samples. This process overcomes the problem that valuable samples are diluted by a large number of conventional data in the traditional method.

[0017] The present application adopts different labeling processes for data with different values. For a small number of complex samples with high value, accurate operation and environment strong association label pairs are generated through hierarchical analysis of operation data clustering and environment data clustering. For massive ordinary samples, an efficient global clustering and label mapping mechanism is adopted. This labeling strategy can reduce labeling cost and improve labeling efficiency.

[0018] Other features and advantages of the present application will be described in the following description, and some will become apparent from the description, or will be learned from practice of the present application. The purpose and other advantages of the present application can be achieved and obtained by the structures specifically pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some of the embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor.

[0020] Figure 1 is a flowchart of the industrial equipment state recognition and health management method described in the embodiments of the present application;

[0021] Figure 2 is a structural diagram of the industrial equipment state recognition and health management device described in the embodiments of the present application;

[0022] Figure 3 is a structural diagram of the industrial equipment state recognition and health management device described in the embodiments of the present application. DETAILED DESCRIPTION

[0023] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the following will combine the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, but not all the embodiments. The components of the embodiments of the present application described and shown in the drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0024] It should be noted that: similar numbers or letters represent similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of the present application, the terms "first", "second" and the like are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0025] Embodiment 1

[0026] As shown in Figure 1 , the present embodiment provides an industrial equipment state recognition and health management method, which comprises steps S1, S2 and S3.

[0027] Step S1, obtaining comprehensive data of the industrial equipment at each historical time in a preset first historical period, the comprehensive data comprising operation data and environment data;

[0028] In this step, the first historical period can be the past six months, the running data represents the data of the dynamic behavior and internal state of the equipment itself, and the environment data represents the working condition parameters of the working conditions and external loads where the equipment is located, for example, when the industrial equipment is a numerical control machine tool, the running data can include spindle vibration, axis current, servo error, spindle speed, etc., and the environment data can include environment temperature, environment humidity, spindle bearing temperature, cooling system pressure and flow, etc.

[0029] Step S2, according to the running data and environment data at each historical moment, the feature information is calculated, all feature information is screened to obtain target feature information, the corresponding comprehensive data of the target feature information is searched, and the model is trained based on the corresponding comprehensive data of the target feature information to obtain the industrial equipment state information recognition model.

[0030] In this step, according to the running data and environment data at each historical moment, the feature information is calculated, and the specific implementation steps of obtaining the target feature information by screening all feature information include:

[0031] Step S21, the running data and environment data at each historical moment are spliced, the feature information of the spliced data is extracted, and the corresponding feature information of each historical moment is obtained; based on the full connection algorithm, by setting a first similarity threshold, all feature information is initially clustered to obtain a plurality of initial state clustering clusters; all initial state clustering clusters are traversed, the initial state clustering cluster containing the most feature information is found, and the number of feature information contained in the initial state clustering cluster is recorded as n;

[0032] In this step, splicing the running data and environment data at each historical moment can be understood as:

[0033] Suppose at time t, the running data is [vibration value, current value, speed value]; the environment data vector is [temperature value, humidity value], and the spliced data is [vibration value, current value, speed value, temperature value, humidity value]; the spliced data can also be understood as a vector, and the feature information is extracted, that is, the feature information of each historical moment is obtained, wherein, the spliced vector can be directly input into a full connection neural network layer, and mapped to a higher dimensional feature space through nonlinear transformation, and the output of the layer is the feature information. The conventional feature extraction method can also be used for feature extraction, and the present application will not be repeated here;

[0034] Meanwhile, in this step, based on the full linkage algorithm, the initial clustering of all feature information is performed by setting a first similarity threshold, and a plurality of initial state clustering clusters are obtained. It can be understood as follows: a complete linkage clustering algorithm is used to calculate the cosine similarity between any two feature information; the clustering with the highest similarity is iteratively merged, and when the maximum similarity between all clusters is less than the preset first similarity threshold, the clustering is terminated, thereby obtaining a plurality of initial state clustering clusters.

[0035] Step S22, calculate the bit number m of n, create m continuous number intervals based on the bit number m, wherein when i (i-1) i - 1], when i = m, the ith interval is [10 (m-1) , n]; analyze the number of feature information contained in each initial state clustering cluster, combine the initial state clustering clusters whose numbers are in the same number interval, obtain a combined cluster, and record the combined cluster containing the initial state clustering cluster whose number is not less than 3 as a target combined cluster; and filter out target feature information according to the target combined cluster.

[0036] This step can be understood as:

[0037] Calculate the bit number m: find n, calculate its decimal bit number m. For example, n = 258, then m = 3, create m = 3 intervals:

[0038] i = 1 (i < m): [10 (1-1) , 10 1 - 1] = [10 0 , 10 - 1] = [1, 9]

[0039] i = 2 (i < m): [10 (2-1) , 10 2 - 1] = [10 1 , 100 - 1] = [10, 99]

[0040] i = 3 (i = m): [10 (3-1) , n] = [10 2 , 258] = [100, 258]

[0041] Result: the interval is divided into [1, 9], [10, 99], [100, 258];

[0042] Meanwhile, in this step, the specific implementation steps of filtering out target feature information according to the target combined cluster include step S221 and step S222.

[0043] ​Step S221, for each target combination cluster, count the number of feature information contained in each initial state clustering cluster in the target combination cluster, and sort the initial state clustering clusters in order from most to least according to the number, and select the top 3 initial state clustering clusters as candidate clustering clusters; calculate the mean feature information of all feature information contained in each candidate clustering cluster, calculate the similarity between the mean feature information corresponding to each pair of candidate clustering clusters, and merge the two candidate clustering clusters corresponding to the maximum similarity to obtain a target clustering cluster;

[0044] Step S222, for each target clustering cluster, the remaining feature information not belonging to the target clustering cluster is recorded as first feature information; calculate the similarity between each first feature information and each feature information in each target clustering cluster, and record the maximum value in the obtained similarity as the attribution similarity of the first feature information with respect to the target clustering cluster, and record the first feature information corresponding to the attribution similarity greater than the second similarity threshold as the target feature information, the second similarity threshold is less than the first similarity threshold.

[0045] Step S221 and step S222 together realize the screening of key data: first, the target clustering cluster representing the mainstream is obtained by step S221; second, the samples highly similar to the target clustering cluster but not attributed to it, i.e. some boundary samples and difficult samples, are found back based on the target clustering cluster as a reference by step S222.

[0046] In step S2, the comprehensive data corresponding to the target feature information is found, and the industrial equipment state information recognition model is trained based on the comprehensive data corresponding to the target feature information. The specific implementation steps of the industrial equipment state information recognition model include step S23:

[0047] Step S23, collect the running data in the comprehensive data corresponding to all target feature information to obtain a first set; cluster the running data in the first set to obtain a plurality of first clustering clusters; for each first clustering cluster, assign a running mode label to each first clustering cluster, and the label of the running data is the same as the label of the first clustering cluster to which it belongs; find the environment data corresponding to each running data in the first clustering cluster, wherein the running data and the environment data corresponding thereto are acquired at the same time, collect all the environment data corresponding to each first clustering cluster to obtain a second set; cluster the environment data in each second set to obtain a second clustering result, and assign an environment mode label to each second clustering cluster, and the label of the environment data is the same as the label of the second clustering cluster to which it belongs; train an industrial equipment state information recognition model based on the running data, the environment data, the running mode label and the environment mode label corresponding to all target feature information.

[0048] In this step, the operation mode label is obtained by manually analyzing the response signal operation data of the equipment, directly reflecting the health or abnormal condition of the equipment, and the environment mode label is obtained by manually analyzing the environment data of the equipment, explaining the background and inducement of the equipment state, for example, in the field of numerical control machine tools, the operation mode label can be healthy steady state, spindle bearing wear, tool chatter and the like, and the environment mode label can be standard working condition, high temperature and humidity environment and the like;

[0049] In this step, the specific implementation steps of training the industrial equipment state information recognition model according to the operation data, environment data, operation mode label and environment mode label corresponding to all target feature information include step S231;

[0050] In step S231, for the comprehensive data corresponding to all target feature information, the operation data and environment data at the same historical moment are found, the operation data and environment data are spliced into a complex sample, and the operation mode label corresponding to the operation data and the environment mode label corresponding to the environment data are taken as the label corresponding to the complex sample; the operation data and environment data corresponding to each feature information except the target feature information are spliced into an ordinary sample, all ordinary samples are clustered by using a clustering algorithm, a plurality of third clustering clusters are obtained, an operation mode label and an environment mode label are matched for each third clustering cluster, and the operation mode label and the environment mode label of the ordinary sample are the same as the operation mode label and the environment mode label of the third clustering cluster to which the ordinary sample belongs; a convolutional neural network model is trained by using the complex sample and the ordinary sample with the operation mode label and the environment mode label, and an industrial equipment state information recognition model is obtained.

[0051] The contents in step S23 and step S231 are that, for a small amount of complex samples with high value, a hierarchical clustering analysis method is used for fine disassembly, the operation mode label is determined through operation data clustering, and the inducing condition is identified through environment data clustering, to form a strong association label pair of operation mode and environment mode. For a large number of ordinary samples with low value density, a global clustering and label mapping mechanism is used. In this way, the labeling efficiency can be improved. Meanwhile, during training, the complex sample and the ordinary sample are input, and the operation mode label and the environment mode label are output for training;

[0052] In step S3, the comprehensive data of the industrial equipment at the current moment is input into the industrial equipment state recognition model, the state information corresponding to the industrial equipment at the current moment is obtained, and a health management method is generated according to the state information.

[0053] In this step, the specific implementation steps of generating a health management method according to the state information include step S31 and step S32;

[0054] Step S31, a first table in which operation mode scores correspond to operation mode labels and a second table in which environment mode scores correspond to environment mode labels are constructed; according to the operation mode label and the environment mode label corresponding to the industrial equipment at the current time, the corresponding operation mode score and the environment mode score are found from the first table and the second table; at the same time, the operation mode label and the environment mode label corresponding to the industrial equipment at the current time are combined to form a current combined label; a dangerous combined label library is constructed, the dangerous combined label library contains a plurality of combined labels and a synergistic risk multiplier corresponding to each combined label;

[0055] In this step, the first table, the second table and the dangerous combined label library are artificially constructed in the system, wherein the dangerous combined label library can be understood as a pre-defined blacklist library, which lists known and particularly dangerous combinations, and assigns a synergistic risk multiplier to each dangerous combination. For example: the synergistic risk multiplier corresponding to the label (early bearing wear, insufficient cooling) is 1.5, and the synergistic risk multiplier corresponding to the label (severe tool wear, heavy cutting) is 1.8; the synergistic risk multiplier amplifies the risk of a particular combination. For example, 1.5 means that the overall risk of the combination will increase by 50% on the basis of the basic risk. This captures the complex effect of some environments and operating conditions intensifying each other, leading to accelerated deterioration of failure.

[0056] Step S32, determining whether the current combined label exists in the dangerous combined label library, calculating the working condition risk comprehensive index according to the determination result, determining the risk level according to the value interval in which the working condition risk comprehensive index is located, and taking different health management measures to manage the industrial equipment according to the different risk levels, wherein each value interval corresponds to a risk level.

[0057] This step can be understood as:

[0058] Hierarchical mapping: preset risk level intervals, the risk levels can be low risk, medium risk and high risk, for example:

[0059] 0 ≤ R (working condition risk comprehensive index) <0.5: low risk (normal)

[0060] 0.5 ≤ R <1.0: medium risk (warning)

[0061] R ≥ 1.0: high risk (alarm)

[0062] Decision execution: according to the finally determined risk level, different health management methods are sent to the staff, for example:

[0063] Low risk: send information "record log, continuous monitoring".

[0064] Medium risk: send information as "warning, specific operation mode label and environment mode label, please check after this shift"

[0065] High risk: send information as "emergency alert! Operation mode label and environment mode label, suggest immediate shutdown for maintenance!"

[0066] At the same time, in this step, it is judged whether the current combination label exists in the dangerous combination label library, and the specific implementation steps of calculating the working condition risk comprehensive index according to the judgment result include step S321;

[0067] Step S321, judge whether the current combination label exists in the dangerous combination label library, if it exists, the synergistic influence factor is 1, and the working condition risk comprehensive index calculation formula includes:

[0068] ;

[0069] In the formula, R represents the working condition risk comprehensive index; represents the operation mode score; represents the environment mode score; represents the operation mode score threshold (manually defined); represents the environment mode score threshold; represents the synergistic risk multiplier; represents the synergistic influence factor; is the abnormal proportion; wherein, the operation mode score threshold and the environment mode score threshold are manually defined;

[0070] Step S322, the calculation method of the abnormal proportion includes: obtaining the comprehensive data of the industrial equipment at each time in a preset second historical period, the start time of the second historical period is later than the end time of the first historical period, and the end time of the second historical period is the current time; input all the comprehensive data in the second historical period and the first historical period into the industrial equipment state information recognition model, get the operation mode label and the environment mode label corresponding to each comprehensive data, and combine the operation mode label and the environment mode label corresponding to each comprehensive data to form a combination label to be processed; find the same combination label to be processed as the current combination label in all combination labels to be processed and record it as the target combination label, count the first number of times of the target combination label in the second historical period, count the second number of times of the target combination label in the second historical period, calculate the ratio of the first number and the second number, and analyze the ratio, wherein, when the ratio is greater than a preset ratio threshold, the ratio is recorded as the abnormal proportion, otherwise, the abnormal proportion is zero.

[0071] In this step, the first historical period can be the past week; if the target combination label accidentally appears once or twice, sporadic appearance is considered as normal background noise, and it is not worth upgrading the risk level, so no risk amplification processing is performed; but if it continues to appear repeatedly, it means that it can be a real signal of a systematic and trend problem, so a proportion threshold is set, that is, the target combination label must reach a certain frequency of appearance in the recent period to be considered worthy of attention, and then risk amplification processing is performed;

[0072] Meanwhile, the specific implementation steps of step S32 for judging whether the current combination label exists in the dangerous combination label library and calculating the working condition risk comprehensive index according to the judgment result further include step S323.

[0073] Step S323, judging whether the current combination label exists in the dangerous combination label library, if not, the synergistic influence factor is 0, and the working condition risk comprehensive index calculation formula includes:

[0074]

[0075] In the formula, R represents the working condition risk comprehensive index; represents the operation mode score; represents the environment mode score; represents the operation mode score threshold; represents the environment mode score threshold, represents the synergistic risk multiplier; represents the synergistic influence factor.

[0076] In steps S321-S323, first, by introducing the synergistic risk multiplier mechanism, the coupling effect of the operation mode and the environment condition is effectively identified, and the limitation of single-dimensional risk assessment in the traditional method is solved. When a high-risk combination label is detected, the risk amplification mechanism is automatically triggered to accurately reflect the potential harm of the compound fault. Secondly, the abnormal proportion of the time dimension is introduced, and the threshold analysis of the abnormal proportion is also performed. Finally, the multi-dimensional device state information is converted into a unified quantitative index, which provides a basis for formulating a differentiated health management strategy.

[0077] Embodiment 2

[0078] As shown in Figure 2 , the embodiment provides an industrial equipment state recognition and health management system, which comprises an acquisition module 1, a training module 2 and a management module 3.

[0079] The acquisition module 1 is used for acquiring comprehensive data of the industrial equipment at each historical moment in a preset first historical period, and the comprehensive data includes operation data and environment data.

[0080] The training module 2 is configured to calculate feature information according to the operation data and the environment data at each historical moment, filter all the feature information to obtain target feature information, search for comprehensive data corresponding to the target feature information, train a model based on the comprehensive data corresponding to the target feature information, and obtain an industrial equipment state information recognition model.

[0081] The management module 3 is configured to input the comprehensive data of the industrial equipment at the current moment into the industrial equipment state recognition model to obtain state information corresponding to the industrial equipment at the current moment, and generate a health management method according to the state information.

[0082] In one specific embodiment of the present disclosure, the training module 2 further includes a splicing unit 21 and a combination unit 22.

[0083] The splicing unit 21 is configured to splice the operation data and the environment data at each historical moment, extract feature information of the spliced data, and obtain feature information corresponding to each historical moment; based on a full linkage algorithm, initial clustering of all the feature information is performed by setting a first similarity threshold to obtain a plurality of initial state clustering clusters; all the initial state clustering clusters are traversed to find an initial state clustering cluster containing the most feature information, and the number of feature information contained in the initial state clustering cluster is recorded as n;

[0084] The combination unit 22 is configured to calculate a bit number m of n, create m continuous number intervals based on the bit number m, wherein when i < m, the ith interval is [10 (i-1) , 10 i - 1], and when i = m, the ith interval is [10 (m-1) , n]; analyze the number of feature information contained in each initial state clustering cluster, combine initial state clustering clusters having the number of feature information in the same number interval to obtain combination clusters, record combination clusters containing the number of initial state clustering clusters not less than 3 as target combination clusters, and filter target feature information according to the target combination clusters.

[0085] In one specific embodiment of the present disclosure, the combination unit 22 further includes a first calculation unit 221 and a second calculation unit 222.

[0086] The first calculation unit 221 is configured to, for each target combination cluster, count the number of feature information contained in each initial state clustering cluster in the target combination cluster, sort the initial state clustering clusters in descending order of the number, and select the first three initial state clustering clusters as candidate clustering clusters after sorting; calculate mean feature information of all the feature information contained in each candidate clustering cluster, calculate the similarity between the mean feature information corresponding to each pair of candidate clustering clusters, and merge two candidate clustering clusters corresponding to the maximum similarity to obtain a target clustering cluster.

[0087] The second calculation unit 222 is configured to, for each target clustering cluster, record remaining feature information not belonging to the target clustering cluster as first feature information; calculate similarity between each first feature information and each feature information in each target clustering cluster; record a maximum value in the obtained similarity as a belonging similarity of the first feature information with respect to the target clustering cluster; and record first feature information corresponding to a belonging similarity greater than a second similarity threshold as target feature information, the second similarity threshold being less than the first similarity threshold.

[0088] In one specific embodiment of the present disclosure, the training module 2 further comprises a collection unit 23.

[0089] The collection unit 23 is configured to collect running data in comprehensive data corresponding to all target feature information to obtain a first collection; cluster the running data in the first collection to obtain a plurality of first clustering clusters; for each first clustering cluster, assign a running mode label to each first clustering cluster, the label of the running data being the same as the label of the first clustering cluster to which the running data belongs; find environment data corresponding to each running data in the first clustering cluster, wherein the running data and the environment data corresponding thereto are acquired at the same time; collect all environment data corresponding to each first clustering cluster to obtain a second collection; cluster the environment data in each second collection to obtain a second clustering result; assign an environment mode label to each second clustering cluster, the label of the environment data being the same as the label of the second clustering cluster to which the environment data belongs; and train an industrial equipment state information recognition model according to the running data, the environment data, the running mode label and the environment mode label corresponding to all target feature information.

[0090] In one specific embodiment of the present disclosure, the collection unit 23 further comprises a training unit 231.

[0091] The training unit 231 is configured to, for comprehensive data corresponding to all target feature information, find running data and environment data at the same historical moment; splice the running data and the environment data into a complex sample, the running mode label corresponding to the running data and the environment mode label corresponding to the environment data being labels corresponding to the complex sample; splice running data and environment data corresponding to each feature information except the target feature information into an ordinary sample; cluster all ordinary samples by using a clustering algorithm to obtain a plurality of third clustering clusters; match a running mode label and an environment mode label to each third clustering cluster, the running mode label and the environment mode label of the ordinary sample being the same as the running mode label and the environment mode label of the third clustering cluster to which the ordinary sample belongs; and train a convolutional neural network model by using the complex sample and the ordinary sample with the running mode label and the environment mode label to obtain an industrial equipment state information recognition model.

[0092] It should be noted that the specific manner in which the various modules perform operations in the system in the above embodiments has been described in detail in the embodiments related to the method, and will not be described in detail here.

[0093] Embodiment 3

[0094] Corresponding to the above method embodiments, the embodiments of the disclosure also provide an industrial equipment state identification and health management device. The industrial equipment state identification and health management device described below can be mutually corresponding with reference to the industrial equipment state identification and health management method described above.

[0095] Figure 3 is a block diagram of an industrial equipment state identification and health management device 300 according to an exemplary embodiment. As shown, the industrial equipment state identification and health management device 300 can include a processor 301, a memory 302. The industrial equipment state identification and health management device 300 can also include one or more of a multimedia component 303, an I / O interface 304, and a communication component 305. Figure 3

[0096] ​The processor 301 is configured to control overall operation of the industrial equipment state identification and health management device 300 to accomplish all or part of the steps of the above-mentioned industrial equipment state identification and health management method. The memory 302 is configured to store various types of data to support operation of the industrial equipment state identification and health management device 300, which can include, for example, instructions for any application or method operating on the industrial equipment state identification and health management device 300, and application-related data, such as contact data, sent and received messages, pictures, audio, video, and the like. The memory 302 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic storage, a flash memory, a magnetic disk or an optical disk. The multimedia component 303 can include a screen and an audio component. The screen can be, for example, a touch screen, and the audio component is configured to output and / or input audio signals. For example, the audio component can include a microphone configured to receive external audio signals. The received audio signals can be further stored in the memory 302 or transmitted through the communication component 305. The audio component also includes at least one speaker configured to output audio signals. The I / O interface 304 provides an interface between the processor 301 and other interface modules, which can be a keyboard, a mouse, a button, and the like. The buttons can be virtual buttons or physical buttons. The communication component 305 is configured to enable wired or wireless communication between the industrial equipment state identification and health management device 300 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, near field communication (NFC), 2G, 3G or 4G, or a combination of one or more of them, so the corresponding communication component 305 can include a Wi-Fi module, a Bluetooth module, an NFC module.

[0097] In an example embodiment, the industrial equipment state recognition and health management device 300 can be implemented by one or more Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), controller, microcontroller, microprocessor or other electronic elements for executing the above-mentioned industrial equipment state recognition and health management method.

[0098] In another example embodiment, a computer readable storage medium including program instructions that, when executed by a processor, implement the steps of the above-mentioned industrial equipment state recognition and health management method is also provided. For example, the computer readable storage medium can be the above-mentioned memory 302 including program instructions, which can be executed by the processor 301 of the industrial equipment state recognition and health management device 300 to complete the above-mentioned industrial equipment state recognition and health management method.

[0099] Embodiment 4

[0100] Corresponding to the above method embodiments, the embodiments of the present disclosure also provide a readable storage medium, which can be referred to below in conjunction with the above-mentioned industrial equipment state recognition and health management method.

[0101] A readable storage medium, on which a computer program is stored, the computer program being executed by a processor to implement the steps of the industrial equipment state recognition and health management method of the above-mentioned method embodiments.

[0102] The readable storage medium can specifically be a U disk, a mobile hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and various readable storage media that can store program codes.

[0103] The above only describes preferred embodiments of the present disclosure and is not used to limit the present disclosure. For those skilled in the art, the present disclosure can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for industrial equipment status identification and health management, characterized in that, include: Acquire comprehensive data of industrial equipment at each historical moment within a preset first historical period. The comprehensive data includes operational data and environmental data. The operational data and environmental data at each historical moment are concatenated, and the feature information of the concatenated data is extracted to obtain the feature information corresponding to each historical moment. Based on the full-chain algorithm, by setting a first similarity threshold, all feature information is initially clustered to obtain multiple initial state clusters. All initial state clusters are traversed, and the initial state cluster containing the most feature information is found, and the number of feature information contained in it is recorded as n. The number of digits m of n is calculated, and based on the number of digits m, m consecutive quantity intervals are created, where when i < m, the i-th interval is [10]. (i-1) , 10 i -1], when i = m, the i-th interval is [10 (m-1) [, n]; Analyze the range of the number of feature information contained in each initial state cluster, and combine the initial state clusters with the same number of feature information to obtain a combined cluster. The combined cluster containing at least 3 initial state clusters is recorded as the target combined cluster. For each target combined cluster, count the number of feature information contained in each initial state cluster in the target combined cluster, and sort the initial state clusters in descending order of the number of feature information. Select the top 3 initial state clusters as candidate clusters. Calculate the mean feature information of all feature information contained in each candidate cluster, calculate the similarity between the mean feature information of each pair of candidate clusters, and select the most... Two candidate clusters with high similarity are merged to obtain the target cluster. For each target cluster, the remaining feature information that does not belong to the target cluster is recorded as the first feature information. The similarity between each first feature information and each feature information in each target cluster is calculated. The maximum value of the obtained similarity is recorded as the belonging similarity of the first feature information relative to the target cluster. The first feature information corresponding to the belonging similarity greater than the second similarity threshold is recorded as the target feature information. The second similarity threshold is less than the first similarity threshold. The comprehensive data corresponding to the target feature information is found. The model is trained based on the comprehensive data corresponding to the target feature information to obtain the industrial equipment status information recognition model. The comprehensive data of the industrial equipment at the current moment is input into the industrial equipment status identification model to obtain the status information of the industrial equipment at the current moment, and a health management method is generated based on the status information.

2. The industrial equipment status identification and health management method according to claim 1, characterized in that, Find the comprehensive data corresponding to the target feature information, train the model based on the comprehensive data corresponding to the target feature information, and obtain the industrial equipment status information recognition model, including: The operational data from the comprehensive data corresponding to all target feature information are aggregated to obtain a first set. The operational data in the first set are clustered to obtain multiple first clusters. For each first cluster, an operational mode label is assigned, and the label of the operational data is the same as the label of the first cluster to which it belongs. The environmental data corresponding to each operational data in the first cluster is found, wherein the acquisition time of the operational data and the corresponding environmental data is the same. All environmental data corresponding to each first cluster are aggregated to obtain a second set. The environmental data in each second set are clustered to obtain a second clustering result. An environmental mode label is assigned to each second cluster, and the label of the environmental data is the same as the label of the second cluster to which it belongs. An industrial equipment status information recognition model is trained based on the operational data, environmental data, operational mode labels, and environmental mode labels corresponding to all target feature information.

3. The industrial equipment status identification and health management method according to claim 2, characterized in that, An industrial equipment status information recognition model is trained based on the operational data, environmental data, operational mode labels, and environmental mode labels corresponding to all target feature information, including: For the comprehensive data corresponding to all target feature information, the operational data and environmental data at the same historical moment are searched. The operational data and environmental data are concatenated into a complex sample, and the operational mode label corresponding to the operational data and the environmental mode label corresponding to the environmental data are used as the label of the complex sample. The operational data and environmental data corresponding to each feature information other than the target feature information are concatenated into a normal sample. Clustering algorithms are used to cluster all normal samples to obtain multiple third clusters. Operational mode labels and environmental mode labels are matched for each third cluster. The operational mode labels and environmental mode labels of the normal samples are the same as those of the third cluster to which they belong. A convolutional neural network model is trained using the complex samples and normal samples with operational mode labels and environmental mode labels to obtain the industrial equipment status information recognition model.

4. An industrial equipment status identification and health management system, characterized in that, include: The acquisition module is used to acquire comprehensive data of industrial equipment at each historical moment within a preset first historical period. The comprehensive data includes operational data and environmental data. The training module is used to concatenate the runtime data and environmental data at each historical moment, extract the feature information of the concatenated data, and obtain the feature information corresponding to each historical moment; based on the full chain algorithm, by setting a first similarity threshold, all feature information is initially clustered to obtain multiple initial state clusters; all initial state clusters are traversed to find the initial state cluster containing the most feature information, and the number of feature information contained in it is recorded as n; the number of digits m of n is calculated, and based on the number of digits m, m consecutive intervals of quantity are created, where when i < m, the i-th interval is [10]. (i -1) , 10 i -1], when i = m, the i-th interval is [10 (m-1) [, n]; Analyze the range of the number of feature information contained in each initial state cluster, and combine the initial state clusters with the same number of feature information to obtain a combined cluster. The combined cluster containing at least 3 initial state clusters is recorded as the target combined cluster. For each target combined cluster, count the number of feature information contained in each initial state cluster in the target combined cluster, and sort the initial state clusters in descending order of the number of feature information. Select the top 3 initial state clusters as candidate clusters. Calculate the mean feature information of all feature information contained in each candidate cluster, calculate the similarity between the mean feature information of each pair of candidate clusters, and select the most... Two candidate clusters with high similarity are merged to obtain the target cluster. For each target cluster, the remaining feature information that does not belong to the target cluster is recorded as the first feature information. The similarity between each first feature information and each feature information in each target cluster is calculated. The maximum value of the obtained similarity is recorded as the belonging similarity of the first feature information relative to the target cluster. The first feature information corresponding to the belonging similarity greater than the second similarity threshold is recorded as the target feature information. The second similarity threshold is less than the first similarity threshold. The comprehensive data corresponding to the target feature information is found. The model is trained based on the comprehensive data corresponding to the target feature information to obtain the industrial equipment status information recognition model. The management module is used to input the comprehensive data of the industrial equipment at the current moment into the industrial equipment status identification model to obtain the status information of the industrial equipment at the current moment, and generate a health management method based on the status information.

5. The industrial equipment status identification and health management system according to claim 4, characterized in that, The training module includes: The aggregation unit is used to aggregate the operational data from the comprehensive data corresponding to all target feature information to obtain a first set; the operational data in the first set is clustered to obtain multiple first clusters; for each first cluster, an operational mode label is assigned to each first cluster, and the label of the operational data is the same as the label of the first cluster to which it belongs; the environmental data corresponding to each operational data in the first cluster is found, wherein the acquisition time of the operational data and the corresponding environmental data is the same; all environmental data corresponding to each first cluster are aggregated to obtain a second set; the environmental data in each second set are clustered to obtain a second clustering result, and an environmental mode label is assigned to each second cluster, and the label of the environmental data is the same as the label of the second cluster to which it belongs; the industrial equipment status information recognition model is trained based on the operational data, environmental data, operational mode labels, and environmental mode labels corresponding to all target feature information.

6. The industrial equipment status identification and health management system according to claim 5, characterized in that, The set of units includes: The training unit is used to find the operational and environmental data at the same historical moment for the comprehensive data corresponding to all target feature information. The operational and environmental data are concatenated into a complex sample, and the operational mode label corresponding to the operational data and the environmental mode label corresponding to the environmental data are used as the label of the complex sample. The operational and environmental data corresponding to each feature information other than the target feature information are concatenated into a normal sample. Clustering algorithms are used to cluster all normal samples to obtain multiple third clusters. Operational mode labels and environmental mode labels are matched for each third cluster. The operational mode labels and environmental mode labels of the normal samples are the same as those of the third cluster to which they belong. The convolutional neural network model is trained using the complex samples and normal samples with operational mode labels and environmental mode labels to obtain the industrial equipment status information recognition model.

Images