CN115906642B

CN115906642B - Bearing production detection control method and device

Info

Publication number: CN115906642B
Application number: CN202211496474.0A
Authority: CN
Inventors: 胡金福; 吕文轩; 周学兵
Original assignee: Dongguan Keda Hardware Products Co ltd
Current assignee: Dongguan Keda Hardware Products Co ltd
Priority date: 2022-11-28
Filing date: 2022-11-28
Publication date: 2023-07-28
Anticipated expiration: 2042-11-28
Also published as: CN115906642A

Abstract

The invention provides a bearing production detection control method and device, and relates to the technical field of data processing. In the invention, the operation data of the bearing production equipment are extracted; loading the operation data of the bearing production equipment into an operation data analysis neural network of the optimizing equipment to perform operation data feature mining processing so as to output operation data mining feature representation, and performing operation data feature restoration processing on the operation data mining feature representation so as to output key equipment operation behaviors, key behavior semantic data and key equipment operation behavior representative data; and analyzing and outputting a first equipment operation behavior abnormality analysis result according to the key equipment operation behavior, the key behavior semantic data and the key equipment operation behavior representation data, and determining quality representation data of the bearing product correspondingly produced according to the first equipment operation behavior abnormality analysis result. Based on the above, the convenience and reliability of the bearing production detection can be improved.

Description

Bearing production detection control method and device

Technical Field

The invention relates to the technical field of data processing, in particular to a bearing production detection control method and device.

Background

The bearing has more application scenes and higher quality requirements, so that the bearing needs to be detected and controlled, such as quality detection and the like, in the production and processing process of the bearing. However, in the prior art, quality monitoring is generally implemented based on video monitoring, that is, corresponding monitoring equipment is deployed to production equipment of bearings or produced bearings, so that the problem of low convenience of production detection exists.

Disclosure of Invention

In view of the above, the present invention is directed to a method and a device for detecting and controlling bearing production, so as to improve the convenience and reliability of bearing production detection.

In order to achieve the above purpose, the embodiment of the present invention adopts the following technical scheme:

a bearing production inspection control method, comprising:

extracting a bearing production detection instruction, wherein the bearing production detection instruction comprises bearing production equipment operation data corresponding to target bearing production equipment;

loading the bearing production equipment operation data into an optimizing equipment operation data analysis neural network to perform operation data feature mining processing so as to output corresponding operation data mining feature representation, performing operation data feature restoration processing on the operation data mining feature representation so as to output corresponding key equipment operation behaviors, key behavior semantic data and key equipment operation behavior representative data, loading the typical bearing production equipment operation data into the equipment operation data analysis neural network to perform network optimization forming, performing operation data feature mining processing and operation data feature restoration processing through the equipment operation data analysis neural network to perform operation data feature mining processing and operation data feature restoration processing so as to output typical key equipment operation behaviors, typical key behavior semantic data and typical key equipment operation behavior representative data, analyzing and outputting corresponding equipment operation behavior abnormality analysis results according to the typical key equipment operation behaviors, the typical key operation behavior semantic data and the typical key equipment operation behavior representative data, performing operation behavior feature mining feature representation so as to output corresponding typical key equipment operation behavior analysis results, determining actual operation behavior analysis values according to the equipment operation behavior configuration analysis results and learning values, determining the actual operation abnormality analysis values according to the equipment operation behavior configuration analysis results, performing network optimization on the equipment operation data analysis neural network to be optimized according to the anomaly analysis learning cost value and the behavior criticality analysis learning cost value to form a corresponding optimized equipment operation data analysis neural network;

And analyzing and outputting a corresponding first equipment operation behavior abnormality analysis result according to the key equipment operation behavior, the key behavior semantic data and the key equipment operation behavior representation data, and determining quality representation data of the bearing product correspondingly produced by the target bearing production equipment according to the first equipment operation behavior abnormality analysis result.

In some preferred embodiments, in the above bearing production detection control method, the optimization apparatus operation data analysis neural network includes an optimization data feature mining model and an optimization data feature restoration model; the step of loading the operation data of the bearing production equipment into an operation data analysis neural network of the optimizing equipment to perform operation data feature mining processing so as to output corresponding operation data mining feature representations, and then performing operation data feature restoration processing on the operation data mining feature representations so as to output corresponding operation behavior of the key equipment, semantic data of the key behavior and operation behavior representative data of the key equipment, comprises the following steps:

loading the operation data of the bearing production equipment into the optimized data feature mining model to perform operation data feature mining processing so as to output corresponding operation data mining feature representation;

And loading the operation data mining feature representation into the optimized data feature reduction model to perform operation data feature reduction processing so as to output corresponding key equipment operation behaviors, key behavior semantic data and key equipment operation behavior representative data.

In some preferred embodiments, in the above-described bearing production detection control method, further comprising:

extracting typical bearing production equipment operation data, equipment operation behavior actual results and key equipment operation behavior actual results;

loading the operation data of the typical bearing production equipment into an operation data analysis neural network of the equipment to be optimized for operation data feature mining processing so as to output corresponding typical operation data mining feature representations, and then carrying out operation data feature restoration processing on the typical operation data mining feature representations so as to output corresponding typical key equipment operation behaviors, typical key behavior semantic data and typical key equipment operation behavior representative data;

analyzing and outputting corresponding equipment operation behavior abnormality analysis results according to the typical key equipment operation behaviors, the typical key behavior semantic data and the typical key equipment operation behavior representative data, and carrying out operation behavior criticality analysis processing on the typical operation data mining characteristic representation so as to output corresponding typical key equipment operation behavior analysis results;

Analyzing and determining the learning cost value according to the equipment operation behavior abnormal analysis result and the equipment operation behavior actual result so as to output a corresponding abnormal analysis learning cost value, and analyzing and determining the learning cost value according to the key equipment operation behavior actual result and the typical key equipment operation behavior analysis result so as to output a corresponding behavior criticality analysis learning cost value;

and performing network optimization on the equipment operation data analysis neural network to be optimized according to the abnormal analysis learning cost value and the behavior critical analysis learning cost value to form a candidate equipment operation data analysis neural network, re-marking the candidate equipment operation data analysis neural network as the equipment operation data analysis neural network to be optimized, executing the steps of extracting the typical bearing production equipment operation data, the equipment operation behavior actual results and the key equipment operation behavior actual results in a rotary mode, and forming the optimized equipment operation data analysis neural network under the condition that a network optimization target is reached.

In some preferred embodiments, in the above bearing production detection control method, the step of loading the typical bearing production equipment operation data into an equipment operation data analysis neural network to be optimized to perform operation data feature mining processing to output a corresponding typical operation data mining feature representation, and performing operation data feature restoration processing on the typical operation data mining feature representation to output a corresponding typical critical equipment operation behavior, typical critical behavior semantic data, and typical critical equipment operation behavior representative data includes:

Carrying out ordered processing on the operation data of the typical bearing production equipment to form a corresponding ordered set of operation behaviors of the typical equipment, and loading the ordered set of operation behaviors of the typical equipment to an equipment operation data analysis neural network to be optimized;

analyzing a neural network by using the equipment operation data to be optimized, and mining data mining feature representations corresponding to the ordered set of typical equipment operation behaviors so as to output corresponding typical operation data mining feature representations;

analyzing a neural network by using the equipment operation data to be optimized, performing operation data feature restoration processing on the typical operation data mining feature representation to output corresponding typical key equipment operation behaviors, typical key behavior semantic data and candidate equipment operation behavior representation data, and determining corresponding operation behavior indication data of each key equipment according to the typical key equipment operation behaviors;

and analyzing a neural network by using the equipment operation data to be optimized, carrying out data comparison processing on the candidate equipment operation behavior representing data and each key equipment operation behavior indicating data, and carrying out marking processing on the candidate equipment operation behavior representing data under the condition that the candidate equipment operation behavior representing data is consistent with the key equipment operation behavior indicating data so as to form corresponding typical key equipment operation behavior representing data.

In some preferred embodiments, in the above method for detecting and controlling bearing production, the step of performing an ordering process on the running data of the typical bearing production equipment to form a corresponding ordered set of running behaviors of the typical equipment, and loading the ordered set of running behaviors of the typical equipment into a neural network for analyzing the running data of the equipment to be optimized includes:

performing operation behavior separation processing on the operation data of the typical bearing production equipment to output corresponding operation behaviors of each typical equipment;

extracting configured head-end operation behavior indication information and tail-end operation behavior indication information, combining the head-end operation behavior indication information, the tail-end operation behavior indication information and each typical equipment operation behavior according to a precedence relationship in the typical bearing production equipment operation data to form a corresponding typical equipment operation behavior ordered set, and loading the typical equipment operation behavior ordered set into an equipment operation data analysis neural network to be optimized.

In some preferred embodiments, in the bearing production detection control method, the step of analyzing the neural network by using the device operation data to be optimized, performing a data comparison process on the candidate device operation behavior representing data and each of the key device operation behavior indicating data, and in a case that the candidate device operation behavior representing data coincides with the key device operation behavior indicating data, performing a marking process on the candidate device operation behavior representing data to form corresponding typical key device operation behavior representing data includes:

Constructing a corresponding key equipment operation behavior indication data knowledge graph according to each key equipment operation behavior indication data;

screening the candidate equipment operation behavior representing data from the key equipment operation behavior indication data knowledge graph, and marking the candidate equipment operation behavior representing data under the condition that the candidate equipment operation behavior representing data is screened to form corresponding typical key equipment operation behavior representing data.

In some preferred embodiments, in the above bearing production inspection control method, the representative operational data mining feature representation includes each representative operational behavior mining feature representation;

the step of performing operation behavior criticality analysis processing on the typical operation data mining feature representation to output a corresponding typical key equipment operation behavior analysis result comprises the following steps:

traversing in sequence in each representative operation behavior mining feature representation to form traversed representative operation behavior mining feature representations;

performing mapping output processing on the traversed typical operation behavior mining feature representation to output a corresponding mapping operation behavior mining feature representation, and performing operation behavior criticality analysis processing on the mapping operation behavior mining feature representation to output typical key equipment operation behavior analysis probability corresponding to the traversed typical operation behavior mining feature representation;

Analyzing the traversed typical operation behavior mining features to represent corresponding traversed typical operation behavior analysis results according to the typical key equipment operation behavior analysis probability;

and representing corresponding typical operation behavior analysis results according to each typical operation behavior mining characteristic to form corresponding typical key equipment operation behavior analysis results.

In some preferred embodiments, in the above bearing production detection control method, the step of performing analysis determination of the learning cost value according to the actual result of the operation behavior of the key device and the analysis result of the operation behavior of the typical key device to output a corresponding behavior criticality analysis learning cost value includes:

analyzing and determining the learning cost value of the corresponding typical operation behavior analysis result and the corresponding equipment operation behavior actual result in the key equipment operation behavior actual result according to each typical operation behavior mining feature representation so as to output the corresponding local behavior key analysis learning cost value;

and analyzing and outputting the corresponding behavior criticality analysis learning cost value according to each local behavior criticality analysis learning cost value.

In some preferred embodiments, in the bearing production detection control method, the device operation data analysis neural network to be optimized includes a data feature mining model to be optimized, a data feature restoring model to be optimized, and a critical analysis processing model to be optimized;

loading the operation data of the typical bearing production equipment into the data feature mining model to be optimized so as to perform operation data feature mining processing and output corresponding typical operation data mining feature representations;

loading the typical operation data mining feature representation into the data feature reduction model to be optimized so as to perform operation data feature reduction processing, outputting corresponding typical key equipment operation behaviors, typical key behavior semantic data and typical key equipment operation behavior representative data, and analyzing and outputting corresponding equipment operation behavior abnormality analysis results according to the typical key equipment operation behaviors, the typical key behavior semantic data and the typical key equipment operation behavior representative data;

loading the typical operation data mining characteristic representation into the to-be-optimized key analysis processing model to perform operation behavior key analysis processing, and outputting a corresponding typical key equipment operation behavior analysis result;

And constructing a corresponding optimizing equipment operation data analysis neural network according to the optimized data characteristic mining model to be optimized and the optimized data characteristic restoring model to be optimized.

The embodiment of the invention also provides a bearing production detection control device, which comprises a processor and a memory, wherein the memory is used for storing a computer program, and the processor is used for executing the computer program so as to realize the bearing production detection control method.

The bearing production detection control method and device provided by the embodiment of the invention can extract the operation data of bearing production equipment; loading the operation data of the bearing production equipment into an operation data analysis neural network of the optimizing equipment to perform operation data feature mining processing so as to output operation data mining feature representation, and performing operation data feature restoration processing on the operation data mining feature representation so as to output key equipment operation behaviors, key behavior semantic data and key equipment operation behavior representative data; and analyzing and outputting a first equipment operation behavior abnormality analysis result according to the key equipment operation behavior, the key behavior semantic data and the key equipment operation behavior representation data, and determining quality representation data of the bearing product correspondingly produced according to the first equipment operation behavior abnormality analysis result. Based on the foregoing, because the analysis processing is performed based on the operation data of the bearing production equipment, compared with the conventional technical scheme of performing video monitoring on the bearing or equipment, no additional monitoring equipment is required to be deployed, so that the detection convenience is higher.

In order to make the above objects, features and advantages of the present invention more comprehensible, preferred embodiments accompanied with figures are described in detail below.

Drawings

Fig. 1 is a block diagram of a bearing production detection control device according to an embodiment of the present invention.

Fig. 2 is a schematic flow chart of each step included in the bearing production detection control method according to the embodiment of the present invention.

Fig. 3 is a schematic diagram of each module included in the bearing production detection control system according to the embodiment of the present invention.

Description of the embodiments

For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, but not all embodiments of the present invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.

Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

As shown in fig. 1, an embodiment of the present invention provides a bearing production detection control device. Wherein, the bearing production detection control device can comprise a memory and a processor.

In detail, the memory and the processor are electrically connected directly or indirectly to realize transmission or interaction of data. For example, electrical connection may be made to each other via one or more communication buses or signal lines. The memory may store at least one software functional module (computer program) that may exist in the form of software or firmware. The processor may be configured to execute an executable computer program stored in the memory, thereby implementing a bearing production detection control method provided by an embodiment of the present invention (as described later).

It should be appreciated that in some embodiments, the Memory may be, but is not limited to, random access Memory (Random Access Memory, RAM), read Only Memory (ROM), programmable Read Only Memory (Programmable Read-Only Memory, PROM), erasable Read Only Memory (Erasable Programmable Read-Only Memory, EPROM), electrically erasable Read Only Memory (Electric Erasable Programmable Read-Only Memory, EEPROM), and the like. The processor may be a general purpose processor including a central processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), a System on Chip (SoC), etc.; but also Digital Signal Processors (DSPs), application Specific Integrated Circuits (ASICs), field Programmable Gate Arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

It should be appreciated that in some embodiments, the bearing production detection control device may be a server with data processing capabilities.

With reference to fig. 2, the embodiment of the invention further provides a bearing production detection control method, which can be applied to the bearing production detection control device. The method steps defined by the flow related to the bearing production detection control method can be realized by the bearing production detection control device.

The specific flow shown in fig. 2 will be described in detail.

And step S110, extracting a bearing production detection instruction.

In the embodiment of the invention, the bearing production detection control device can extract a bearing production detection instruction. The bearing production detection instruction comprises bearing production equipment operation data corresponding to the target bearing production equipment (such as extraction from the operation log data of the bearing production equipment).

Step S120, loading the operation data of the bearing production equipment into an operation data analysis neural network of the optimizing equipment to perform operation data feature mining processing so as to output corresponding operation data mining feature representations, and performing operation data feature restoration processing on the operation data mining feature representations so as to output corresponding operation behavior of the key equipment, semantic data of the key behavior and operation behavior representative data of the key equipment.

In the embodiment of the invention, the bearing production detection control device can load the bearing production equipment operation data into the optimizing equipment operation data analysis neural network to perform operation data feature mining processing so as to output corresponding operation data mining feature representations, and then perform operation data feature reduction processing on the operation data mining feature representations so as to output corresponding key equipment operation behaviors, key behavior semantic data and key equipment operation behavior representative data. The optimizing equipment operation data analysis neural network is formed by loading typical bearing production equipment operation data into an equipment operation data analysis neural network to be optimized for network optimization, performing operation data feature mining processing and operation data feature reduction processing through the equipment operation data analysis neural network to be optimized for outputting typical key equipment operation behaviors, typical key behavior semantic data and typical key equipment operation behavior representative data, then analyzing and outputting corresponding equipment operation behavior exception analysis results according to the typical key equipment operation behaviors, the typical key behavior semantic data and the typical key equipment operation behavior representative data, performing operation behavior criticality analysis processing on the outputted typical operation data mining feature representation for outputting corresponding typical key equipment operation behavior analysis results, and then performing learning cost value analysis determination according to the equipment operation exception analysis results and configured equipment operation behavior actual results for outputting corresponding exception analysis learning cost value, and performing learning cost value analysis determination according to the configured key equipment operation behavior actual results and the typical key equipment operation behavior analysis results for outputting corresponding behavior exception analysis results, and performing learning cost value analysis on the neural network to be optimized according to the equipment operation data to be optimized.

Step S130, analyzing and outputting a corresponding first equipment operation behavior abnormality analysis result according to the key equipment operation behavior, the key behavior semantic data and the key equipment operation behavior representation data, and determining quality representation data of the bearing product correspondingly produced by the target bearing production equipment according to the first equipment operation behavior abnormality analysis result.

In the embodiment of the invention, the bearing production detection control device can analyze and output a corresponding first device operation behavior abnormality analysis result according to the key device operation behavior, the key operation behavior semantic data and the key device operation behavior representation data, and then determine quality representation data (for example, the quality of a produced bearing is abnormal due to the fact that the operation behavior of the device is abnormal) of a bearing product correspondingly produced by the target bearing production device according to the first device operation behavior abnormality analysis result, in some applications, the quality representation data determined based on the first device operation behavior abnormality analysis result can be used as quality representation data of one dimension or initial quality representation data, and therefore, further quality analysis and other processing can be performed by combining data of other dimensions.

Based on the foregoing, because the analysis processing is performed based on the operation data of the bearing production equipment, compared with the conventional technical scheme of performing video monitoring on the bearing or equipment, no additional monitoring equipment is required to be deployed, so that the detection convenience is higher.

It should be understood that, in some embodiments, step S110 above may include some of the following sub-steps (i.e., detailed implementation) that may be implemented:

extracting a bearing production detection instruction, extracting bearing production equipment operation data from the bearing production detection instruction, and performing staged decomposition processing on the bearing production equipment operation data to form a plurality of corresponding original equipment operation behavior log data, wherein each original equipment operation behavior log data corresponds to one production stage (or production procedure, production link, etc.) of a bearing;

Performing feature mining (feature extraction, etc.) on the original equipment operation log data for each of the original equipment operation log data to obtain a log data feature representation corresponding to the original equipment operation log data, performing feature representation similarity calculation on the log data feature representation corresponding to the original equipment operation log data and the log data feature representation corresponding to each adjacent original equipment operation log data of the original equipment operation log data to output a corresponding feature representation similarity, and performing feature association processing on the log data feature representation corresponding to each adjacent original equipment operation log data and the log data feature representation corresponding to the original equipment operation log data according to the feature representation similarity to obtain an associated log data feature representation of the original equipment operation log data (illustratively, weighting the log data feature representation corresponding to the adjacent original equipment operation log data based on the corresponding feature representation similarity, and then performing feature stacking on each weighted log data feature representation and the log data feature representation corresponding to obtain an associated log data feature representation of the original equipment operation log data;

For each piece of original equipment operation behavior log data, determining standard equipment operation behavior log data corresponding to the original equipment operation behavior log data (the standard equipment operation behavior log data can be preconfigured), and determining associated standard log data characteristic representation corresponding to the standard equipment operation behavior log data (determining mode is as described above);

and for each piece of original equipment operation behavior log data, performing feature representation distance calculation processing (the distance can be inversely related to the similarity) on the associated log data feature representation of the original equipment operation behavior log data and the associated standard log data feature representation corresponding to the standard equipment operation behavior log data corresponding to the original equipment operation behavior log data so as to obtain the feature representation distance corresponding to the original equipment operation behavior log data (the feature representation distance can be used for aggregation, such as weighted summation, of equipment operation behavior anomalies obtained through subsequent calculation so as to obtain final equipment operation behavior anomalies, thereby realizing reliable analysis of the behavior anomalies).

It should be appreciated that, in some embodiments, the optimization device operation data analysis neural network includes an optimization data feature mining model and an optimization data feature restoration model, based on which, step S120 above may include the following sub-steps (i.e., detailed implementation process) that may be implemented:

It should be understood that, in some embodiments, the above bearing production detection control method may further include the following steps (i.e., detailed implementation process) that may be implemented:

extracting typical bearing production equipment operation data, an equipment operation behavior actual result and a key equipment operation behavior actual result (exemplarily, the typical bearing production equipment operation data, the equipment operation behavior actual result and the key equipment operation behavior actual result are taken as optimization basis combinations, the equipment operation behavior actual result is used for reflecting real behavior abnormality results of the typical bearing production equipment operation data, namely, whether abnormality or abnormality degree and the like, and the equipment operation behavior actual result can be integrated or can be respectively used for reflecting actual key equipment operation behaviors in the typical bearing production equipment operation data according to whether abnormality or abnormality degree and the like of each equipment operation behavior, and the key equipment operation behavior actual result can be formed through manual labeling);

Loading the typical bearing production equipment operation data into an equipment operation data analysis neural network to be optimized to perform operation data feature mining processing so as to output corresponding typical operation data mining feature representation, and performing operation data feature restoration processing on the typical operation data mining feature representation so as to output corresponding typical key equipment operation behaviors, typical key behavior semantic data and typical key equipment operation behavior representative data (for example, the typical key equipment operation behaviors may refer to key behaviors in a plurality of typical equipment operation behaviors included in the typical bearing production equipment operation data, the typical key behavior semantic data may be used for reflecting behavior attribute information of the typical key equipment operation behaviors and may represent anomaly degree, and the typical key equipment operation behavior representative data may be used for reflecting influence of the typical key equipment operation behaviors on subsequent equipment operation behaviors);

according to the typical key device operation behaviors, the typical key behavior semantic data and the typical key device operation behavior representation data, analyzing and outputting corresponding device operation behavior abnormality analysis results (for example, the typical operation data mining characteristics can be used for carrying out operation behavior criticality analysis processing on the basis of the influence of the previous typical key device operation behaviors, fusing the behavior attribute information of the previous typical key device operation behaviors and the current typical key device operation behaviors to obtain fused behavior attribute information corresponding to the current typical key device operation behaviors, namely obtaining device operation behavior abnormality analysis results;

and performing network optimization on the equipment operation data analysis neural network to be optimized according to the abnormal analysis learning cost value and the behavior critical analysis learning cost value to form a candidate equipment operation data analysis neural network, re-marking the candidate equipment operation data analysis neural network as the equipment operation data analysis neural network to be optimized, executing the steps of extracting the typical bearing production equipment operation data, the equipment operation behavior actual results and the key equipment operation behavior actual results in a rotary mode, and forming the optimized equipment operation data analysis neural network under the condition that a network optimization target is reached (such as the learning cost value is smaller than a threshold value).

It should be appreciated that, in some embodiments, the above steps of loading the typical bearing production equipment operation data into the equipment operation data analysis neural network to be optimized to perform the operation data feature mining process to output the corresponding typical operation data mining feature representation, and then performing the operation data feature restoration process on the typical operation data mining feature representation to output the corresponding typical critical equipment operation behavior, typical critical behavior semantic data, and typical critical equipment operation behavior representative data may include the following sub-steps (i.e., detailed implementation process) that may be implemented:

utilizing the device operation data analysis neural network to be optimized to mine the data mining feature representations corresponding to the ordered set of typical device operation behaviors so as to output corresponding typical operation data mining feature representations (illustratively, each typical device operation behavior in the ordered set of typical device operation behaviors can be mined, namely, feature mining can be performed on the data mining feature representations so as to obtain the data mining feature representations corresponding to each typical device operation behavior, and then the data mining feature representations corresponding to each typical device operation behavior can be combined so as to form the data mining feature representations corresponding to the ordered set of typical device operation behaviors);

analyzing a neural network by using the equipment operation data to be optimized, performing operation data feature restoration processing on the typical operation data mining feature representation to output corresponding typical key equipment operation behaviors, typical key behavior semantic data and candidate equipment operation behavior representation data, and determining corresponding each key equipment operation behavior indication data according to the typical key equipment operation behaviors (each key equipment operation behavior indication data is used for reflecting at least one behavior influence of the typical key equipment operation behaviors and can be found out from historical behavior data);

Wherein, it should be understood that, in some embodiments, the step of analyzing the neural network by using the device operation data to be optimized, and performing the operation data feature restoration processing on the representative operation data mining feature representation to output the corresponding representative key device operation behavior, representative key behavior semantic data and candidate device operation behavior representative data may include the following sub-steps (i.e. detailed implementation process) that may be implemented:

analyzing a neural network by using the device operation data to be optimized, performing operation data feature restoration processing on the configured head-end operation behavior identification feature representation and the typical operation data mining feature representation to output a corresponding first head-end operation behavior identification feature representation (illustratively, the head-end operation behavior identification feature representation may be a feature representation of assigned identification information, which is used for indicating the first typical device operation behavior; in addition, the head-end operation behavior identification feature representation and the typical operation data mining feature representation may be fused, such as performing processing such as splicing or superposition, and then performing operation data feature restoration processing on a processing result, which may be implemented based on RNN or long short-term memory (LSTM), for example);

Analyzing a neural network by using the equipment operation data to be optimized, performing operation data characteristic restoration processing on the first head-end operation behavior identification characteristic representation and the typical operation data mining characteristic representation to output a corresponding first central operation behavior identification characteristic representation (illustratively, the first head-end operation behavior identification characteristic representation and the typical operation data mining characteristic representation can be fused first, then, the operation data characteristic restoration processing is performed on the fusion result to output a corresponding first central operation behavior identification characteristic representation;

analyzing a neural network by using the equipment operation data to be optimized, performing operation data characteristic restoration processing on the first central operation behavior identification characteristic representation and the typical operation data mining characteristic representation to output a corresponding first end operation behavior identification characteristic representation (in the case that the first central operation behavior identification characteristic representation is a plurality of, performing operation data characteristic restoration processing on the last first central operation behavior identification characteristic representation and the typical operation data mining characteristic representation to output a corresponding first end operation behavior identification characteristic representation);

Analyzing the neural network by using the device operation data to be optimized, analyzing the first head-end operation behavior identification feature representation, the first center operation behavior identification feature representation and the first tail-end operation behavior identification feature representation, outputting corresponding typical key device operation behaviors, typical key behavior semantic data and candidate device operation behavior representation data (based on the analysis result, each operation behavior identification feature representation fuses the previous operation behavior identification feature, namely, the influence of the previous device operation behaviors on the subsequent device operation behaviors is considered as a whole, so that the information of each operation behavior identification feature representation is richer and more reliable, and in addition, the first head-end operation behavior identification feature representation, each first center operation behavior identification feature representation and the first tail-end operation behavior identification feature representation can be spliced to form a spliced operation behavior identification feature representation, and then, the analysis neural network by using the device operation data to be optimized analyzes the spliced operation behavior identification feature representation, the corresponding typical key device operation behaviors, typical key behavior data and candidate device operation behavior representation data are estimated.

Wherein, it should be understood that, in some embodiments, the above steps of performing analysis determination of learning cost value according to the device operation behavior anomaly analysis result and the device operation behavior actual result to output corresponding anomaly analysis learning cost value may include the following sub-steps (i.e. detailed implementation process) that may be implemented:

analyzing and outputting a head-end operation behavior analysis learning cost value corresponding to the first head-end operation behavior identification feature representation according to the first head-end operation behavior identification feature representation and the equipment operation behavior actual result (illustratively, an equipment operation behavior abnormality analysis result corresponding to the equipment operation behavior can be analyzed and output according to the first head-end operation behavior identification feature representation, and then, a corresponding end operation behavior analysis learning cost value is analyzed and output based on the equipment operation behavior abnormality analysis result and the equipment operation behavior actual result corresponding to the equipment operation behavior);

analyzing and outputting a center operation behavior analysis learning cost value corresponding to the first center operation behavior identification characteristic representation according to the first center operation behavior identification characteristic representation and the device operation behavior actual result (analysis of corresponding device operation behaviors as described above);

Analyzing and outputting an end operation behavior analysis learning cost value corresponding to the first end operation behavior identification characteristic representation according to the first end operation behavior identification characteristic representation and the device operation behavior actual result (analysis of corresponding device operation behaviors as described above);

and analyzing and outputting corresponding abnormal analysis learning cost values according to the head-end operation behavior analysis learning cost value, the central operation behavior analysis learning cost value and the tail-end operation behavior analysis learning cost value (for example, the head-end operation behavior analysis learning cost value, the central operation behavior analysis learning cost value and the tail-end operation behavior analysis learning cost value can be weighted and summed to obtain corresponding abnormal analysis learning cost values).

It should be appreciated that, in some embodiments, the step of ordering the typical bearing production equipment operation data to form a corresponding ordered set of typical equipment operation behaviors, and loading the ordered set of typical equipment operation behaviors into the equipment operation data analysis neural network to be optimized, may include the following possible sub-steps (i.e., detailed implementation):

extracting configured head-end operation behavior indication information and tail-end operation behavior indication information, combining the head-end operation behavior indication information, the tail-end operation behavior indication information and each typical equipment operation behavior according to a precedence relationship in the typical bearing production equipment operation data to form a corresponding typical equipment operation behavior ordered set (in the typical equipment operation behavior ordered set, the head-end operation behavior indication information belongs to first information in the set, the tail-end operation behavior indication information belongs to last information in the set), and loading the typical equipment operation behavior ordered set into an equipment operation data analysis neural network to be optimized.

It should be understood that, in some embodiments, the step of analyzing the neural network by using the device operation data to be optimized, performing data comparison processing on the candidate device operation behavior representing data and each of the key device operation behavior indicating data, and, in the case that the candidate device operation behavior representing data corresponds to the key device operation behavior indicating data, performing marking processing on the candidate device operation behavior representing data to form corresponding typical key device operation behavior representing data, may include the following possible substeps (i.e. detailed implementation process):

It should be appreciated that, in some embodiments, the exemplary operational data mining feature representation includes each exemplary operational behavior mining feature representation, based on which the step of subjecting the exemplary operational data mining feature representation to operational behavior criticality analysis processing to output corresponding exemplary critical device operational behavior analysis results described above may include the following sub-steps (i.e., detailed implementation):

performing a sequential traversal in each of the representative behavior mining feature representations (e.g., traversing a first representative behavior mining feature representation, traversing a second representative behavior mining feature representation, traversing a third representative behavior mining feature representation, traversing a fourth representative behavior mining feature representation) to form a traversed representative behavior mining feature representation;

Performing mapping output processing on the traversed typical operation behavior mining feature representation to output a corresponding mapping operation behavior mining feature representation (illustratively, the typical operation behavior mining feature table may be weighted based on a first parameter matrix, then the weighted results may be summed based on a second parameter matrix to implement shifting to obtain a mapping operation behavior mining feature representation; in addition, the first parameter matrix and the second parameter matrix belong to optimization parameters of a neural network), performing operation behavior criticality analysis processing on the mapping operation behavior mining feature representation to output a typical key device operation behavior analysis probability corresponding to the traversed typical operation behavior mining feature representation (the typical key device operation behavior analysis probability may refer to a probability that the typical operation behavior mining feature representation corresponds to a key device operation behavior, that is, an estimated probability value; in addition, the mapping operation behavior mining feature representation may be processed based on a softmax function to output a corresponding typical key device operation behavior analysis probability);

It should be appreciated that, in some embodiments, the above step of performing the analysis determination of the learning cost value according to the actual result of the operation behavior of the key device and the analysis result of the operation behavior of the typical key device to output the corresponding operation criticality analysis learning cost value may include the following sub-steps (i.e., detailed implementation process) that may be implemented:

analyzing and determining a learning cost value for each representative operation behavior mining feature representation corresponding representative operation behavior analysis result and a corresponding device operation behavior actual result in the key device operation behavior actual results to output each corresponding local behavior criticality analysis learning cost value (namely, for each representative operation behavior mining feature representation, performing difference analysis for the representative operation behavior analysis result corresponding to the representative operation behavior mining feature representation and a device operation behavior actual result corresponding to the representative operation behavior representation in the key device operation behavior actual result to obtain a local behavior criticality analysis learning cost value);

And analyzing and outputting corresponding behavior criticality analysis learning cost values according to each local behavior criticality analysis learning cost value (illustratively, a sum technology can be carried out on each local behavior criticality analysis learning cost value to obtain the corresponding behavior criticality analysis learning cost value).

It should be appreciated that, in some embodiments, the device operational data analysis neural network to be optimized may include a data feature mining model to be optimized, a data feature restoration model to be optimized, and a critical analysis processing model to be optimized, based on which the foregoing steps may be implemented based on the data feature mining model to be optimized, the data feature restoration model to be optimized, and the critical analysis processing model to be optimized, and detailed implementation procedures thereof may include:

loading the operation data of the typical bearing production equipment into the data feature mining model to be optimized so as to perform operation data feature mining processing and output corresponding typical operation data mining feature representations; loading the typical operation data mining feature representation into the data feature reduction model to be optimized so as to perform operation data feature reduction processing, outputting corresponding typical key equipment operation behaviors, typical key behavior semantic data and typical key equipment operation behavior representative data, and analyzing and outputting corresponding equipment operation behavior abnormality analysis results according to the typical key equipment operation behaviors, the typical key behavior semantic data and the typical key equipment operation behavior representative data; loading the typical operation data mining characteristic representation into the to-be-optimized key analysis processing model to perform operation behavior key analysis processing, and outputting a corresponding typical key equipment operation behavior analysis result; according to the optimized data feature mining model and the optimized data feature restoring model, a corresponding optimizing equipment operation data analysis neural network is constructed (for example, an optimizing key analysis processing model in the optimized equipment operation data analysis neural network can be directly discarded to obtain the optimizing equipment operation data analysis neural network, and the optimizing equipment operation data analysis neural network can also be constructed according to model parameters of the optimized data feature mining model and the optimized data feature restoring model).

With reference to fig. 3, an embodiment of the present invention further provides a bearing production detection control system, which is applicable to the above bearing production detection control device. Wherein, the bearing production detection control system may include:

the production detection instruction extraction module is used for extracting a bearing production detection instruction, wherein the bearing production detection instruction comprises bearing production equipment operation data corresponding to target bearing production equipment;

the data feature processing module is used for loading the bearing production equipment operation data into an optimized equipment operation data analysis neural network to perform operation data feature mining processing so as to output corresponding operation data mining feature representation, then performing operation data feature restoration processing on the operation data mining feature representation so as to output corresponding key equipment operation behaviors, key behavior semantic data and key equipment operation behavior representation data, the optimized equipment operation data analysis neural network is formed by loading typical bearing production equipment operation data into the equipment operation data analysis neural network to perform network optimization, the equipment operation data analysis neural network to be optimized is used for performing operation data feature mining processing and operation data feature restoration processing so as to output typical key equipment operation behaviors, typical key behavior semantic data and typical key equipment operation behavior representation data, then analyzing and outputting corresponding equipment operation behavior abnormal analysis results according to the typical key equipment operation behaviors, the typical key behavior semantic data and the typical key equipment operation behavior representation data, and performing operation behavior characteristic analysis processing on the output typical operation behavior representation data so as to output corresponding typical key equipment operation behavior analysis results, and determining actual operation value analysis results according to the typical operation behavior analysis results and learning value analysis results, outputting a corresponding behavior critical analysis learning cost value, and performing network optimization on the equipment operation data analysis neural network to be optimized according to the abnormal analysis learning cost value and the behavior critical analysis learning cost value to form a corresponding optimization equipment operation data analysis neural network;

The quality representative data determining module is used for analyzing and outputting a corresponding first equipment operation behavior abnormal analysis result according to the key equipment operation behavior, the key behavior semantic data and the key equipment operation behavior representative data, and determining quality representative data of the bearing product correspondingly produced by the target bearing production equipment according to the first equipment operation behavior abnormal analysis result.

It should be appreciated that the production test instruction extraction module, the data feature processing module, and the quality representative data determination module may all be software functional modules. Correspondingly, the bearing production detection control system may be a software functional system.

In summary, the method and the device for detecting and controlling the bearing production provided by the invention can extract the operation data of the bearing production equipment; loading the operation data of the bearing production equipment into an operation data analysis neural network of the optimizing equipment to perform operation data feature mining processing so as to output operation data mining feature representation, and performing operation data feature restoration processing on the operation data mining feature representation so as to output key equipment operation behaviors, key behavior semantic data and key equipment operation behavior representative data; and analyzing and outputting a first equipment operation behavior abnormality analysis result according to the key equipment operation behavior, the key behavior semantic data and the key equipment operation behavior representation data, and determining quality representation data of the bearing product correspondingly produced according to the first equipment operation behavior abnormality analysis result. Based on the foregoing, because the analysis processing is performed based on the operation data of the bearing production equipment, compared with the conventional technical scheme of performing video monitoring on the bearing or equipment, no additional monitoring equipment is required to be deployed, so that the detection convenience is higher.

The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. The bearing production detection control method is characterized by comprising the following steps of:

extracting a bearing production detection instruction, wherein the bearing production detection instruction comprises bearing production equipment operation data corresponding to target bearing production equipment; performing staged decomposition processing on the bearing production equipment operation data to form a plurality of corresponding original equipment operation behavior log data, wherein each original equipment operation behavior log data corresponds to one production stage, production procedure or production link of the bearing;

loading the bearing production equipment operation data into an optimizing equipment operation data analysis neural network to perform operation data feature mining processing so as to output corresponding operation data mining feature representation, performing operation data feature restoration processing on the operation data mining feature representation so as to output corresponding key equipment operation behaviors, key behavior semantic data and key equipment operation behavior representative data, loading the typical bearing production equipment operation data into the equipment operation data analysis neural network to perform network optimization forming, performing operation data feature mining processing and operation data feature restoration processing through the equipment operation data analysis neural network to perform operation data feature mining processing and operation data feature restoration processing so as to output typical key equipment operation behaviors, typical key behavior semantic data and typical key equipment operation behavior representative data, analyzing and outputting corresponding equipment operation behavior abnormality analysis results according to the typical key equipment operation behaviors, the typical key operation behavior semantic data and the typical key equipment operation behavior representative data, performing operation behavior feature mining feature representation so as to output corresponding typical key equipment operation behavior analysis results, determining actual operation behavior analysis values according to the equipment operation behavior configuration analysis results and learning values, determining the actual operation abnormality analysis values according to the equipment operation behavior configuration analysis results, and performing network optimization on the equipment operation data analysis neural network to be optimized according to the anomaly analysis learning cost value and the behavior criticality analysis learning cost value to form a corresponding optimized equipment operation data analysis neural network, wherein the typical operation data mining feature representation comprises each typical operation behavior mining feature representation, and the operation behavior criticality analysis processing comprises: traversing successively in each representative operational behavior mining feature representation;

Traversing the first representative operational behavioral mining feature representation, traversing the second representative operational behavioral mining feature representation, traversing the third representative operational behavioral mining feature representation, traversing the fourth representative operational behavioral mining feature representation;

forming a traversed typical operation behavior mining feature representation, and carrying out mapping output processing on the traversed typical operation behavior mining feature representation to output a corresponding mapping operation behavior mining feature representation;

weighting the typical operation behavior mining feature table based on a first parameter matrix, and then summing the weighted results based on a second parameter matrix to realize shifting so as to obtain a mapping operation behavior mining feature representation; the first parameter matrix and the second parameter matrix belong to optimization parameters of a neural network;

performing operation behavior criticality analysis processing on the mapping operation behavior mining feature representation to output typical key equipment operation behavior analysis probability corresponding to the traversed typical operation behavior mining feature representation;

the analysis probability of the typical key equipment operation behaviors refers to the probability that the corresponding typical equipment operation behaviors belong to the key equipment operation behaviors, namely the estimated probability value, by the typical operation behavior mining characteristics; processing the mapping operation behavior mining feature representation based on a softmax function, and outputting corresponding typical key equipment operation behavior analysis probability;

Analyzing the traversed typical operation behavior mining features to represent corresponding traversed typical operation behavior analysis results according to the typical key equipment operation behavior analysis probability, and forming corresponding typical key equipment operation behavior analysis results according to each typical operation behavior mining feature to represent corresponding typical operation behavior analysis results;

2. The bearing production detection control method according to claim 1, wherein the optimizing equipment operation data analysis neural network comprises an optimizing data feature mining model and an optimizing data feature restoring model; the step of loading the operation data of the bearing production equipment into an operation data analysis neural network of the optimizing equipment to perform operation data feature mining processing so as to output corresponding operation data mining feature representations, and then performing operation data feature restoration processing on the operation data mining feature representations so as to output corresponding operation behavior of the key equipment, semantic data of the key behavior and operation behavior representative data of the key equipment, comprises the following steps:

3. The bearing production inspection control method according to claim 1, further comprising:

and performing network optimization on the equipment operation data analysis neural network to be optimized according to the abnormal analysis learning cost value and the behavior critical analysis learning cost value to form a candidate equipment operation data analysis neural network, and re-marking the candidate equipment operation data analysis neural network as the equipment operation data analysis neural network to be optimized, so that the steps of extracting typical bearing production equipment operation data, equipment operation behavior actual results and key equipment operation behavior actual results are continuously executed, and the optimized equipment operation data analysis neural network is formed under the condition that a network optimization target is reached.

4. The method for detecting and controlling bearing production according to claim 3, wherein the step of loading the typical bearing production equipment operation data into an equipment operation data analysis neural network to be optimized to perform operation data feature mining processing to output a corresponding typical operation data mining feature representation, and performing operation data feature restoration processing on the typical operation data mining feature representation to output a corresponding typical critical equipment operation behavior, typical critical behavior semantic data, and typical critical equipment operation behavior representative data comprises the steps of:

5. The bearing production inspection control method according to claim 4, wherein the step of analyzing a neural network using the device operation data to be optimized, performing data comparison processing on the candidate device operation behavior representation data and each of the key device operation behavior indication data, and, in the case where the candidate device operation behavior representation data coincides with the key device operation behavior indication data, performing a marking processing on the candidate device operation behavior representation data to form corresponding typical key device operation behavior representation data, comprises:

6. The bearing production detection control method as claimed in claim 3, wherein the step of performing analysis determination of the learning cost value according to the actual results of the operation behaviors of the key device and the analysis results of the operation behaviors of the typical key device to output the corresponding behavior criticality analysis learning cost value includes:

7. The bearing production detection control method according to claim 3, wherein the equipment operation data analysis neural network to be optimized comprises a data feature mining model to be optimized, a data feature restoration model to be optimized and a key analysis processing model to be optimized;

Constructing a corresponding optimizing equipment operation data analysis neural network according to the optimized data characteristic mining model and the optimized data characteristic restoring model;

the step of constructing a corresponding optimizing equipment operation data analysis neural network according to the optimized data feature mining model to be optimized and the optimized data feature restoring model to be optimized comprises the following steps:

discarding the optimized key analysis processing model in the equipment operation data analysis neural network to be optimized so as to obtain the optimized equipment operation data analysis neural network.

8. A bearing production inspection control device comprising a processor and a memory, the memory for storing a computer program, the processor for executing the computer program to implement the bearing production inspection control method of any one of claims 1-7.