A method and system for enterprise production safety evaluation based on data mining

Abstract

The present invention discloses a method and system for enterprise production safety evaluation based on data mining, which belongs to the field of production safety management, wherein the method includes: using a monitoring device to collect operation images each time an electric box is operated; performing image data mining processing to generate an operation image sequence; performing safety evaluation of the posture of the electric box and the tool used by the electric box, constructing a posture safety evaluation result sequence and a tool safety evaluation result sequence, generating an optimized posture safety evaluation result sequence and an optimized tool safety evaluation result sequence; performing safety evaluation of the use of the electric box in production, and obtaining a production safety evaluation result. This application solves the technical problem that the prior art cannot implement dynamic safety evaluation of the entire production process, resulting in low accuracy and reliability of the safety evaluation, and achieves the technical effect of realizing the intelligentization and dynamicization of enterprise production safety evaluation through data-driven methods, and improving the accuracy and reliability of safety evaluation.

Description

Technical Field

[0001] The present invention relates to the field of production safety management, and in particular to a method and system for evaluating enterprise production safety based on data mining. Background Art

[0002] With the continuous improvement of the degree of industrial automation, the number of electrical control devices used in the production process of enterprises is increasing day by day. As an important electrical interface device, the use and operation safety of the electrical control box (referred to as the electrical box) directly affects the safe and stable operation of the production process. At present, the safety evaluation and risk identification of the electrical box mainly rely on the experience judgment and regular spot checks of on-site personnel. This evaluation method is highly subjective, and the evaluation results are difficult to objectively reflect the actual safety hazards in the use of the electrical box, resulting in poor accuracy of the enterprise production safety evaluation results. At the same time, the existing enterprise production safety evaluation only uses a single operation for evaluation, and cannot analyze the long-term operation of the electrical box, resulting in poor reliability of the enterprise production safety evaluation results. Summary of the invention

[0003] This application provides a method and system for enterprise production safety evaluation based on data mining, aiming to solve the technical problem that the existing technology cannot implement dynamic safety evaluation of the entire production process, resulting in low accuracy and reliability of safety evaluation.

[0004] In view of the above problems, the present application provides a method and system for enterprise production safety evaluation based on data mining.

[0005] The first aspect disclosed in the present application provides a method for enterprise production safety evaluation based on data mining, the method comprising: by deploying a monitoring device at a designated electrical box in a target enterprise, collecting operation images each time the electrical box is operated; performing image data mining processing on the operation images to generate an operation image sequence, wherein the image data mining processing includes similarity mining; performing safety evaluation on the use posture of the electrical box and safety evaluation on the tools used by the electrical box according to the operation image sequence, and obtaining posture safety evaluation results and tool safety evaluation results; performing data mining and evaluation on the operation images of multiple operations of the electrical box within a preset time period to obtain multiple posture safety evaluation results and multiple tool safety evaluation results, and constructing a posture safety evaluation result sequence and a tool safety evaluation result sequence; performing evaluation data mining processing on the posture safety evaluation result sequence and the tool safety evaluation result sequence to generate an optimized posture safety evaluation result sequence and an optimized tool safety evaluation result sequence, wherein the evaluation data mining processing includes deviation mining; performing safety evaluation on the use of the electrical box in production based on the optimized posture safety evaluation result sequence and the optimized tool safety evaluation result sequence to obtain a production safety evaluation result.

[0006] Another aspect disclosed in the present application provides an enterprise production safety assessment system based on data mining, the system comprising: an operation image acquisition module, which is used to collect operation images each time the electric box is operated by a monitoring device deployed on a designated electric box in a target enterprise; an image data mining module, which is used to perform image data mining processing on the operation images to generate an operation image sequence, wherein the image data mining processing includes similarity mining; a use safety assessment module, which is used to perform a safety assessment on the use posture of the electric box and a safety assessment on the use tools of the electric box according to the operation image sequence, and obtain a posture safety assessment result and a tool safety assessment result; an assessment result sequence module, which is used to perform a safety assessment on the use posture of the electric box and a tool safety assessment result based on a preset time cycle. The operation images of the electrical box are operated multiple times during the period and the data mining and evaluation are performed to obtain multiple posture safety evaluation results and multiple tool safety evaluation results, and a posture safety evaluation result sequence and a tool safety evaluation result sequence are constructed; an evaluation data mining module is used to perform evaluation data mining processing on the posture safety evaluation result sequence and the tool safety evaluation result sequence to generate an optimized posture safety evaluation result sequence and an optimized tool safety evaluation result sequence, wherein the evaluation data mining processing includes deviation mining; an evaluation result acquisition module is used to perform safety evaluation on the use of the electrical box in production based on the optimized posture safety evaluation result sequence and the optimized tool safety evaluation result sequence to obtain a production safety evaluation result.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] The monitoring device is installed at the designated electrical box in the target enterprise to monitor and collect the operation images of the electrical box in real time every time the electrical box is operated, so as to obtain the first-hand image information of the operation of the electrical box; the image data mining and processing are performed on the large number of collected operation images to generate an image sequence describing the operation process of the electrical box, and the main operation images of the electrical box are extracted from a large number of original monitoring videos; according to the extracted operation image sequence, the posture recognition and tool detection are used to judge the operator's posture and the tools used, and the posture safety evaluation results and tool safety evaluation results are given to determine the safety risk of a single operation behavior; within a certain time span, the safety evaluation is performed on multiple operation processes of the electrical box to form a posture safety evaluation result sequence and a tool safety evaluation result sequence, which reflect the changing trend of the safety status of the electrical box operation; the posture Evaluation data mining is carried out on the safety evaluation result sequence and the tool safety evaluation result sequence to find out the abnormal points in the evaluation and the deviation correlation between the evaluation indicators, generate the optimized posture safety evaluation result sequence and the optimized tool safety evaluation result sequence, and improve the accuracy of the safety evaluation result sequence; comprehensively consider the optimized posture safety evaluation result sequence and the optimized tool safety evaluation result sequence, make an overall judgment on the safety of the use of electrical boxes in the enterprise production, obtain the production safety evaluation results, and implement a technical solution for safety evaluation based on data mining, which solves the technical problem that the existing technology cannot implement dynamic safety evaluation of the entire production process, resulting in low accuracy and reliability of safety evaluation, and achieves the technical effect of realizing the intelligent and dynamic enterprise production safety evaluation through data-driven means and improving the accuracy and reliability of safety evaluation.

[0009] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A flow chart of a method for evaluating enterprise production safety based on data mining is provided for an embodiment of the present application;

[0011] Figure 2 A flowchart of generating an operation image sequence in a method for evaluating enterprise production safety based on data mining is provided for an embodiment of the present application;

[0012] Figure 3 A structural schematic diagram of an enterprise production safety evaluation system based on data mining is provided for an embodiment of the present application.

[0013] Explanation of reference numerals: operation image acquisition module 11 , image data mining module 12 , use safety evaluation module 13 , evaluation result sequence module 14 , evaluation data mining module 15 , evaluation result acquisition module 16 . DETAILED DESCRIPTION

[0014] The overall idea of ​​the technical solution provided by this application is as follows:

[0015] The embodiment of the present application provides a method and system for enterprise production safety evaluation based on data mining. First, the operation images of the enterprise's internal electrical control device, namely the electric box, are collected by deploying a monitoring device in a designated electric box in the target enterprise. Then, key information is extracted from the operation image, and an operation image sequence is constructed in chronological order. Next, the posture safety evaluation of the electric box and the safety evaluation of the tools used by the electric box are evaluated, and a time series of evaluation results is formed for multiple operation images of the same electric box according to a certain time span, namely, the posture safety evaluation result sequence and the tool safety evaluation result sequence. Subsequently, significant anomalies and deviations are found in the posture safety evaluation result sequence and the tool safety evaluation result sequence, and an optimized posture safety evaluation result sequence and an optimized tool safety evaluation result sequence are generated. After that, finally, the posture safety evaluation result sequence and the tool safety evaluation result sequence are comprehensively optimized, and the production safety evaluation results of the electric box operation are output, which realizes the use of data-driven methods to monitor and evaluate the operational safety of enterprise production, and comprehensively improve the accuracy and reliability of enterprise production safety evaluation.

[0016] After introducing the basic principles of the present application, various non-limiting implementation methods of the present application will be specifically described below in conjunction with the drawings in the specification.

[0017] Embodiment 1

[0018] like Figure 1 As shown, the embodiment of the present application provides a method for evaluating enterprise production safety based on data mining, the method comprising:

[0019] By deploying monitoring devices at designated power boxes in target enterprises, operation images are collected every time the power boxes are operated;

[0020] In the embodiment of the present application, in order to collect data on the operation process of a designated electrical box in a target enterprise, a monitoring device is deployed on the designated electrical box in the target enterprise to realize image monitoring of the entire operation process of the electrical box. The monitoring device can be an image acquisition device such as an industrial camera or a mobile visual device. Preferably, the resolution of the monitoring device is above 10 million pixels, the frame rate is greater than 25 frames per second, and the electrical box operation image is collected in real time.

[0021] At the same time, an acquisition control module is deployed for each monitoring device. This module monitors all monitoring devices in real time. When operator activity is detected, the acquisition control module automatically controls the corresponding monitoring device to start acquisition, and obtains the operation image of each operation of the electrical box, providing a first-hand data source for image data mining and safety evaluation.

[0022] Performing image data mining processing on the operation image to generate an operation image sequence, wherein the image data mining processing includes similarity mining;

[0023] Further, such as Figure 2 As shown, this step includes:

[0024] Dividing the operation image into frames to obtain an operation image set;

[0025] Combining adjacent operation images in the operation image set in pairs to obtain an operation image group set;

[0026] Performing image data mining on a plurality of operation image groups in the operation image group set to obtain an image mining result, wherein the image mining result includes a plurality of mined operation image groups;

[0027] The operation image sequence is generated based on a plurality of drilling operation images within the plurality of drilling operation image groups.

[0028] Furthermore, the image mining results are obtained, including:

[0029] Obtaining a set of sample operation image groups, and performing similarity analysis on two sample operation images in each sample operation image group to obtain a sample similarity set;

[0030] Based on the twin network, the sample operation image group set and the sample similarity set are used to construct an operation image mining analyzer;

[0031] Based on the operation image mining analyzer, multiple operation image groups in the operation image group set are analyzed to obtain multiple similarities;

[0032] The operation image groups corresponding to the smallest K similarities among the multiple similarities are output as the multiple mining operation image groups, and the output is the image mining result, where K is an integer greater than 1.

[0033] In a feasible implementation, after obtaining the operation image, multiple frames of operation images in the operation image are processed by image data mining, and representative frames are extracted to form an operation image sequence, wherein the image data mining adopts a similarity mining method.

[0034] First, the operation image is parsed into an image data stream containing multiple frames to obtain the original video frame data, wherein each video frame is an operation image, and each operation image has a corresponding operation time, thereby obtaining an operation image set. Then, according to the time sequence of the operation images in the operation image set, each operation image is paired and combined with its adjacent previous and next operation images to obtain an operation image group set. For example, if the operation image set has images A, B, C, and D in order of operation time, the paired combinations are: the combination of A and B, the combination of B and C, and the combination of C and D, which is an operation image group set, wherein each combination is an operation image group.

[0035] In order to realize image data mining of multiple operation image groups in the operation image group set, an operation image mining analyzer is pre-established. First, image data containing representative operation processes are selected from historical operation images, and frame division processing is performed to obtain a historical operation image set. Then, adjacent operation images in the historical operation image set are combined in pairs to obtain a sample operation image group set. Secondly, for each operation image group in the sample operation image group set, the visual features of the two operation images in the group are analyzed, and their similarities are annotated to obtain a sample similarity set. Among them, the sample similarity of the two sample operation images in the sample operation image group with strong matching obtains a high score, and the sample similarity with large differences obtains a low score. Then, the twin network model structure is defined for learning image feature representation and realizing similarity regression prediction tasks. The twin network model is supervised for training using the sample operation image group set and the corresponding sample similarity set, and the loss function is designed as the regression loss between the predicted similarity and the true similarity. During the training process, the gradient descent method is used to gradually update the network parameters. After multiple rounds of iterations, the model gradually converges, so that the operation image mining analyzer can calculate and output the similarity score for any new input operation image group, and realize the similarity judgment of two operation images in the operation image group.

[0036] Subsequently, the multiple operation image groups in the acquired operation image group set are sequentially input into the operation image mining analyzer, and the operation image mining analyzer outputs the similarity of the two operation images in each operation image group. Repeat the process to form a similarity result set of each operation image group in the operation image group set, which is defined as multiple similarities. After that, the multiple similarities are sorted in order from small to large, the first K similarities after sorting are selected, and the K operation image groups corresponding to these K similarities are queried. In these operation image groups with lower similarities, the two operation images contained have large changes, which are key node information in the operation process of the electric box. Among them, the K value is an empirical value to ensure the balance between the integrity of the results and the computational efficiency, and is an integer greater than 1. Then, the obtained K operation image groups are used as multiple mining operation image groups to constitute the image mining results of this time. Thereafter, all the mining operation images contained in the multiple mining operation image groups in the image mining results are taken out, and they are reordered and combined in order according to the acquisition timestamp information of each mining operation image to generate an operation image sequence.

[0037] By performing similarity analysis on multiple frames of operation images within the operation image, adjacent operation images with small similarity are retained to provide a data basis for accurate safety evaluation analysis. At the same time, the analysis of images with high similarity is reduced, computing power is saved, and safety evaluation efficiency is improved.

[0038] According to the operation image sequence, a posture safety evaluation of the use of the electric box and a tool safety evaluation of the use of the electric box are performed to obtain a posture safety evaluation result and a tool safety evaluation result;

[0039] Furthermore, this step includes:

[0040] Extracting operation images from the operation image sequence according to a first extraction span, and performing downsampling processing on the extracted operation images to obtain a first operation image sequence;

[0041] Extracting operation images from the operation image sequence according to a second extraction span to obtain a second operation image sequence, wherein the second extraction span is greater than the first extraction span;

[0042] According to the first operation image sequence and the second operation image sequence, based on the slowfast network, a posture safety evaluation of the use of the power box and a tool safety evaluation of the use of the power box are performed to obtain the posture safety evaluation result and the tool safety evaluation result.

[0043] Further, obtaining the posture safety evaluation result and the tool safety evaluation result includes:

[0044] According to the usage monitoring data of the designated electric box, a sample first operation image sequence set and a sample second operation image sequence set are obtained, and a sample posture safety evaluation result set and a sample tool safety evaluation result set are obtained;

[0045] Based on the slowfast network, a posture safety evaluation branch and a tool safety evaluation branch are constructed, wherein both the posture safety evaluation branch and the tool safety evaluation branch include a slow path and a fast path;

[0046] Using the sample first operation image sequence set and the sample second operation image sequence set, respectively combined with the sample posture safety evaluation result set and the sample tool safety evaluation result set, supervised training is performed on the posture safety evaluation branch and the tool safety evaluation branch until convergence, to obtain an image safety evaluator;

[0047] Based on the image safety evaluator, the first operation image sequence and the second operation image sequence are respectively subjected to safety evaluation of posture using the electric box and safety evaluation of tools using the electric box to obtain the posture safety evaluation result and the tool safety evaluation result.

[0048] In a preferred embodiment, after obtaining a complete operation image sequence, first, the extraction span of the slow sequence is set, for example, 1 frame is extracted every 10 frames, and the operation image sequence is sampled according to the set extraction span of the slow sequence to obtain discrete operation images, and the discrete operation images are downsampled using downsampling algorithms such as average pooling to reduce the resolution of the discrete operation images to obtain a first operation image sequence. At the same time, the extraction span of the fast sequence is set, for example, 1 frame is extracted every 2 frames, and the extraction span is significantly higher than the extraction span of the slow sequence. The operation image sequence is sampled according to the extraction span of the fast sequence to obtain a second operation image sequence. Compared with the first operation image sequence, the second operation image sequence contains more operation images and has a higher image resolution.

[0049] The image safety evaluator is pre-built. First, according to the historical monitoring operation images of the same type of electric boxes as the specified electric box, sampling and processing are performed to obtain the sample first operation image sequence set and the sample second operation image sequence set. At the same time, safety assessment rules are set, such as when operating the electric box, the switch needs to be operated with one hand sideways; the face should be kept at a certain distance from the electric box, and the electric box should not be operated facing the front; and insulating protective clothing and other tools should be worn. Subsequently, professional assessors are asked to manually annotate these sample sequences according to the safety assessment rules to form a sample posture safety evaluation result set and a sample tool safety evaluation result set, and each evaluation result has an evaluation level. Then, a posture safety evaluation branch is constructed, which contains a slow path and a fast path. The slow path inputs the first operation image sequence to extract the stable features of the posture, and the fast path inputs the second operation image sequence to extract the instantaneous change features of the posture. At the same time, a tool safety evaluation branch is constructed, and its structure and input and output are similar to those of the posture safety evaluation branch. After that, the network layer parameters of the posture safety evaluation branch and the tool safety evaluation branch are customized to ensure that they learn and evaluate different safety indicators, and a slowFast network is constructed. Next, supervised learning of the slowFast network is performed using the sample first operation image sequence set, the sample second operation image sequence set, and the sample posture safety evaluation result set and the sample tool safety evaluation result set. The operation image sequences in the sample first operation image sequence set and the sample second operation image sequence set are input into the posture safety evaluation branch and the tool safety evaluation branch respectively. The sample predicted posture safety evaluation results and the sample predicted tool safety evaluation results output by the branch are compared with the sample posture safety evaluation results and the sample tool safety evaluation results in the corresponding sample posture safety evaluation result set and the sample tool safety evaluation result set to calculate the loss. At the same time, the parameters of the slowFast network are adjusted through back propagation, and the loss is gradually reduced after multiple iterations. When the evaluation losses of the two branches converge, the SlowFast network optimized layer by layer is obtained, which is defined as the image safety evaluator.

[0050] After acquiring the image safety evaluator, when a new first operation image sequence and a second operation image sequence are generated, the first operation image sequence and the second operation image sequence are input into the image safety evaluator, and the first operation image sequence and the second operation image sequence of the image safety evaluator are simultaneously sent to the posture safety evaluation branch and the tool safety evaluation branch. Afterwards, the posture safety evaluation branch transmits the first operation image sequence to the slow path therein, transmits the second operation image sequence to the fast path therein, and outputs the posture safety evaluation result; the tool safety evaluation branch transmits the first operation image sequence to the slow path therein, transmits the second operation image sequence to the fast path therein, and outputs the tool safety evaluation result.

[0051] Based on the operation images of multiple operations of the electric box within a preset time period, multiple posture safety evaluation results and multiple tool safety evaluation results are obtained by data mining and evaluation, and a posture safety evaluation result sequence and a tool safety evaluation result sequence are constructed;

[0052] In the embodiment of the present application, after obtaining the posture safety evaluation result sequence and tool safety evaluation result of a single electric box operation, they are stored. Then, a preset time period is extracted, such as the most recent week, and the posture safety evaluation result and tool safety evaluation result of the operation image of each operation of the electric box within the preset time period are queried to obtain multiple posture safety evaluation results and multiple tool safety evaluation results. Subsequently, the multiple posture safety evaluation results and multiple tool safety evaluation results are sorted according to the time sequence of the evaluation results, and the posture safety evaluation result sequence and tool safety evaluation result sequence are obtained to reduce the impact of accidental fluctuations in a single operation, and lay the foundation for obtaining more objective and stable production safety evaluation results.

[0053] Performing evaluation data mining processing on the posture safety evaluation result sequence and the tool safety evaluation result sequence to generate an optimized posture safety evaluation result sequence and an optimized tool safety evaluation result sequence, wherein the evaluation data mining processing includes deviation degree mining;

[0054] Furthermore, this step includes:

[0055] Combining adjacent posture safety evaluation results in the posture safety evaluation result sequence in pairs to obtain a posture safety evaluation result group set;

[0056] Performing deviation analysis on multiple groups of posture safety evaluation results in the posture safety evaluation result group set to obtain a deviation set;

[0057] Select the posture safety evaluation result group corresponding to the largest J deviations in the deviation set to generate the optimized posture safety evaluation result sequence, where J is an integer greater than 1;

[0058] The tool safety evaluation result sequence is subjected to evaluation data mining processing to generate the optimization tool safety evaluation result sequence.

[0059] In a feasible implementation, after obtaining the posture safety evaluation result sequence, according to the order of the posture safety evaluation results in the posture safety evaluation result sequence, select every two adjacent posture safety evaluation results for combination. For example, results A and B form a group in pairs, B and C form a group in pairs, and so on, repeat the combination process until all adjacent posture safety evaluation results in the posture safety evaluation result sequence are organized in the form of pairs, and a posture safety evaluation result group set is obtained. Then, the two posture safety evaluation results in each posture safety evaluation result group in the posture safety evaluation result group set are analyzed, and the absolute value of the numerical difference of their evaluation levels is calculated as the deviation of the posture safety evaluation result group. The deviation of the two posture safety evaluation results in each posture safety evaluation result group in the posture safety evaluation result group set is repeatedly calculated to obtain a deviation set. Subsequently, all the deviations in the deviation set are sorted from large to small according to the deviation. The larger the deviation, the greater the difference between the two adjacent posture safety evaluation results and the greater the change in the safety situation. After that, the posture safety evaluation result group corresponding to the first J deviations after sorting is selected, and all posture safety evaluation results contained in the selected posture safety evaluation result group are taken out and reassembled in chronological order to form an optimized posture safety evaluation result sequence, where J is an integer greater than 1. At the same time, in a manner of generating an optimized posture safety evaluation result sequence from a posture safety evaluation result sequence, the tool safety evaluation result sequence is subjected to evaluation data mining processing to obtain an optimized tool safety evaluation result sequence.

[0060] Based on the optimization posture safety evaluation result sequence and the optimization tool safety evaluation result sequence, a safety evaluation of the use of the electric box in production is performed to obtain a production safety evaluation result.

[0061] Furthermore, this step includes:

[0062] Obtain a sample optimization posture safety evaluation result sequence set and a sample optimization tool safety evaluation result sequence set, and obtain a sample production safety evaluation result set;

[0063] Using the sample optimization posture safety evaluation result sequence set, the sample optimization tool safety evaluation result sequence set and the sample production safety evaluation result set, a production safety evaluator is constructed, wherein the production safety evaluator includes a first evaluation branch, a second evaluation branch and a weighted calculation branch;

[0064] The first evaluation result and the second evaluation result are obtained by inputting the optimization posture safety evaluation result sequence and the optimization tool safety evaluation result sequence into the first evaluation branch and the second evaluation branch, and then inputting them into the weighted calculation branch to calculate and obtain the production safety evaluation result.

[0065] In a feasible implementation, before realizing the automated production safety evaluation, first, historical operation images are selected for processing to obtain a sample optimization posture safety evaluation result sequence set and a sample optimization tool safety evaluation result sequence set. At the same time, professional assessors manually evaluate the production safety level of the corresponding operation process to form a sample production safety evaluation result set. Then, a network structure including a first evaluation branch, a second evaluation branch and a weighted calculation branch is defined as a production safety evaluator. Among them, the first evaluation branch is used to perform a safety evaluation on the optimized posture safety result sequence, and obtain a safety score for the posture when operating the electric box as the first evaluation result; the second evaluation branch is used to perform a safety evaluation on the optimized tool safety result sequence, and obtain a safety score for the use of the tool when operating the electric box as the second evaluation result; the weighted calculation branch performs weighted processing on the first evaluation result and the second evaluation result to obtain the final production safety evaluation result. The production safety evaluator is trained according to the obtained sample optimization posture safety evaluation result sequence set, the sample optimization tool safety evaluation result sequence set and the sample production safety evaluation result set, and the network parameters of the production safety evaluator are continuously optimized through supervised learning to achieve the mapping and approximation of the safety evaluation, thereby obtaining a production safety evaluator that meets the requirements.

[0066] After obtaining the newly generated optimization posture safety evaluation result sequence and optimization tool safety evaluation result sequence, the optimization posture safety evaluation result sequence and the optimization tool safety evaluation result sequence are input into the production safety evaluator, and the production safety evaluator inputs the optimization posture safety evaluation result sequence into the first evaluation branch and the optimization tool safety evaluation result sequence into the second evaluation branch to obtain the first evaluation result and the second evaluation result. After that, the first evaluation result and the second evaluation result are transmitted to the weighted calculation branch, and the first evaluation result and the second evaluation result are weighted to obtain an accurate and reliable production safety evaluation result.

[0067] In summary, the enterprise production safety evaluation method based on data mining provided by the embodiment of the present application has the following technical effects:

[0068] By deploying monitoring devices on designated electric boxes in target enterprises, operation images are collected every time the electric box is operated, and first-hand electric box operation image data is obtained to provide the original information source for subsequent analysis. Image data mining and processing are performed on the operation images to generate an operation image sequence to characterize the entire process of the electric box operation. According to the operation image sequence, the safety evaluation of the electric box use posture and the safety evaluation of the electric box use tools are performed, and the posture safety evaluation results and tool safety evaluation results are obtained to achieve a comprehensive safety evaluation of the operation image. Based on the operation images of multiple operation of the electric box within a preset time period, multiple posture safety evaluation results and multiple tool safety evaluation results are obtained by data mining and evaluation, and a posture safety evaluation result sequence and a tool safety evaluation result sequence are constructed. Evaluation data mining and processing are performed on the posture safety evaluation result sequence and the tool safety evaluation result sequence to generate an optimized posture safety evaluation result sequence and an optimized tool safety evaluation result sequence, so as to optimize and adjust the evaluation result sequence and improve accuracy. Based on the optimized posture safety evaluation result sequence and the optimized tool safety evaluation result sequence, the safety evaluation of the electric box use in production is performed to obtain the production safety evaluation results, and an accurate and reliable electric box operation safety evaluation is completed to provide decision support for enterprise production safety management.

[0069] Embodiment 2

[0070] Based on the same inventive concept as the enterprise production safety evaluation method based on data mining in the above-mentioned embodiment, Figure 3 As shown, the embodiment of the present application provides an enterprise production safety evaluation system based on data mining, the system comprising:

[0071] The operation image acquisition module 11 is used to collect the operation image each time the electric box is operated by using the monitoring device deployed in the designated electric box in the target enterprise;

[0072] An image data mining module 12, used for performing image data mining processing on the operation image to generate an operation image sequence, wherein the image data mining processing includes similarity mining;

[0073] A use safety evaluation module 13 is used to perform a safety evaluation on the posture of the electric box and a safety evaluation on the tool used by the electric box according to the operation image sequence, and obtain a posture safety evaluation result and a tool safety evaluation result;

[0074] An evaluation result sequence module 14 is used to construct a posture safety evaluation result sequence and a tool safety evaluation result sequence based on multiple posture safety evaluation results and multiple tool safety evaluation results obtained by performing data mining and evaluation on operation images of multiple operations of the electric box within a preset time period;

[0075] An evaluation data mining module 15 is used to perform evaluation data mining processing on the posture safety evaluation result sequence and the tool safety evaluation result sequence to generate an optimized posture safety evaluation result sequence and an optimized tool safety evaluation result sequence, wherein the evaluation data mining processing includes deviation degree mining;

[0076] The evaluation result acquisition module 16 is used to perform safety evaluation on the use of the electric box in production based on the optimization posture safety evaluation result sequence and the optimization tool safety evaluation result sequence to obtain the production safety evaluation result.

[0077] Furthermore, the image data mining module 12 includes the following execution steps:

[0078] Dividing the operation image into frames to obtain an operation image set;

[0079] Combining adjacent operation images in the operation image set in pairs to obtain an operation image group set;

[0080] Performing image data mining on a plurality of operation image groups in the operation image group set to obtain an image mining result, wherein the image mining result includes a plurality of mined operation image groups;

[0081] The operation image sequence is generated based on a plurality of drilling operation images within the plurality of drilling operation image groups.

[0082] Furthermore, the image data mining module 12 also includes the following execution steps:

[0083] Obtaining a set of sample operation image groups, and performing similarity analysis on two sample operation images in each sample operation image group to obtain a sample similarity set;

[0084] Based on the twin network, the sample operation image group set and the sample similarity set are used to construct an operation image mining analyzer;

[0085] Based on the operation image mining analyzer, multiple operation image groups in the operation image group set are analyzed to obtain multiple similarities;

[0086] The operation image groups corresponding to the smallest K similarities among the multiple similarities are output as the multiple mining operation image groups, and the output is the image mining result, where K is an integer greater than 1.

[0087] Further, using the safety evaluation module 13 includes the following execution steps:

[0088] Extracting operation images from the operation image sequence according to a first extraction span, and performing downsampling processing on the extracted operation images to obtain a first operation image sequence;

[0089] Extracting operation images from the operation image sequence according to a second extraction span to obtain a second operation image sequence, wherein the second extraction span is greater than the first extraction span;

[0090] According to the first operation image sequence and the second operation image sequence, based on the slowfast network, a posture safety evaluation of the use of the power box and a tool safety evaluation of the use of the power box are performed to obtain the posture safety evaluation result and the tool safety evaluation result.

[0091] Furthermore, the use of the safety evaluation module 13 also includes the following execution steps:

[0092] According to the usage monitoring data of the designated electric box, a sample first operation image sequence set and a sample second operation image sequence set are obtained, and a sample posture safety evaluation result set and a sample tool safety evaluation result set are obtained;

[0093] Based on the slowfast network, a posture safety evaluation branch and a tool safety evaluation branch are constructed, wherein both the posture safety evaluation branch and the tool safety evaluation branch include a slow path and a fast path;

[0094] Using the sample first operation image sequence set and the sample second operation image sequence set, respectively combined with the sample posture safety evaluation result set and the sample tool safety evaluation result set, supervised training is performed on the posture safety evaluation branch and the tool safety evaluation branch until convergence, to obtain an image safety evaluator;

[0095] Based on the image safety evaluator, the first operation image sequence and the second operation image sequence are respectively subjected to safety evaluation of posture using the electric box and safety evaluation of tools using the electric box to obtain the posture safety evaluation result and the tool safety evaluation result.

[0096] Furthermore, the evaluation data mining module 15 includes the following execution steps:

[0097] Combining adjacent posture safety evaluation results in the posture safety evaluation result sequence in pairs to obtain a posture safety evaluation result group set;

[0098] Performing deviation analysis on multiple groups of posture safety evaluation results in the posture safety evaluation result group set to obtain a deviation set;

[0099] Select the posture safety evaluation result group corresponding to the largest J deviations in the deviation set to generate the optimized posture safety evaluation result sequence, where J is an integer greater than 1;

[0100] The tool safety evaluation result sequence is subjected to evaluation data mining processing to generate the optimization tool safety evaluation result sequence.

[0101] Furthermore, the evaluation result acquisition module 16 includes the following execution steps:

[0102] Obtain a sample optimization posture safety evaluation result sequence set and a sample optimization tool safety evaluation result sequence set, and obtain a sample production safety evaluation result set;

[0103] Using the sample optimization posture safety evaluation result sequence set, the sample optimization tool safety evaluation result sequence set and the sample production safety evaluation result set, a production safety evaluator is constructed, wherein the production safety evaluator includes a first evaluation branch, a second evaluation branch and a weighted calculation branch;

[0104] The first evaluation result and the second evaluation result are obtained by inputting the optimization posture safety evaluation result sequence and the optimization tool safety evaluation result sequence into the first evaluation branch and the second evaluation branch, and then inputting them into the weighted calculation branch to calculate and obtain the production safety evaluation result.

[0105] Any step of the method described above can be stored as a computer instruction or program in an unlimited computer memory, and can be called and recognized by an unlimited computer processor to implement any method in the embodiments of the present application, without any unnecessary restrictions.

[0106] Furthermore, the first or second mentioned above may not only represent an order relationship, but may also represent a specific concept, and / or refer to multiple elements that can be selected individually or in full. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A method for evaluating enterprise production safety based on data mining, characterized in that: The method comprises: By deploying monitoring devices at designated power boxes in target enterprises, operation images are collected every time the power boxes are operated; Performing image data mining processing on the operation image to generate an operation image sequence, wherein the image data mining processing includes similarity mining; According to the operation image sequence, a posture safety evaluation of the use of the electric box and a tool safety evaluation of the use of the electric box are performed to obtain a posture safety evaluation result and a tool safety evaluation result; Based on the operation images of multiple operations of the electric box within a preset time period, multiple posture safety evaluation results and multiple tool safety evaluation results are obtained by data mining and evaluation, and a posture safety evaluation result sequence and a tool safety evaluation result sequence are constructed; Performing evaluation data mining processing on the posture safety evaluation result sequence and the tool safety evaluation result sequence to generate an optimized posture safety evaluation result sequence and an optimized tool safety evaluation result sequence, wherein the evaluation data mining processing includes deviation degree mining; Based on the optimization posture safety evaluation result sequence and the optimization tool safety evaluation result sequence, a safety evaluation of the use of the electric box in production is performed to obtain a production safety evaluation result; Wherein, according to the operation image sequence, the safety evaluation of the posture of using the power box and the safety evaluation of the tool of using the power box are performed, including: Extracting operation images from the operation image sequence according to a first extraction span, and performing downsampling processing on the extracted operation images to obtain a first operation image sequence; Extracting operation images from the operation image sequence according to a second extraction span to obtain a second operation image sequence, wherein the second extraction span is greater than the first extraction span; According to the first operation image sequence and the second operation image sequence, based on the slowfast network, a posture safety evaluation of the use of the electric box and a tool safety evaluation of the use of the electric box are performed to obtain the posture safety evaluation result and the tool safety evaluation result; According to the first operation image sequence and the second operation image sequence, based on the slowfast network, a safety evaluation of the posture of using the power box and a safety evaluation of the tool of using the power box are performed, including: According to the usage monitoring data of the designated electric box, a sample first operation image sequence set and a sample second operation image sequence set are obtained, and a sample posture safety evaluation result set and a sample tool safety evaluation result set are obtained; Based on the slowfast network, a posture safety evaluation branch and a tool safety evaluation branch are constructed, wherein both the posture safety evaluation branch and the tool safety evaluation branch include a slow path and a fast path; Using the sample first operation image sequence set and the sample second operation image sequence set, respectively combined with the sample posture safety evaluation result set and the sample tool safety evaluation result set, supervised training is performed on the posture safety evaluation branch and the tool safety evaluation branch until convergence, to obtain an image safety evaluator; Based on the image safety evaluator, the first operation image sequence and the second operation image sequence are respectively subjected to safety evaluation of posture using the electric box and safety evaluation of tools using the electric box to obtain the posture safety evaluation result and the tool safety evaluation result.

2. The method according to claim 1, characterized in that Performing image data mining processing on the operation image to generate an operation image sequence includes: Dividing the operation image into frames to obtain an operation image set; Combining adjacent operation images in the operation image set in pairs to obtain an operation image group set; Performing image data mining on a plurality of operation image groups in the operation image group set to obtain an image mining result, wherein the image mining result includes a plurality of mined operation image groups; The operation image sequence is generated based on a plurality of drilling operation images within the plurality of drilling operation image groups.

3. The method according to claim 2, characterized in that Performing image data mining on a plurality of operation image groups in the operation image group set to obtain image mining results includes: Obtaining a set of sample operation image groups, and performing similarity analysis on two sample operation images in each sample operation image group to obtain a sample similarity set; Based on the twin network, the sample operation image group set and the sample similarity set are used to construct an operation image mining analyzer; Based on the operation image mining analyzer, multiple operation image groups in the operation image group set are analyzed to obtain multiple similarities; The operation image groups corresponding to the smallest K similarities among the multiple similarities are output as the multiple mining operation image groups, and the output is the image mining result, where K is an integer greater than 1.

4. The method according to claim 1, characterized in that: Performing evaluation data mining processing on the posture safety evaluation result sequence and the tool safety evaluation result sequence to generate an optimized posture safety evaluation result sequence and an optimized tool safety evaluation result sequence, including: Combining adjacent posture safety evaluation results in the posture safety evaluation result sequence in pairs to obtain a posture safety evaluation result group set; Performing deviation analysis on multiple groups of posture safety evaluation results in the posture safety evaluation result group set to obtain a deviation set; Select the posture safety evaluation result group corresponding to the largest J deviations in the deviation set to generate the optimized posture safety evaluation result sequence, where J is an integer greater than 1; The tool safety evaluation result sequence is subjected to evaluation data mining processing to generate the optimization tool safety evaluation result sequence.

5. The method according to claim 1, characterized in that Based on the optimization posture safety evaluation result sequence and the optimization tool safety evaluation result sequence, a safety evaluation of the electric box in production is performed, including: Obtain a sample optimization posture safety evaluation result sequence set and a sample optimization tool safety evaluation result sequence set, and obtain a sample production safety evaluation result set; Using the sample optimization posture safety evaluation result sequence set, the sample optimization tool safety evaluation result sequence set and the sample production safety evaluation result set, a production safety evaluator is constructed, wherein the production safety evaluator includes a first evaluation branch, a second evaluation branch and a weighted calculation branch; The first evaluation result and the second evaluation result are obtained by inputting the optimization posture safety evaluation result sequence and the optimization tool safety evaluation result sequence into the first evaluation branch and the second evaluation branch, and then inputting them into the weighted calculation branch to calculate and obtain the production safety evaluation result.

6. An enterprise production safety evaluation system based on data mining, characterized in that: The system is used to implement the enterprise production safety evaluation method based on data mining as described in any one of claims 1 to 5, and comprises: An operation image acquisition module, which is used to collect operation images each time the power box is operated through a monitoring device deployed at a designated power box in the target enterprise; An image data mining module, the image data mining module is used to perform image data mining processing on the operation image to generate an operation image sequence, wherein the image data mining processing includes similarity mining; A use safety evaluation module, wherein the use safety evaluation module is used to perform a safety evaluation on the use posture of the electric box and a safety evaluation on the use tool of the electric box according to the operation image sequence, and obtain a posture safety evaluation result and a tool safety evaluation result; An evaluation result sequence module, wherein the evaluation result sequence module is used to perform data mining and evaluation on multiple posture safety evaluation results and multiple tool safety evaluation results obtained based on operation images of multiple operations of the electric box within a preset time period, and to construct a posture safety evaluation result sequence and a tool safety evaluation result sequence; An evaluation data mining module, the evaluation data mining module is used to perform evaluation data mining processing on the posture safety evaluation result sequence and the tool safety evaluation result sequence to generate an optimized posture safety evaluation result sequence and an optimized tool safety evaluation result sequence, wherein the evaluation data mining processing includes deviation mining; An evaluation result acquisition module is used to perform a safety evaluation on the use of electrical boxes in production based on the optimization posture safety evaluation result sequence and the optimization tool safety evaluation result sequence to obtain a production safety evaluation result.

Images