A power high-low voltage cabinet production and manufacturing process monitoring management system

Abstract

The application discloses a kind of power high-low voltage cabinet production manufacturing procedure monitoring management system, it is related to power equipment manufacturing monitoring technical field, including data acquisition, correlation analysis, strategy generation, real-time regulation and optimization report module.Data acquisition module acquires multi-process site real-time manufacturing data;Correlation analysis module uses improved process correlation degree algorithm, excavates internal contact of data and identifies quality correlation and efficiency constraint relationship between processes;Strategy generation module generates dynamic monitoring strategy containing key process quality monitoring frequency, material supply priority adjustment instruction accordingly;Real-time regulation module controls production using the strategy and collects response data;Optimization report module iteratively optimizes strategy and generates report.The system can clearly identify inter-process correlation constraints, achieve dynamic adaptation of monitoring and control, form a closed-loop management, make production connection smoother, and ensure the stability and continuity of production process.

Description

Technical Field

[0001] This invention belongs to the field of power equipment manufacturing monitoring technology, specifically a monitoring and management system for the production process of high and low voltage switchgear. Background Technology

[0002] In the conventional production and manufacturing process of high and low voltage switchgear, process monitoring often adopts a single process independent data acquisition mode. Production data is organized through simple data statistics or manual recording. The monitoring strategy is mostly a pre-set fixed scheme, which can only monitor the basic operating parameters of a single process and cannot achieve collaborative analysis and dynamic control between multiple process stations.

[0003] Existing technologies lack specialized correlation analysis algorithms, making it difficult to effectively uncover the inherent connections between real-time manufacturing data from multiple process stations. This hinders the identification of quality correlations and efficiency constraints between processes, resulting in insufficient targeting of monitoring strategies. The quality monitoring frequency for key processes remains fixed, and material supply priorities are not dynamically adjusted based on inter-process constraints. Even after adjustments, monitoring strategies cannot be optimized based on production response data. This easily leads to problems such as poor process integration, failure to promptly detect potential quality issues, and a disconnect between material supply and production needs, affecting the stability and continuity of the production process.

[0004] It is necessary to analyze real-time data from multiple processes using specific algorithms, clarify the relationships and constraints between processes, and realize the dynamic generation and iterative optimization of monitoring strategies so that monitoring and control can accurately adapt to actual production needs. Summary of the Invention

[0005] This invention aims to solve at least one of the technical problems existing in the prior art; Therefore, this invention proposes a monitoring and management system for the manufacturing process of high and low voltage switchgear, comprising: The data acquisition module is used to collect real-time manufacturing data from multiple process stations on the power high and low voltage switchgear production line; The correlation analysis module is used to analyze the collected real-time manufacturing data based on the improved process correlation algorithm to identify the quality correlation and efficiency constraint relationships between process stations. The strategy generation module is used to generate dynamic monitoring strategies based on the identified correlations and constraints. The dynamic monitoring strategies include instructions for adjusting the quality monitoring frequency of key processes and instructions for adjusting the material supply priority. The real-time control module is used to control the production process of high and low voltage switchgear in real time using the dynamic monitoring strategy, and to collect the production response data after control. The optimization report module is used to iteratively optimize dynamic monitoring strategies based on production response data and preset production target models, and generate process monitoring management reports.

[0006] Furthermore, the collection of real-time manufacturing data from multiple process stations on the high and low voltage switchgear production line specifically includes: Data acquisition terminals are deployed at the sheet metal processing, busbar assembly, component installation, wiring and debugging, withstand voltage testing and shell spraying stations of the power high and low voltage switch production line; The data acquisition terminal can be used to obtain in real time the equipment operating parameters, material consumption rate, work-in-process images, operator identification, and process completion timestamps of each process station. The equipment operating parameters, material consumption rate, work-in-process images, operator identification, and process completion timestamps are aligned and packaged according to a unified time base to generate a process raw data package with time sequence tags. The original data packets of the aforementioned process are parsed to extract the structured and unstructured data. The extracted unstructured data, including images of work-in-process, is subjected to feature extraction to identify material type, assembly location compliance, and label clarity features in the images, and then converted into structured feature data. The structured data and the transformed structured feature data are merged and verified to form standardized real-time manufacturing data.

[0007] Furthermore, the improved process correlation algorithm analyzes the collected real-time manufacturing data to identify quality correlations and efficiency constraints between process stations, specifically including: Obtain standardized real-time manufacturing data, and construct an initial production network model based on process stations, with processes as nodes and logistics and information flows as virtual edges; In the initial production network model, the spatiotemporal correlation coefficient and quality fluctuation transmission coefficient of manufacturing data between any two process nodes are calculated; The improved process correlation algorithm calculates the impact weight of the quality anomaly of the preceding process on the quality pass rate of the subsequent process, and the impact weight of the delay in the completion time of the preceding process on the start-up waiting time of the subsequent process, based on the spatiotemporal correlation coefficient and the quality fluctuation transmission coefficient. The calculated impact weights and delay impact weights are used as the quality correlation strength and efficiency constraint strength, respectively, and assigned to the corresponding virtual edges in the initial production network model. Based on the weighted virtual edges, process node pairs with quality association strength exceeding the first threshold are identified in the initial production network model, and their relationship is defined as a strong quality association relationship. In the initial production network model, process node pairs with efficiency constraints exceeding the second threshold are identified, and their relationships are defined as key efficiency constraints.

[0008] Furthermore, the improved process correlation algorithm calculates the impact weight of quality anomalies in preceding processes on the quality pass rate of subsequent processes based on the spatiotemporal correlation coefficient and the quality fluctuation transmission coefficient, specifically including: The improved process correlation algorithm extracts the quality pass rate sequence and key process parameter sequence of each process node within a specified historical time period from standardized real-time manufacturing data. The quality pass rate sequence is time-aligned with the key process parameter sequence, and the conditional mutual information between the quality pass rate of the preceding process node and the quality pass rate of the following process node is calculated as a preliminary measure of the quality dependence between the preceding and following process nodes. The improved process correlation algorithm incorporates the spatiotemporal correlation coefficient and the quality fluctuation transmission coefficient as adjustment factors into the calculation process of the conditional mutual information. When calculating conditional mutual information, if a quality anomaly event is detected in the preceding process node, the algorithm will amplify the statistical impact of the quality anomaly event on the quality pass rate sequence of subsequent process nodes within the time window based on the quality fluctuation transmission coefficient. Meanwhile, the algorithm will weight the transmission time of quality anomaly events based on the spatiotemporal correlation coefficient. The closer the process nodes are and the more closely connected the logistics are, the greater the weighting value of the transmission time of quality anomaly events. Finally, the improved process correlation algorithm outputs a composite weight value that integrates direct statistical dependence, the intensity of abnormal event impact, and spatiotemporal transmission effect, as the impact weight.

[0009] Furthermore, based on the identified correlations and constraints, a dynamic monitoring strategy is generated. This dynamic monitoring strategy includes instructions for adjusting the quality monitoring frequency of key processes and instructions for adjusting the material supply priority, specifically including: Based on the identified strong quality correlations, the process nodes that are at the core of the relationship and have the greatest impact on the quality of the final product are identified and marked as quality-critical processes. Based on the identified key efficiency constraints, the process nodes with the most obvious bottlenecks are determined and marked as efficiency-critical processes. Obtain the current preset quality monitoring frequency for key quality processes, calculate the quality risk coefficient based on its influence weight, and adjust the preset quality monitoring frequency proportionally upwards or downwards according to the quality risk coefficient to form a quality monitoring frequency adjustment instruction. Obtain the current material supply plan for efficiency-critical processes, calculate the material shortage risk coefficient based on its delay impact weight, and based on the material shortage risk coefficient, increase the delivery priority of the materials required for the efficiency-critical processes in the material supply ranking, thus forming a material supply priority adjustment instruction. The quality monitoring frequency adjustment command and the material supply priority adjustment command are integrated and packaged into a dynamic monitoring strategy command set.

[0010] Furthermore, the dynamic monitoring strategy is used to control the production process of high and low voltage switchgear in real time, and to collect production response data after the control is implemented, specifically including: The dynamic monitoring strategy instruction set is sent to the production line execution layer, wherein the quality monitoring frequency adjustment instruction is sent to the corresponding quality inspection station, and the material supply priority adjustment instruction is sent to the material distribution system. The system receives sampling inspection result data from the quality inspection station based on the adjusted frequency. The sampling inspection result data includes the sampling time, description of non-conformities, and handling measures. The system receives material arrival information from the material distribution system based on the adjusted priority. The material arrival information includes the material code, actual delivery time, and deviation from the planned time. Real-time monitoring of production efficiency data and work-in-process queue length changes at each process station during the execution of the dynamic monitoring strategy; Collect overall production cycle data and final first pass rate data of high and low voltage switchgear at the production line execution layer under strategy control; The sampling inspection results, material arrival information, production efficiency data, work-in-process queue length change data, overall production cycle data, and final first pass rate data are summarized into production response data.

[0011] Furthermore, the iterative optimization of the dynamic monitoring strategy based on production response data and a preset production target model specifically includes: The production response data is acquired, and key performance indicators are extracted from it, including average production cycle, work-in-process inventory level, and quality cost ratio. The key performance indicators are input into a preset production target model, which defines the expected target value and weight of each indicator. Calculate the overall deviation between the current key performance indicator values ​​and the expected target values ​​in the production target model; If the overall deviation exceeds the optimization threshold, the strategy optimization process is initiated to analyze the contribution of each sub-item in the production response data to the overall deviation. Based on the analysis results, it was determined whether the quality monitoring frequency adjustment instruction, the material supply priority adjustment instruction, or both instructions needed optimization. For instructions that need optimization, trace back to the process association and constraint relationships identified by the improved process association algorithm, and fine-tune the historical data time window or threshold used when calculating the impact weight and delay impact weight; Using the fine-tuned parameters, the process correlation analysis is re-executed to generate updated correlation and constraint relationships, and an updated dynamic monitoring strategy is generated accordingly. The calculation of the comprehensive deviation between the current key performance indicator value and the expected target value in the production target model specifically includes: From the production response data, read the actual average production cycle, work-in-process inventory level, and quality cost ratio. From the preset production target model, read the expected values ​​of average production cycle, work-in-process inventory level, quality cost ratio, and the corresponding weight coefficients of each indicator. Calculate the absolute deviations between the actual and expected values ​​of the average production cycle, the actual and expected values ​​of the work-in-process inventory level, and the actual and expected values ​​of the quality cost ratio, respectively. Divide the absolute deviation of each indicator by its corresponding expected value to obtain the relative deviation rate of the indicator. Multiply the relative deviation rate of each indicator by the weight coefficient of the corresponding indicator in the production target model to obtain the weighted deviation value of the corresponding indicator. The weighted deviation values ​​of all indicators are summed to obtain the comprehensive deviation.

[0012] Furthermore, the generated process monitoring and management report specifically includes: It aggregates real-time manufacturing data, identifies process relationships and constraints, generates updated dynamic monitoring strategies, production response data, and strategy iteration and optimization records; Organize the aggregated data and records in chronological order to form a complete log that includes data collection periods, analysis process summaries, strategy execution details, and effect feedback; Extract key information from the complete logs, including the quality anomaly event tracking chain and efficiency bottleneck transfer path, to form a special analysis chapter; Fill in and format the complete log and the special analysis section according to the preset report template; The report includes a rating of the current production line status based on the overall deviation. Generate a process monitoring and management report document that includes process backtracking, special analysis, and status rating.

[0013] Furthermore, the method for constructing the preset production target model specifically includes: Obtain historical production data of the power high and low voltage switch production line, and extract historical key performance indicators from the historical production data. The historical key performance indicators include historical average production cycle, historical work-in-process inventory level, historical quality cost ratio, historical first pass rate and historical equipment comprehensive utilization rate. The system receives user-inputted production line target expectations, which include average production cycle target, work-in-process inventory level target, quality cost ratio target, first pass rate target, and equipment utilization rate target. Based on the extracted historical key performance indicators, the historical average and historical best values ​​are calculated, and the historical average and historical best values ​​are used as reference benchmarks for constructing the production target model. The target expectation value of the production line input by the user is compared with the reference benchmark. If the target expectation value of the production line exceeds the preset percentage range of the historical best value, the target expectation value of the production line is reasonably corrected so that it falls into the feasible range based on the historical best value. Based on the revised production line target expectation, determine the target value of each indicator, and assign corresponding weight coefficients according to the relative importance of each indicator in production management. The target values ​​of each indicator and their corresponding weight coefficients are integrated to construct a mathematical evaluation model that includes quantitative targets and a weight system, which serves as a preset production target model.

[0014] Furthermore, the process node pairs identified in the initial production network model whose efficiency constraint strength exceeds a second threshold, and whose relationship is defined as a critical efficiency constraint relationship, specifically includes: In the initial production network model after assigning weights, all virtual edges are traversed, and virtual edges with efficiency constraint strength values ​​greater than the second preset threshold are selected. For each selected virtual edge, trace the two process nodes it connects to determine the preceding and following process nodes. Analyze the historical completion time data of the preceding process node and the historical start-up waiting time data of the following process node to verify the statistical delay relationship between the preceding and following process nodes; For the verified process nodes, further check whether they are located on the main production path to eliminate the influence of auxiliary or parallel processes; The process node pairs that have passed the verification and are located on the main production path are officially marked as key efficiency constraints, and their preceding process nodes are marked as bottleneck candidate processes. All key efficiency constraints marked and their corresponding efficiency constraint strengths are stored in a relation knowledge base for priority determination when generating dynamic monitoring strategies.

[0015] Compared with the prior art, the beneficial effects of the present invention are: An improved process correlation algorithm is used to analyze real-time manufacturing data from multiple process stations on the power high and low voltage switchgear production line. This algorithm can accurately capture the quality correlation and efficiency constraints between different process stations, breaking the limitations of independent analysis of a single process. It avoids omissions of correlations and misjudgments of constraints caused by isolated data, and allows the mutual influence between processes to be clearly identified. This makes data processing more targeted and accurate, and compared with conventional single data statistics methods, it can more comprehensively reflect the internal connections of each link in the production process.

[0016] Based on the identified quality correlations and efficiency constraints, a dynamic monitoring strategy is generated, which includes instructions for adjusting the frequency of quality monitoring for key processes and instructions for adjusting the priority of material supply. This enables dynamic adaptation of monitoring frequency and material supply, avoiding the lag and irrationality of fixed monitoring strategies. At the same time, the dynamic monitoring strategy is iteratively optimized using the adjusted production response data and the preset production target model, so that the monitoring strategy can be continuously adjusted according to changes in production status, forming a closed loop of monitoring, control, and optimization. Compared with the conventional fixed monitoring mode, this makes the production process smoother and the monitoring more in line with actual production needs. Attached Figure Description

[0017] Figure 1 This is a timing diagram of a power high and low voltage switchgear manufacturing process monitoring and management system according to the present invention; Figure 2 A flowchart illustrating the process of collecting real-time manufacturing data; Figure 3 A flowchart for calculating the influence weights in the improved process correlation algorithm. Detailed Implementation

[0018] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] See Figure 1 This invention provides a monitoring and management system for the manufacturing process of high and low voltage switchgear, specifically including: The system comprises a data acquisition module, a correlation analysis module, a strategy generation module, a real-time control module, and an optimization report module. The data acquisition module deploys data acquisition terminals at multiple process stations along the high and low voltage switchgear production line to collect equipment operating parameters, material consumption rates, work-in-process images, operator identification, and process completion timestamps for sheet metal processing, busbar assembly, component installation, wiring debugging, withstand voltage testing, and casing painting processes. The collected data is standardized to form real-time manufacturing data. The correlation analysis module, based on the standardized real-time manufacturing data, uses an improved process correlation algorithm to calculate the spatiotemporal correlation coefficient and quality fluctuation transmission coefficient between processes. This leads to the derivation of quality correlation strength and efficiency constraint strength, constructing a weighted production network model to identify strong quality correlations and key efficiency constraints. The strategy generation module, based on the identified correlations and constraints, locates quality-critical and efficiency-critical processes, calculates quality risk coefficients and material shortage risk coefficients, and generates a dynamic monitoring strategy that includes quality monitoring frequency adjustment instructions and material supply priority adjustment instructions. The real-time control module distributes the strategy instruction set to the production line execution layer, driving adjustments in the quality inspection station and material distribution system. During strategy execution, it simultaneously collects sampling inspection results, material arrival information, production efficiency data, work-in-process queue length changes, overall production cycle time data, and final first-pass yield data, summarizing them into production response data. The optimization report module extracts key performance indicators from the production response data, inputs them into a preset production target model to calculate the comprehensive deviation. When the deviation exceeds a threshold, it reverses the adjustment of the correlation algorithm parameters and updates the strategy. Simultaneously, based on the entire process data, it generates a process monitoring and management report containing complete logs, specialized analysis, and status ratings.

[0020] In one embodiment of the present invention, see [reference] Figure 2Industrial IoT gateways and vision sensors with Ethernet communication capabilities are deployed at each workstation in sheet metal processing, busbar assembly, component installation, wiring debugging, withstand voltage testing, and shell painting processes to form a data acquisition terminal array. These terminals continuously capture equipment current and voltage signals, tool feed speed, and fastening torque readings as equipment operating parameters at each station. They record the unit-time consumption of sheet metal and auxiliary materials as material consumption rates, use high-resolution industrial cameras to periodically capture images of semi-finished products during assembly as work-in-process images, scan RFID tags on operator badges to obtain operator identification, and record the time each workpiece leaves the workstation as a process completion timestamp triggered by the PLC clock. After data transmission to the edge server, the NTP protocol is used to synchronize and calibrate all time sources, packaging the multi-source heterogeneous data into a process raw data packet with a global sequence number at the millisecond time level. The data packet is separated into a structured parameter table and an image binary stream by a parsing service. A convolutional neural network model is applied to the image data to extract contour and texture features, identify the busbar overlap direction, conductor color code consistency, and nameplate affixing coordinates, and output standardized material model codes, assembly position compliance Boolean values, and label clarity scores. The structured process parameters and image feature vectors undergo field matching and integrity checks in the verification module. After outliers are removed, the data is written to the real-time database to form a unified real-time manufacturing dataset.

[0021] In practice, the sheet metal processing station of the high and low voltage switchgear production line is equipped with an industrial intelligent camera of model SICK-VT120 and a Siemens S7-1200 PLC controller to form a data acquisition terminal. The bus assembly process deploys a Pepperl+Fuchs RFID reader and an Omron temperature control instrument integrated terminal. The component installation process installs a Keyence LJ-V7080 laser displacement sensor in conjunction with a Mitsubishi FX5U controller to build a data acquisition node. The data acquisition terminal samples equipment operating parameters once per second, including the main cylinder pressure value of the sheet metal hydraulic press, the angle setting value of the busbar bending machine, and the instantaneous torque reading of the automatic tightening gun; it calculates material consumption once per minute, recording the length reduction of cold-rolled steel sheet coils and the weight difference of copper busbar profiles; the in-process image is captured by an industrial camera 500ms after the workpiece enters the positioning fixture, acquiring a 1024×768 pixel RGB color image; when an operator approaches the workstation, the RFID antenna reads the unique ID code embedded in the chip of the wearable badge as the operator's identification; when the workpiece completes all steps and exits the sensing area, the photoelectric switch triggers the PLC's internal RTC clock to record a process completion timestamp accurate to milliseconds. In some embodiments, the data acquisition terminal for the withstand pressure test process is additionally connected to the pulse counting output of a partial discharge detector, and the atomization pressure of the electrostatic spray gun and the cumulative value of the paint flow meter are collected for the shell spraying process. Timestamps from all sources are synchronized at the microsecond level at the network switch layer using the PTP precision clock protocol. The resulting packets are packaged into raw data packets with 64-bit nanosecond-level time tags. The structure of a single data packet includes a header identifier, the MAC address of the acquisition terminal, a time sequence number, multiple sets of parameter key-value pairs, and an image MD5 checksum.

[0022] Optionally, the original process data packet is transmitted to the workshop edge computing gateway for parsing. The JSON format is used to decompose the structured fields, including floating-point arrays of equipment operating parameters, integer variables of material consumption rates, strings representing operator identification, and ISO8601 time strings for process completion timestamps. Unstructured data, including JPEG binary streams of work-in-process images, is independently extracted and stored in an object storage bucket. Image feature extraction uses a pre-trained ResNet-34 convolutional neural network model. The input image is normalized to 224×224 pixels, and the output layer activation function is Softmax. The recognition categories cover two material types: busbar TMY-100×10 and TMY-80×8; Boolean judgment on the compliance of bolt orientation perpendicularity at the lap joint; and readability score for the nameplate QR code. The structured feature data conversion formula is expressed as: in: This represents the confidence score of the transformed structured features. Representing the Model weight coefficients for each feature channel, The first fully connected layer corresponding to the last layer of the neural network 3D feature vector elements, For feature mapping function, The bias constant is This means the output feature dimension is fixed at 256. The converted data generates standardized key-value pairs containing material model code strings, compliance enumeration values, and clarity percentages. The structured data and the converted structured feature data are merged in a memory cache. Verification rules require that all channel values ​​of equipment operating parameters must be within the manufacturer-defined safe operating range, material consumption must be non-negative and not exceed the total inventory of the previous period, operator identification must be present in the HR system whitelist, timestamps must monotonically increase without rollback, and image feature confidence scores above a threshold of 0.85 are required to pass verification. Finally, the data is written to a time-series database to form standardized real-time manufacturing data records.

[0023] It is understandable that, in a sample data collection of a certain batch of sheet metal processing, an abnormal peak value of 185 bar was collected when the hydraulic press main cylinder pressure was within the normal range of 180±5 bar. The material consumption record showed that the length of the cold-rolled steel sheet coil decreased by 2.35 meters, corresponding to a negative deviation of 0.05 meters from the theoretical consumption of 2.40 meters. Image recognition showed that the excessive burrs were caused by wear of the punching die. The comparative data of the same batch of busbar assembly process showed that the bending angle was set at 45° but the actual measurement was 47°. The copper busbar consumption weight deviation was +1.2kg, exceeding the tolerance limit. The RFID read the ID code of an unauthorized temporary operator. In both data collection examples, the standardized processing steps marked the abnormal fields and retained the original timestamp alignment relationship. The feature extraction output busbar model TMY-100×10 recognition accuracy rate was 98.7%, which was better than the comparison benchmark of 95.2% for manual visual inspection. The compliance judgment response time of 200ms was 15 times faster than the traditional quality inspector's 3-second single-piece inspection.

[0024] In one embodiment of the present invention, a standardized real-time manufacturing data input modeling engine establishes an initial production network model with a directed graph structure, using process codes as node indices, along the material flow direction. For any two process nodes, time-series data from 30 consecutive production batches are selected, and the Pearson correlation coefficient between equipment start-up / shutdown events and output rhythm within a sliding window is calculated to obtain the spatiotemporal correlation coefficient. The conditional probability increment of the rejection rate in the subsequent process when a process deviation occurs in the preceding process is statistically analyzed, and after normalization, the quality fluctuation transmission coefficient is obtained. (See also...) Figure 3The improved process correlation algorithm calls upon the historical one-year quality pass rate curve and critical dimension tolerance sequence. After time alignment, it calculates the conditional mutual information between the pass rates of preceding and subsequent nodes. In this statistical calculation, a quality fluctuation transmission coefficient is introduced as a gain factor. When the flag of a preceding abnormal event is detected as true, the covariance contribution of the subsequent pass rate fluctuation within the time window is amplified by the transmission coefficient. Simultaneously, the spatiotemporal correlation coefficient is combined with an exponentially decaying weighting of the lag time step, giving higher timeliness weights to process pairs that are physically adjacent and have short logistics handover intervals. The algorithm ultimately outputs a composite value that integrates conditional mutual information, abnormal gain, and timeliness weighting, which is assigned as the quality correlation strength. An analogous method is used to calculate the mutual information between preceding completion delays and subsequent start-up waiting times, and the efficiency constraint strength is obtained by superimposing the logistics connection frequency weight. After assigning weights, all virtual edges in the production network are traversed, and edges with efficiency constraint strength greater than the second threshold of 0.85 are selected. The nodes of the punching and bending processes connected to these edges are backtracked, and the Spearman rank correlation between the punching delay time and the bending waiting time in the historical logs is verified. After confirming that the correlation is statistically significant, the node is determined to be located on the main sheet metal production path. It is marked as a key efficiency constraint relationship, and the punching process is marked as a bottleneck candidate process and entered into the relation knowledge base.

[0025] In the production process of high and low voltage switchgear, the precision of the cabinet structure formed by sheet metal processing directly affects the subsequent busbar lap gap and insulation distance. In the busbar assembly process, the flatness of the copper busbar lap surface and the tightening torque directly determine the contact resistance and withstand voltage test pass rate of the entire cabinet. In the component installation process, the positional deviation of circuit breakers, transformers, and other devices changes the phase-to-phase and phase-to-ground insulation gaps. In the wiring and commissioning process, the wire stripping length and crimping quality affect the loop resistance and insulation strength. In the withstand voltage test process, the leakage current and partial discharge values ​​directly reflect the overall quality level of the preceding processes. In the shell coating process, the coating thickness and adhesion affect the cabinet's corrosion resistance and protection level. Based on these physical and electrical characteristics, the improved process correlation algorithm, when calculating the quality fluctuation transmission coefficient, uses the breakdown voltage threshold and partial discharge pulse number of the withstand voltage test process as key process parameters for the quality pass rate sequence, and uses the change in contact resistance of the lap surface in the busbar assembly process and the phase-to-phase distance deviation in the component installation process as feature vectors of quality anomalies in the preceding processes. The algorithm presets different baseline values ​​for the quality fluctuation conduction coefficient based on the voltage level of the high- and low-voltage switchgear (e.g., 10kV, 35kV), with higher voltage levels corresponding to larger conduction coefficients to reflect the progressive amplification effect of insulation failure risk. Simultaneously, in calculating the spatiotemporal correlation coefficient, the algorithm sets the material flow connection time thresholds between consecutive processes such as sheet metal processing, busbar assembly, and component installation to the same order of magnitude, based on the fixed workstation layout and conveyor belt cycle time of the high- and low-voltage switchgear production line. For process pairs with buffer zones, such as drying after spraying, an exponential decay weighting method is used to reduce their time-related weighting values. Through these methods, the input data, parameter settings, and weighting rules of the improved process correlation algorithm are all closely coupled with the material characteristics, electrical safety requirements, and process flow of the high- and low-voltage switchgear, rather than using general data processing methods.

[0026] In practical implementation, the correlation analysis module receives standardized real-time manufacturing data from the data acquisition module. This data covers records of 30 consecutive production batches across six process stations: sheet metal processing, busbar assembly, component installation, wiring debugging, withstand voltage testing, and casing painting. The module initializes an initial production network model. The node set includes the six process nodes, and the virtual edge set is set according to the process route diagram: five directed edges for sheet metal processing → busbar assembly → component installation → wiring debugging → withstand voltage testing → casing painting, and three parallel auxiliary edges. The spatiotemporal correlation coefficient calculation selects the end time sequence of the punching machine cycle in the sheet metal processing process and the start time sequence of the bending machine in the busbar assembly process. The time window is set to a 24-hour production shift. The Pearson correlation coefficient between the two sequences within the window is calculated, yielding a value of 0.78. The quality fluctuation transmission coefficient is calculated based on 300 historical anomaly records, showing a conditional probability increment of 0.62 when a punching diameter deviation event occurs in the sheet metal processing process, corresponding to a bending angle deviation in the busbar assembly process. In some embodiments, the spatiotemporal correlation coefficient between the terminal crimping failure event sequence of the wiring commissioning process and the insulation breakdown event sequence of the withstand voltage test process is 0.71, and the quality fluctuation transmission coefficient is 0.58.

[0027] Optionally, the improved process correlation algorithm extracts the daily quality pass rate sequences of the sheet metal processing and bus assembly processes over the past year, aligning the time to the day. It calculates the conditional mutual information of the bus assembly process's quality pass rate under a given condition for the sheet metal processing process's quality pass rate, serving as a preliminary measure of quality dependence. The algorithm introduces a quality fluctuation transmission coefficient of 0.62 as an adjustment factor. When a quality anomaly event is detected in the sheet metal processing process on a given day, the covariance contribution of the bus assembly process's subsequent three-day quality pass rate fluctuation is amplified in the conditional mutual information calculation. Simultaneously, a spatiotemporal correlation coefficient of 0.78 is introduced to weight the transmission time lag. The weighted value for the transmission time efficiency of adjacent processes is set to the reciprocal of the coefficient value, thus enhancing the transmission effect between the sheet metal processing and bus assembly processes. The efficiency constraint strength calculation is performed analogously, extracting the completion time delay sequence of the sheet metal processing process and the start-up waiting time sequence of the bus assembly process. After calculating the conditional mutual information, a logistics connection frequency weight is superimposed. The logistics connection frequency is calculated by dividing the average daily material flow frequency by the maximum capacity frequency. The final output quality correlation strength and efficiency constraint strength are assigned to the corresponding virtual edges of the initial production network model. The quality correlation strength of the edge from the sheet metal processing process to the busbar assembly process is recorded as 0.83, and the efficiency constraint strength is recorded as 0.79.

[0028] Understandably, in the initial production network model after assigning weights, edges with efficiency constraint strength greater than the second threshold of 0.75 are filtered through the virtual edges of the five main paths. The edge from the sheet metal processing process to the bus assembly process has an efficiency constraint strength of 0.79, which meets the condition. Tracing the nodes at both ends confirms that the sheet metal processing process is the preceding process node and the bus assembly process is the following process node. Historical production logs are queried to extract the completion delay distribution of the sheet metal processing process over the past 50 shifts and the corresponding waiting time distribution of the bus assembly process. The Spearman rank correlation coefficient of 0.81 is calculated to verify the statistical delay relationship. The process flow diagram is checked to confirm that this node belongs to the main production path. After eliminating the interference of the deburring auxiliary process, the relationship between the sheet metal processing process and the bus assembly process is officially marked as a key efficiency constraint relationship, and the sheet metal processing process is marked as a bottleneck candidate process. The relationship record is stored in the relationship knowledge base, with fields including the preceding process code, the following process code, the efficiency constraint strength value, the verification timestamp, and the path attribute identifier.

[0029] In practical implementation, the composite weight formula is defined in the process of calculating the quality correlation strength as follows: in: This represents the final output value of the mass correlation strength. Represents the sequence of quality pass rates of preceding processes. Sequence of quality pass rates of subsequent processes In the process parameter sequence Conditional mutual information under conditions The scaling factor for the spatiotemporal correlation coefficient is fixed at 0.5. It is the spatiotemporal correlation coefficient input value. The gain coefficient of the mass fluctuation transmission coefficient is fixed at 1.2. This is the input value for the quality fluctuation transmission coefficient. The formula calculates the relationship between sheet metal processing and busbar assembly processes. The value is 0.57. Substituting the parameters, we get... After normalization and mapping to the [0,1] interval, the final quality association strength is 0.83. The calculation result is written into the production network model metadata for subsequent identification of strong quality associations. When the first threshold is set to 0.70, the node is identified as having a strong quality association.

[0030] The improved process correlation algorithm in the correlation analysis module relies on a conditional mutual information calculation model, which is constructed as follows: The model includes an input layer, a conditional probability estimation layer, a mutual information calculation layer, and a composite weight output layer. The input layer receives standardized real-time manufacturing data, where the quality pass rate sequence is the ratio of the number of qualified parts inspected once per day at each process station to the total number of parts sent for inspection. The key process parameter sequence includes measured values ​​of bending angles in sheet metal processing, torque readings in busbar assembly, and coordinate deviations of installation positions in component installation. The conditional probability estimation layer uses a kernel density estimator, selecting a Gaussian kernel as the kernel function, and determining the bandwidth through ten-fold cross-validation. The mutual information calculation layer uses the quality pass rate sequences of the preceding and subsequent processes as random variables, and the key process parameter sequence as conditional variables to calculate the conditional mutual information value. The algorithm's training process is as follows: daily production records of high and low voltage switchgear production over the past year are extracted from the historical database. After removing abnormal batch data caused by external power outages or material shortages, at least two hundred effective samples from production days are obtained. For each process node pair, the quality pass rate sequences of the preceding and subsequent processes are aligned by time. A sliding window method is used to divide the sequence into subsequences of thirty consecutive production days. Conditional mutual information is calculated for each subsequence, and the median of the results for all subsequences is taken as the initial metric. The training of the spatiotemporal correlation coefficient is based on the timestamp sequence of equipment start-up and shutdown events. The difference sequence between the completion and start-up times of workpieces between adjacent processes is extracted, and the least-squares estimate of the Pearson correlation coefficient is calculated. The training of the quality fluctuation transmission coefficient uses event analysis. Dates where process parameters in the preceding process exceed the three-sigma limit are selected as event days. The decrease in the quality pass rate of the subsequent process relative to the average of the fifteen production days prior to the event day is statistically analyzed within three production days after the event day. The median of the decrease on each event day is normalized to the zero-to-one interval and used as the transmission coefficient. All parameters in the above training steps are calculated independently based on historical data from the power high- and low-voltage switchgear production line, without using common preset values ​​from other industries.

[0031] In one embodiment of the invention, the correlation analysis module outputs a list of strong quality correlations, selecting the busbar crimping process as the core node affecting the overall cabinet insulation performance and marking it as a key quality process; from the set of key efficiency constraints, the wiring and commissioning process is identified as having the highest frequency of stalls due to material shortages and is marked as an efficiency key process. The strategy generator retrieves the current preset quality monitoring frequency of once per hour for the busbar crimping process, multiplies it by its quality correlation strength of 0.92 to obtain a quality risk coefficient of 1.84, and accordingly increases the sampling frequency to once every half hour, generating special inspection instructions for busbar insulation layer thickness and overlap surface flatness. The material scheduler reads the historical material shortage records of the wiring and commissioning process, calculates the material shortage risk coefficient based on its efficiency constraint strength of 0.88, and in the ERP interface, upgrades the replenishment priority of cable terminals and relays from ordinary to urgent, generating priority dispatch orders and AGV direct delivery instructions. The two adjustment instructions are combined into a dynamic monitoring strategy instruction set after format encapsulation, with an effective timestamp and applicable production line segment code.

[0032] In practice, the strategy generation module calls the list of strong quality correlations and the list of key efficiency constraints output by the correlation analysis module. The list of strong quality correlations includes a quality correlation strength of 0.94 between the sheet metal processing process and the busbar assembly process, a quality correlation strength of 0.96 between the busbar assembly process and the component installation process, and a quality correlation strength of 0.99 between the wiring debugging process and the withstand voltage test process. The list of key efficiency constraints includes an efficiency constraint strength of 0.73 between the sheet metal processing process and the busbar assembly process, an efficiency constraint strength of 0.68 between the component installation process and the wiring debugging process, and an efficiency constraint strength of 0.77 between the withstand voltage test process and the shell painting process. The module compares the quality correlation strength values ​​of each node pair, selects the wiring and debugging process and the withstand voltage test process node pair with the highest quality correlation strength and located at the end of the main production path, and determines that the withstand voltage test process is the node with the greatest impact on the final insulation performance of the entire cabinet, and marks it as a quality critical process; it scans the list of critical efficiency constraints and filters the withstand voltage test process and the shell spraying process node pair corresponding to the maximum efficiency constraint strength, and identifies the withstand voltage test process as the bottleneck source with the highest equipment failure downtime frequency, and marks it as an efficiency critical process.

[0033] Optionally, the strategy generator accesses the production management system to obtain the current preset quality monitoring frequency for the withstand voltage testing process, which is once every 2 hours. Based on its quality correlation strength of 0.99, a quality risk coefficient is calculated, using a power function transformation to eliminate the influence of frequency dimensions. The material supply plan query for the withstand voltage testing process shows that the current replenishment priority for insulation materials is at the regular level. Based on its efficiency constraint strength of 0.77, a material shortage risk coefficient is calculated, and the risk level is divided based on a segmented mapping of coefficient values. The logic for generating the quality monitoring frequency adjustment instruction is to double the frequency when the quality risk coefficient is greater than 0.900. The calculated quality risk coefficient for the withstand voltage testing process is 0.991, meeting the condition. The instruction content is modified to change the sampling interval to once every hour, and adds two new items: partial discharge testing and dielectric loss testing. The material supply priority adjustment instruction, based on a material shortage risk coefficient of 0.770 belonging to the medium risk range, upgrades the delivery priority of insulation boards and epoxy resin from regular to high, and specifies a shorter transit time for the AGV transport path.

[0034] It is understandable that the strategy generation module outputs a dynamic monitoring strategy instruction set containing two independent instructions: Instruction 1 is a quality monitoring frequency adjustment instruction, instruction code QMF-20250416-011, applicable to the withstand voltage test station of the process, effective time April 16, 2025, 14:00, parameter sampling cycle 60 minutes, and the test items added are partial discharge test and dielectric loss test; Instruction 2 is a material supply priority adjustment instruction, instruction code MPP-20250416-012, applicable to the materials insulation board GPO-3 and epoxy resin EP-828, target process withstand voltage test station, priority marked high, delivery mode AGV express line. The two instructions are bound to the same strategy batch number STR-20250416-B, packaged as a structured JSON format data packet, the packet header containing the strategy version number V2.1, generation timestamp, and applicable production line code PL-02.

[0035] In practice, the calculation process for the quality risk coefficient and the material shortage risk coefficient is recorded in the strategy generation log, and the quality correlation strength of the withstand voltage test procedure is also recorded. The value is 0.99, representing the quality risk coefficient. The calculation formula is expressed as follows: in: This represents the output value of the dimensionless quality risk coefficient. It is a natural constant. The shape parameter is fixed at 3.0. This represents the input value for the correlation strength of the process quality. Substituting the parameters, the calculation yields... The system has a built-in threshold. It is 0.850, because Trigger frequency adjustment rules. The new frequency after adjustment is set according to the strength grading rules. When the quality correlation strength is greater than 0.975, the frequency is set to 30 minutes / time. When it is between 0.925 and 0.975, the frequency is set to 60 minutes / time. The sampling frequency for the withstand pressure test procedure is updated to 1 hour / time. See Table 1.

[0036] Table 1: Material Priority Adjustment Reference Table for Pressure Resistance Testing Process In some embodiments, when the busbar assembly process is marked as a quality-critical process, its quality correlation strength is 0.964, the calculated quality risk coefficient is 0.943, and if it exceeds the threshold of 0.920, the trigger frequency is increased, and an instruction is generated to increase the copper busbar overlap surface flatness detection frequency to once every 45 minutes; when the sheet metal processing process is marked as an efficiency-critical process, the priority of stainless steel plate delivery is increased from normal to medium, and forklift transfer is changed to direct delivery by electric pallet truck.

[0037] In one embodiment of the present invention, the real-time control module pushes the dynamic monitoring strategy instruction set to the production line MES terminal via the MQTT protocol. The quality monitoring frequency adjustment instruction is distributed to the PLC controller of the withstand voltage test station, triggering the test bench to increase the sampling quantity and shorten the testing cycle. The material supply priority adjustment instruction is sent to the WMS scheduling host to modify the warehouse outbound queue rules. The system listens to the sampling result messages uploaded by the withstand voltage test station, recording the timestamp of each test, the description of the breakdown voltage exceeding the standard, and the retest handling code. It receives the material delivery confirmation slip returned by the WMS, including the cable specification code, the actual arrival time in the buffer zone, and the number of minutes of delay compared to the planned time. Throughout the entire strategy execution cycle, the system polls the number of completed work orders and the working hours statistics for each process in real time, calculates the hourly output rate as production efficiency data, monitors the backlog of workpieces in the buffer area to generate a queue length waveform, collects the total time taken from the first process material input to the finished product warehousing to obtain the overall production cycle time, and summarizes the first-pass yield record of the final inspection station as the final first-pass yield data. All feedback items are aggregated into a structured production response dataset.

[0038] In practical implementation, the real-time control module receives the dynamic monitoring strategy instruction set output by the strategy generation module. This instruction set includes instructions for adjusting the quality monitoring frequency and adjusting the material supply priority, targeting the withstand voltage testing process and the component installation process, respectively. The module connects the Siemens S7-1500 PLC controller at the withstand voltage testing station to the WMS scheduling host of the warehouse management system via an industrial Ethernet switch. Instruction messages are published using the MQTT protocol, with the message payload encapsulating instruction parameters in JSON format. The quality monitoring frequency adjustment instruction is sent to the PLC controller at the withstand voltage testing station. The instruction content is parsed, and the internal timer settings are updated, adjusting the insulation resistance test sampling interval from 120 minutes to 60 minutes, triggering the test bench to automatically perform a full batch sampling every hour on the hour. The material supply priority adjustment instruction is sent to the WMS scheduling host. After decoding, the material scheduling queue sorting rules are modified. The outbound priority field of relay JZC1-44 / 380V and contactor CJX2-2510 is updated from "normal" to "urgent." The AGV scheduling algorithm replans the path, bypassing the intermediate storage area and directly reaching the component installation station's shelf coordinates.

[0039] Optionally, the real-time control module listens to the status messages uploaded by the PLC controller of the withstand voltage test station, with a message frequency of once every 10 seconds. It filters and extracts the sampling result data fields: when the sampling trigger flag is detected as true, it records the sampling timestamp 2025-03-14T09:00:00.000Z, parses the test result register to obtain the measured insulation resistance value of 950MΩ, which is lower than the threshold of 1000MΩ, generates the non-conforming item description "insulation resistance is too low", and records the handling measure code R01 to represent retesting; the module synchronously subscribes to the outbound confirmation message queue of the WMS system, receives material arrival information messages, and the messages carry the material code JZC1-44 / 380V, the actual delivery time 2025-03-14T09:12:30.000Z, the planned delivery time 2025-03-14T09:00:00.000Z, and calculates the time deviation +12 minutes and 30 seconds. The module has a built-in periodic polling task that scans the MES process completion records every 5 minutes, counting the number of work orders completed in the component installation process in the past 5 minutes (6 units, total time 48 minutes), and calculating the hourly output rate of 7.5 units / hour as production efficiency data; it monitors the infrared sensor counting pulses in the buffer area and records the fluctuation data of the queue length decreasing from a peak of 22 units to 18 units; it pulls the entire production line order schedule and calculates the total time from the first sheet metal part to the completion of the pressure test (26.4 hours) as the overall production cycle time; it queries the final inspection station database for the last inspection record and obtains the ratio of 97 qualified products to the total output of 100 units (97%) as the final first pass rate data.

[0040] It is understandable that multi-source data collected during the execution of the dynamic monitoring strategy is aggregated into a production response dataset. The data structure follows a unified timestamp alignment principle, and each record is associated with the strategy batch number STR-20250314-A. The sampling inspection result data table includes fields for timestamp, test item, measured value, judgment result, and disposal code. The material arrival information table includes fields for material code, actual time, planned time, and deviation value. The production efficiency data table stores fields for process code, output quantity, working hours, and efficiency value. The work-in-process queue length change data table records fields for sensor ID, counting time, and queue length. The overall production cycle time data table stores fields for order number, start time, end time, and total duration. The final first-pass yield data table stores fields for batch number, number of qualified products, total number, and ratio. The dataset is written to a time-series database, with the collection time and strategy version number as index fields for subsequent optimization module calls.

[0041] In practice, the hourly output rate is calculated using the standard productivity formula, and the output quantity of each process within the statistical window is used. For 6 items, total working hours For 48 minutes, the hourly output rate is The calculation formula is expressed as follows: in: This represents the output rate per unit hour. This represents the total number of completed workpieces within the statistics window. This represents the total number of working minutes, with a constant of 60 serving as the conversion factor from minutes to hours. Substituting the component installation process data, the calculation yields... The number of pieces per hour is recorded in the production efficiency data table entry, see Table 2.

[0042] Table 2: Data Table of Spot Check Results for Pressure Resistance Testing Station In some embodiments, when the shell spraying process executes the quality monitoring frequency adjustment instruction, the module receives the film thickness detection data uploaded by the spraying robot controller and records the sampling time and thickness deviation value; after the material delivery priority of the bus assembly process is adjusted, the module collects the deviation of the AGV arrival time from the planned time by -3 minutes, and counts that the queue length of the bus crimping process is reduced from 15 pieces to 12 pieces, and the overall production cycle time is shortened from 28 hours to 27.2 hours.

[0043] In one embodiment of the invention, the optimization report module extracts the average production cycle of 28 hours, work-in-process inventory level of 152 units, and quality rework cost as a percentage of total cost of 4.2% from the production response data for the most recent week as key performance indicators. A preset production target model loads the historical three-year best average cycle of 25 hours, minimum work-in-process inventory of 120 units, and minimum quality cost ratio of 3.5% as reference benchmarks. Combined with the management-set target cycle of 26 hours, target inventory of 130 units, and target cost ratio of 3.8%, the expected target value is established after linear interpolation correction, and weighting coefficients of 0.4, 0.3, and 0.3 are assigned. The calculation module calculates the absolute deviation between the actual value and the expected value of each indicator, divides it by the expected value to obtain the relative deviation rate, multiplies it by the weighting coefficient, and sums it to obtain a comprehensive deviation of 0.18. Because this value is greater than the optimization threshold of 0.15, the diagnostic logic determines that the material supply priority adjustment instruction has not sufficiently alleviated the bottleneck. Tracing back to the improved process correlation algorithm, the historical data window for calculating the efficiency constraint intensity is expanded from 90 days to 180 days to include more seasonal fluctuation samples. The updated algorithm recalculates the adjusted relationships and generates a new dynamic monitoring strategy. The report generator summarizes real-time manufacturing data, relationship change records, strategy versions, and production response data at a monthly granularity, and arranges them into an event log containing the collection period, parameter summary, and execution details. It extracts the causal chain of busbar crimping abnormalities leading to withstand voltage failures and the transmission path of punching bottlenecks causing sheet metal section congestion to form a special analysis step. It fills the data with an XML report template, adds a B-level rating of production line status based on deviation, and outputs a PDF format process monitoring and management report.

[0044] In practical implementation, the optimization report module extracts 30 consecutive working days of operational records from the production response data collected by the real-time control module, filtering out key performance indicator values: the average production cycle is calculated as the arithmetic mean of the total time taken for all completed orders from material input in the first process to finished product warehousing, which is 28.5 hours; the work-in-process inventory level is the daily average of 158 units in the buffer area of ​​each process at the end of each day; and the quality cost ratio is calculated as the proportion of rework and scrap costs plus retesting labor costs to the total production cost of the period, which is 4.38%. The preset production target model is stored in the system configuration library in advance. The model parameters are derived from the best performance values ​​of the historical three-year operational data. After review by management, the target values ​​are set as follows: the expected average production cycle is 26.0 hours, the expected work-in-process inventory level is 135 units, and the expected quality cost ratio is 3.8%. The weight coefficients are allocated according to strategic importance, with the average production cycle weighted at 0.42, the work-in-process inventory level weighted at 0.31, and the quality cost ratio weighted at 0.27.

[0045] Optionally, the comprehensive deviation calculation module reads the actual value array of the current key performance indicators and the expected value array defined by the production target model, and calculates the absolute deviation of each indicator: the absolute deviation of the average production cycle is 28.5 minus 26.0 equals 2.5 hours; the absolute deviation of the work-in-process inventory level is 158 minus 135 equals 23 units; and the absolute deviation of the quality cost ratio is 4.38% minus 3.8% equals 0.58 percentage points. The relative deviation rate is calculated by dividing the absolute deviation by the expected value and then multiplying by 100%. The relative deviation rate of the average production cycle is approximately 9.61% (2.5 divided by 26.0), the relative deviation rate of the work-in-process inventory level is approximately 17.04% (23 divided by 135), and the relative deviation rate of the quality cost ratio is approximately 15.26% (0.58 divided by 3.8). The weighted deviation value is calculated by multiplying each relative deviation rate by its corresponding weighting coefficient. The average production cycle weighted deviation value is 9.61% multiplied by 0.42, which is approximately 4.036%. The work-in-process inventory level weighted deviation value is 17.04% multiplied by 0.31, which is approximately 5.282%. The quality cost ratio weighted deviation value is 15.26% multiplied by 0.27, which is approximately 4.120%. The overall deviation is the sum of all weighted deviation values, and the calculated result is approximately 13.438%. The system's optimization threshold is set at 15%. Since the overall deviation is less than the threshold, the strategy optimization process will not be triggered this time, and only a routine monitoring report will be generated.

[0046] Understandably, when the overall deviation exceeds the optimization threshold, the strategy iteration process is initiated. This involves analyzing the contribution of each component of the production response data to the overall deviation, diagnosing that the main deviation stems from the fact that the timely arrival rate of materials only improved by 12% after the material supply priority adjustment instruction was executed, falling short of the expected 20% target. The process then traces back to the historical data time window used by the improved process correlation algorithm to calculate the efficiency constraint strength. The original window length was 90 days, covering the regular fluctuation cycle but not including recent seasonal fluctuations in the supply chain. The parameters were fine-tuned to extend the historical data time window to 180 days. After re-executing the process correlation analysis, updated correlations and constraints were obtained, generating updated dynamic monitoring strategies to increase the stocking coefficient for scarce materials. The process monitoring management report generator summarizes nearly one month's real-time manufacturing data, identified process correlation and constraint version change records, dynamic monitoring strategy execution logs, and production response data, arranged chronologically to form a complete log. The log includes the data collection period from February 15, 2025 to March 16, 2025, an analysis process summary explaining the reasons for recalculating the quality correlation strength, and strategy execution details recording the number of instructions issued and the execution success rate. The specialized analysis section extracts the temperature anomaly event tracing chain from the busbar crimping process, links it to insulation failure cases in subsequent withstand voltage testing processes, and draws a path diagram showing the shift of efficiency bottlenecks from sheet metal processing to component installation. The report template adopts the enterprise standard XML structure, and after data filling, an overall deviation degree corresponding to a status rating of C is attached. The output PDF document is delivered to the production management department for review.

[0047] During the iterative optimization process, when the overall deviation exceeds the optimization threshold, the optimization report module determines whether the quality monitoring frequency adjustment instruction or the material supply priority adjustment instruction needs optimization. The specific criteria for this determination are as follows: Analyzing the contribution of each component of the production response data to the overall deviation, if the sum of the weighted deviations of the average production cycle and work-in-process inventory level accounts for more than 60% of the overall deviation, then the material supply priority adjustment instruction needs optimization; if the weighted deviation of the quality cost percentage accounts for more than 40% of the overall deviation, then the quality monitoring frequency adjustment instruction needs optimization. For material supply priority adjustment instructions requiring optimization, the system traces back to the historical data time window used when calculating the delay impact weight using the improved process correlation algorithm. If the material arrival timeliness rate has decreased more than the set value recently (e.g., the last thirty production days) compared to the arrival rate earlier in the window, the time window length is shortened to highlight recent fluctuations; if the material arrival timeliness rate exhibits periodic fluctuations (e.g., fixed lows weekly or monthly), the time window length is extended to cover at least two complete fluctuation cycles. For quality monitoring frequency adjustment instructions that require optimization, the process traces back to the quality fluctuation transmission coefficient threshold used when calculating the impact weight. If the process pairs corresponding to the currently identified strong quality correlations show no quality anomaly transmission in multiple consecutive batches during actual production, the first threshold is appropriately increased to reduce excessive intervention. If unidentified process pairs exhibit significant quality transmission phenomena in actual production, the first threshold is decreased to increase sensitivity. After fine-tuning the parameters, the process correlation analysis is re-executed. In the generated updated strategy, the sampling interval change step of the quality monitoring frequency adjustment instructions is an integer multiple of the original interval, and the priority upgrade level of the material supply priority adjustment instructions is determined in segments based on the material shortage risk coefficient, thereby ensuring the interpretability of the strategy adjustment and the feasibility of production operations.

[0048] In practice, the pre-set production target model construction process utilizes five years of archived historical database data to calculate historical key performance indicator benchmarks: the best historical average production cycle of 24.8 hours occurred in the fourth quarter of 2023; the lowest historical work-in-process inventory level of 118 units occurred in the third quarter of 2022; and the lowest historical quality cost ratio of 3.55% occurred in the first quarter of 2024. The user input interface receives the production line's target expectations: an average production cycle target of 25.0 hours, a work-in-process inventory target of 125 units, a quality cost ratio target of 3.65%, a first-pass yield target of 98.5%, and an equipment utilization rate target of 89%. The model validation module compares user input values ​​with historical best values. The average production cycle target of 25.0 hours exceeds the historical best of 24.8 hours by 0.8%, which is within the allowable fluctuation range of 5% and requires no correction. The work-in-process inventory level target of 125 units exceeds the historical best of 118 units by 5.9%, triggering the correction logic. After correction, the target value is adjusted to 122 units. The quality cost ratio target of 3.65% exceeds the historical best of 3.55% by 2.8%, which is within the limit. The weighting system is based on production management priority, with the average production cycle having the highest weight because delivery time directly affects customer satisfaction. The quality cost ratio has the second highest weight, reflecting the cost constraints of quality control. The work-in-process inventory level has the lowest weight, balancing liquidity and space occupancy. The final model structure includes an array of target values, an array of weight coefficients, and a correction rule version number.

[0049] In practical implementation, the formula for calculating the overall deviation is defined as follows: in: This indicates the result of the overall deviation calculation. The total number of key performance indicators is fixed at 3. Representing the Actual observed values ​​of key performance indicators The first in the representative production target model The expected target value of each indicator Representing the The weight coefficients of each indicator are set, and the sum of the weights is 1. The formula calculation process clearly distinguishes between the positive and negative characteristics of the indicators, and all indicators are treated as absolute value deviations to ensure that the positive values ​​of the deviations are accumulated.

[0050] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A monitoring and management system for the manufacturing process of high and low voltage switchgear, characterized in that, include: The data acquisition module is used to collect real-time manufacturing data from multiple process stations on the power high and low voltage switchgear production line; The correlation analysis module is used to analyze the collected real-time manufacturing data based on the improved process correlation algorithm to identify the quality correlation and efficiency constraint relationships between process stations. The strategy generation module is used to generate dynamic monitoring strategies based on the identified correlations and constraints. The dynamic monitoring strategies include instructions for adjusting the quality monitoring frequency of key processes and instructions for adjusting the material supply priority. The real-time control module is used to control the production process of high and low voltage switchgear in real time using the dynamic monitoring strategy, and to collect the production response data after control. The optimization report module is used to iteratively optimize dynamic monitoring strategies based on production response data and preset production target models, and generate process monitoring and management reports. The improved process correlation algorithm analyzes the collected real-time manufacturing data to identify the quality correlation and efficiency constraint relationships between process stations, specifically including: Obtain standardized real-time manufacturing data, and construct an initial production network model based on process stations, with processes as nodes and logistics and information flows as virtual edges; In the initial production network model, the spatiotemporal correlation coefficient and quality fluctuation transmission coefficient of manufacturing data between any two process nodes are calculated. The spatiotemporal correlation coefficient is the Pearson correlation coefficient between the equipment start-up and shutdown events and the output rhythm within a sliding window for any two process nodes, taking time series data of consecutive production batches. The quality fluctuation transmission coefficient is the conditional probability increment of the rejection rate of the subsequent process when the preceding process has a process deviation, which is obtained after normalization. The improved process correlation algorithm calculates the impact weight of quality anomalies in preceding processes on the quality pass rate of subsequent processes, and the delay impact weight of the completion time of preceding processes on the start-up waiting time of subsequent processes, based on the spatiotemporal correlation coefficient and the quality fluctuation transmission coefficient. The impact weight is a composite weight value output by calculating the conditional mutual information between the quality pass rate sequences of preceding and subsequent processes, introducing the quality fluctuation transmission coefficient as a gain factor in the calculation process, and combining it with the spatiotemporal correlation coefficient for exponential decay weighting. The delay impact weight is obtained by calculating the mutual information between the completion delay of preceding processes and the start-up waiting time of subsequent processes, and then adding the logistics connection frequency weight. The calculated impact weights and delay impact weights are used as the quality correlation strength and efficiency constraint strength, respectively, and assigned to the corresponding virtual edges in the initial production network model. Based on the weighted virtual edges, process node pairs with quality association strength exceeding the first threshold are identified in the initial production network model, and their relationship is defined as a strong quality association relationship. In the initial production network model, process node pairs with efficiency constraints exceeding the second threshold are identified, and their relationships are defined as key efficiency constraints.

2. The power high and low voltage switchgear manufacturing process monitoring and management system according to claim 1, characterized in that, The collection of real-time manufacturing data from multiple process stations on the high and low voltage switchgear production line specifically includes: Data acquisition terminals are deployed at the sheet metal processing, busbar assembly, component installation, wiring and debugging, withstand voltage testing and shell spraying stations of the power high and low voltage switch production line; The data acquisition terminal can be used to obtain in real time the equipment operating parameters, material consumption rate, work-in-process images, operator identification, and process completion timestamps of each process station. The equipment operating parameters, material consumption rate, work-in-process images, operator identification, and process completion timestamps are aligned and packaged according to a unified time base to generate a process raw data package with time sequence tags. The original data packets of the aforementioned process are parsed to extract the structured and unstructured data. The extracted unstructured data, including images of work-in-process, is subjected to feature extraction to identify material type, assembly location compliance, and label clarity features in the images, and then converted into structured feature data. The structured data and the transformed structured feature data are merged and verified to form standardized real-time manufacturing data.

3. The power high and low voltage switchgear manufacturing process monitoring and management system according to claim 2, characterized in that, The improved process correlation algorithm calculates the impact weight of quality anomalies in preceding processes on the quality pass rate of subsequent processes based on the spatiotemporal correlation coefficient and the quality fluctuation transmission coefficient, specifically including: The improved process correlation algorithm extracts the quality pass rate sequence and key process parameter sequence of each process node within a specified historical time period from standardized real-time manufacturing data. The quality pass rate sequence is time-aligned with the key process parameter sequence, and the conditional mutual information between the quality pass rate of the preceding process node and the quality pass rate of the following process node is calculated as a preliminary measure of the quality dependence between the preceding and following process nodes. The improved process correlation algorithm incorporates the spatiotemporal correlation coefficient and the quality fluctuation transmission coefficient as adjustment factors into the calculation process of the conditional mutual information. When calculating conditional mutual information, if a quality anomaly event is detected in the preceding process node, the algorithm will amplify the statistical impact of the quality anomaly event on the quality pass rate sequence of subsequent process nodes within the time window based on the quality fluctuation transmission coefficient. Meanwhile, the algorithm will weight the transmission time of quality anomaly events based on the spatiotemporal correlation coefficient. The closer the process nodes are and the more closely connected the logistics are, the greater the weighting value of the transmission time of quality anomaly events. Finally, the improved process correlation algorithm outputs a composite weight value that integrates direct statistical dependence, the intensity of abnormal event impact, and spatiotemporal transmission effect, as the impact weight.

4. The power high and low voltage switchgear manufacturing process monitoring and management system according to claim 3, characterized in that, The process involves generating a dynamic monitoring strategy based on the identified relationships and constraints. This dynamic monitoring strategy includes instructions for adjusting the quality monitoring frequency of key processes and adjusting the material supply priority. Specifically, it includes: Based on the identified strong quality correlations, the process nodes that are at the core of the relationship and have the greatest impact on the quality of the final product are identified and marked as quality-critical processes. Based on the identified key efficiency constraints, the process nodes with the most obvious bottlenecks are determined and marked as efficiency-critical processes. Obtain the current preset quality monitoring frequency for key quality processes, calculate the quality risk coefficient based on its influence weight, and adjust the preset quality monitoring frequency proportionally upwards or downwards according to the quality risk coefficient to form a quality monitoring frequency adjustment instruction. Obtain the current material supply plan for efficiency-critical processes, calculate the material shortage risk coefficient based on its delay impact weight, and based on the material shortage risk coefficient, increase the delivery priority of the materials required for the efficiency-critical processes in the material supply ranking, thus forming a material supply priority adjustment instruction. The quality monitoring frequency adjustment command and the material supply priority adjustment command are integrated and packaged into a dynamic monitoring strategy command set.

5. A monitoring and management system for the manufacturing process of high and low voltage switchgear according to claim 4, characterized in that, The dynamic monitoring strategy is used to control the production process of high and low voltage switchgear in real time and collect production response data after the control is implemented. Specifically, this includes: The dynamic monitoring strategy instruction set is sent to the production line execution layer, wherein the quality monitoring frequency adjustment instruction is sent to the corresponding quality inspection station, and the material supply priority adjustment instruction is sent to the material distribution system. The system receives sampling inspection result data from the quality inspection station based on the adjusted frequency. The sampling inspection result data includes the sampling time, description of non-conformities, and handling measures. The system receives material arrival information from the material distribution system based on the adjusted priority. The material arrival information includes the material code, actual delivery time, and deviation from the planned time. Real-time monitoring of production efficiency data and work-in-process queue length changes at each process station during the execution of the dynamic monitoring strategy; Collect overall production cycle data and final first pass rate data of high and low voltage switchgear at the production line execution layer under strategy control; The sampling inspection results, material arrival information, production efficiency data, work-in-process queue length change data, overall production cycle data, and final first pass rate data are summarized into production response data.

6. The power high and low voltage switchgear manufacturing process monitoring and management system according to claim 5, characterized in that, The iterative optimization of the dynamic monitoring strategy based on production response data and a pre-set production target model specifically includes: The production response data is acquired, and key performance indicators are extracted from it. The key performance indicators include average production cycle, work-in-process inventory level, and quality cost ratio. The key performance indicators are input into a preset production target model, which defines the expected target value and weight of each indicator. Calculate the overall deviation between the current key performance indicator values ​​and the expected target values ​​in the production target model; If the overall deviation exceeds the optimization threshold, the strategy optimization process is initiated to analyze the contribution of each sub-item in the production response data to the overall deviation. Based on the analysis results, it was determined whether the quality monitoring frequency adjustment instruction, the material supply priority adjustment instruction, or both instructions needed optimization. For instructions that need optimization, trace back to the process association and constraint relationships identified by the improved process association algorithm, and fine-tune the historical data time window or threshold used when calculating the impact weight and delay impact weight; Using the fine-tuned parameters, the process correlation analysis is re-executed to generate updated correlation and constraint relationships, and an updated dynamic monitoring strategy is generated accordingly. The calculation of the comprehensive deviation between the current key performance indicator value and the expected target value in the production target model specifically includes: From the production response data, read the actual average production cycle, work-in-process inventory level, and quality cost ratio. From the preset production target model, read the expected values ​​of average production cycle, work-in-process inventory level, quality cost ratio, and the corresponding weight coefficients of each indicator. Calculate the absolute deviations between the actual and expected values ​​of the average production cycle, the actual and expected values ​​of the work-in-process inventory level, and the actual and expected values ​​of the quality cost ratio, respectively. Divide the absolute deviation of each indicator by its corresponding expected value to obtain the relative deviation rate of the indicator. Multiply the relative deviation rate of each indicator by the weight coefficient of the corresponding indicator in the production target model to obtain the weighted deviation value of the corresponding indicator. The weighted deviation values ​​of all indicators are summed to obtain the comprehensive deviation.

7. The power high and low voltage switchgear manufacturing process monitoring and management system according to claim 6, characterized in that, The generated process monitoring and management report specifically includes: It aggregates real-time manufacturing data, identifies process relationships and constraints, generates updated dynamic monitoring strategies, production response data, and strategy iteration and optimization records; Organize the aggregated data and records in chronological order to form a complete log that includes data collection periods, analysis process summaries, strategy execution details, and effect feedback; Extract key information from the complete logs, including the quality anomaly event tracking chain and efficiency bottleneck transfer path, to form a special analysis chapter; Fill in and format the complete log and the special analysis section according to the preset report template; The report includes a rating of the current production line status based on the overall deviation. Generate a process monitoring and management report document that includes process backtracking, special analysis, and status rating.

8. A monitoring and management system for the manufacturing process of high and low voltage switchgear according to claim 7, characterized in that, The construction method of the preset production target model specifically includes: Obtain historical production data of the power high and low voltage switch production line, and extract historical key performance indicators from the historical production data. The historical key performance indicators include historical average production cycle, historical work-in-process inventory level, historical quality cost ratio, historical first pass rate and historical equipment comprehensive utilization rate. The system receives user-inputted production line target expectations, which include average production cycle target, work-in-process inventory level target, quality cost ratio target, first pass rate target, and equipment utilization rate target. Based on the extracted historical key performance indicators, the historical average and historical best values ​​are calculated, and the historical average and historical best values ​​are used as reference benchmarks for constructing the production target model. The target expectation value of the production line input by the user is compared with the reference benchmark. If the target expectation value of the production line exceeds the preset percentage range of the historical best value, the target expectation value of the production line is reasonably corrected so that it falls into the feasible range based on the historical best value. Based on the revised production line target expectation, determine the target value of each indicator, and assign corresponding weight coefficients according to the relative importance of each indicator in production management. The target values ​​of each indicator and their corresponding weight coefficients are integrated to construct a mathematical evaluation model that includes quantitative targets and a weight system, which serves as a preset production target model.

9. A monitoring and management system for the manufacturing process of high and low voltage switchgear according to claim 8, characterized in that, The process node pairs identified in the initial production network model whose efficiency constraint strength exceeds the second threshold are defined as key efficiency constraint relationships, specifically including: In the initial production network model after assigning weights, all virtual edges are traversed, and virtual edges with efficiency constraint strength values ​​greater than the second preset threshold are selected. For each selected virtual edge, trace the two process nodes it connects to determine the preceding and following process nodes. Analyze the historical completion time data of the preceding process node and the historical start-up waiting time data of the following process node to verify the statistical delay relationship between the preceding and following process nodes; For the verified process nodes, further check whether they are located on the main production path to eliminate the influence of auxiliary or parallel processes; The process node pairs that have passed the verification and are located on the main production path are officially marked as key efficiency constraints, and their preceding process nodes are marked as bottleneck candidate processes. All key efficiency constraints marked and their corresponding efficiency constraint strengths are stored in a relation knowledge base for priority determination when generating dynamic monitoring strategies.

Images