Bus jack box production monitoring method and system based on multi-sensor fusion

Abstract

The invention discloses a bus jack box production monitoring method and system based on multi-sensor fusion, and relates to the related technical field of production monitoring, and the method comprises the steps: collecting the multi-source operation parameters of a bus jack box in real time through multiple sensors; carrying out adaptive gain control on the heterogeneous sensing data set, and carrying out weighted fusion according to a gain data set to generate a standardized feature vector; performing state identification evaluation based on the standardized feature vector to obtain a state evaluation result; and carrying out production monitoring analysis according to the state evaluation result, and generating a monitoring instruction to carry out real-time regulation and control early warning on the production process of the bus jack box. The technical problems that in the prior art, the multi-dimensional operation state of the bus jack box cannot be comprehensively captured, the data utilization efficiency is low, state evaluation is inaccurate, early warning regulation and control are not timely, and consequently the production efficiency and reliability are insufficient are solved. The technical effects of realizing accurate monitoring and real-time early warning of the production process of the bus jack box and improving the production efficiency and reliability are achieved.

Description

Technical Field

[0001] This invention relates to the field of production monitoring technology, specifically to a method and system for production monitoring of busbar junction boxes based on multi-sensor fusion. Background Technology

[0002] Busbar junction boxes are key components in power distribution systems. Their main function is to safely and efficiently distribute electrical energy from the main busbar to various electrical devices. Their operational reliability and stability directly affect the safety and efficiency of the entire power supply system. As the manufacturing process of busbar junction boxes becomes increasingly complex, traditional production monitoring relies heavily on single sensors or manual inspections, resulting in limited data sources, insufficient real-time performance, and high false alarm rates, making it difficult to meet the demands for high precision and high reliability in production. Existing busbar junction box monitoring typically relies on threshold judgments based on parameters such as temperature, current, or voltage, which cannot comprehensively reflect the overall operating status of the equipment. For example, relying solely on temperature monitoring may overlook potential risks such as current fluctuations or decreased insulation performance, leading to untimely warnings or inaccurate control. Furthermore, due to the complex operating environment of busbar junction boxes, their sensor data often exhibits multi-source heterogeneous characteristics, including electrical, mechanical, and environmental parameters, which differ significantly in format, dimensions, and sampling frequency. Directly performing simple superposition analysis can easily lead to information redundancy or feature loss, thus affecting the accuracy of status assessment.

[0003] Therefore, current technologies suffer from several technical problems, including the inability to fully capture the multi-dimensional operating status of busbar junction boxes, low data utilization efficiency, inaccurate status assessment, and untimely early warning and control, leading to insufficient production efficiency and reliability. Summary of the Invention

[0004] This application provides a busbar plug-in box production monitoring method and system based on multi-sensor fusion, which solves the technical problems in the prior art, such as the inability to fully capture the multi-dimensional operating status of busbar plug-in boxes, low data utilization efficiency, inaccurate status assessment, and untimely early warning and control, resulting in insufficient production efficiency and reliability. It achieves the technical effect of realizing accurate monitoring and real-time early warning of the busbar plug-in box production process and improving production efficiency and reliability.

[0005] This application provides a production monitoring method for busbar plug-in boxes based on multi-sensor fusion. The method includes: collecting multi-source operating parameters of the busbar plug-in box in real time through multiple sensors to construct a heterogeneous sensor data set; performing adaptive gain control on the heterogeneous sensor data set, and weighted fusion based on the gain dataset to generate a standardized feature vector; performing state identification and evaluation based on the standardized feature vector to obtain a state evaluation result; and performing production monitoring analysis based on the state evaluation result to generate monitoring instructions for real-time regulation and early warning of the busbar plug-in box production process.

[0006] In a possible implementation, the busbar plug-in box production monitoring method based on multi-sensor fusion further performs the following processing: traversing the busbar plug-in boxes to perform production impact analysis, and determining multiple key locations based on the impact analysis results; constructing a time synchronization mechanism based on the data acquisition sequence, and deploying a sensor array based on the multiple key locations according to the time synchronization mechanism; acquiring multi-source operating parameters by real-time data acquisition of the busbar plug-in boxes through the sensor array; extracting data timestamps according to the time synchronization mechanism, performing structural analysis on the multi-source operating parameters according to the data timestamps, and determining a preset data structure; encapsulating the multi-source operating parameters according to the preset data structure to construct the heterogeneous sensor data set.

[0007] In a possible implementation, the busbar junction box production monitoring method based on multi-sensor fusion further performs the following processing: the sensor array includes a temperature sensor, a current sensor, a voltage sensor, and a vibration sensor; the temperature sensor performs contact sensing on the conductive connection points of the busbar junction box to obtain contact temperature data; the current sensor performs current sensing on the busbar conductors of the busbar junction box to obtain load current data; the voltage sensor performs voltage sensing on the phase wires and ground wires between the busbar junction boxes to obtain voltage data; the vibration sensor performs vibration sensing on the mechanical structure of the busbar junction box to obtain vibration spectrum data; and the contact temperature data, the load current data, the voltage data, and the vibration spectrum data are integrated to construct the multi-source operating parameters.

[0008] In a possible implementation, the busbar junction box production monitoring method based on multi-sensor fusion further performs the following processing: traversing the heterogeneous sensor data set to perform signal quality assessment on the sensor arrays respectively, and calculating the instantaneous signal strength of the sensor arrays; performing gain analysis based on the instantaneous signal strength, setting gain coefficients, and applying the gain coefficients to perform real-time gain control on the sensor arrays to generate a gain dataset; performing quality analysis based on the gain dataset, setting data quality indicators, and assigning weights to the gain dataset according to the data quality indicators to calculate multiple weight coefficients; performing weighted fusion calculation on the heterogeneous sensor data set based on the multiple weight coefficients to generate primary fusion features; and normalizing the primary fusion features to construct the standardized feature vector.

[0009] In a possible implementation, the busbar junction box production monitoring method based on multi-sensor fusion further performs the following processing: performing intensity extreme value analysis on the sensor array based on the heterogeneous sensor data set to determine the maximum and minimum signal strength values; setting a target intensity range by using the maximum signal strength value as the upper limit of the range and the minimum signal strength value as the lower limit of the range; comparing the instantaneous signal strength with the preset target intensity range; if the instantaneous signal strength is lower than the lower limit of the range, generating an increase command, performing gain analysis based on the increase command, and setting a first gain coefficient; amplifying the sensor array based on the first gain coefficient to generate a first gain dataset; if the instantaneous signal strength is higher than the upper limit of the range, generating a decrease command, performing gain analysis based on the decrease command, and setting a second gain coefficient; attenuating the sensor array based on the second gain coefficient to generate a second gain dataset.

[0010] In a possible implementation, the busbar plug-in box production monitoring method based on multi-sensor fusion further performs the following processing: deep confidence analysis is performed based on the standardized feature vector; feature extraction is performed based on the confidence level to determine deep feature parameters; the busbar plug-in box is classified into operating states to obtain multiple state categories; the distribution of the multiple state categories is calculated based on the deep feature parameters to obtain multiple state category probability distribution parameters; multi-timescale analysis is performed based on the multiple state category probability distribution parameters to obtain multiple timescale features, including short-term transient features, medium-term trend features, and long-term degradation features; a comprehensive evaluation is performed based on the short-term transient features, the medium-term trend features, and the long-term degradation features to obtain multi-timescale state evaluation results for anomaly risk identification, and multiple state anomaly risk levels are classified; the multiple state anomaly levels, the confidence level, and the multi-timescale state evaluation results are added to the state evaluation results.

[0011] In a possible implementation, the busbar plug-in box production monitoring method based on multi-sensor fusion further performs the following processing: constructing a production control strategy library, using the state evaluation results as an index to search and match the production control strategy library, and determining a control scheme based on the matching results; performing production monitoring analysis on the busbar plug-in box according to the control scheme, and generating monitoring instructions; issuing the monitoring instructions to the production line execution unit to perform real-time control feedback on the busbar plug-in box, and generating a control feedback parameter set; performing control analysis based on the control feedback parameter set to obtain real-time control effect data; performing dynamic iterative optimization by backtracking to the monitoring instructions according to the real-time control effect data, generating monitoring optimization instructions to perform closed-loop control on the production process of the busbar plug-in box, and constructing a control log; performing anomaly analysis on the production control of the busbar plug-in box according to the control log, extracting control anomaly data for real-time early warning.

[0012] In a possible implementation, the busbar plug-in box production monitoring method based on multi-sensor fusion further performs the following processing: retrieving historical production control datasets of the busbar plug-in box for state analysis, dividing multiple control state data for strategy matching, and determining normal state strategy sets, early warning state strategy sets, and alarm state strategy sets; integrating the normal state strategy sets, early warning state strategy sets, and alarm state strategy sets to construct a three-level strategy library structure; mapping the historical production control datasets to the three-level strategy library structure to construct a production control strategy library; calculating the state features of the state assessment results based on the multiple state anomaly levels, the confidence level, and the multi-timescale state assessment results, determining the feature contribution degree, and constructing a multi-level index tree based on the feature contribution degree; traversing the production control strategy library based on the multi-level index tree to perform fuzzy matching, obtaining the scheme matching degree, and generating the control scheme based on the scheme matching degree.

[0013] In a possible implementation, the busbar plug-in box production monitoring method based on multi-sensor fusion further performs the following processing: performing time-series parsing on the control log to obtain multiple time-series indicator features, including instruction issuance timestamp, execution start timestamp, and execution completion timestamp; performing control response calculation based on the instruction issuance timestamp, the execution start timestamp, and the execution completion timestamp to obtain control response time parameters; performing execution calculation based on the control response time parameters according to parameter stability indicators to obtain execution deviation; and identifying control anomalies in busbar plug-in box production control according to the execution deviation to determine control anomaly data.

[0014] This application also provides a busbar plug-in box production monitoring system based on multi-sensor fusion. The system includes: an operation parameter acquisition module, used to acquire multi-source operation parameters of the busbar plug-in box in real time through multiple sensors to construct a heterogeneous sensor data set; a gain control module, used to perform adaptive gain control on the heterogeneous sensor data set, perform weighted fusion based on the gain dataset, and generate a standardized feature vector; a state recognition and evaluation module, used to perform state recognition and evaluation based on the standardized feature vector to obtain a state evaluation result; and a production monitoring and analysis module, used to perform production monitoring and analysis based on the state evaluation result, and generate monitoring instructions to perform real-time regulation and early warning of the busbar plug-in box production process.

[0015] This application proposes a multi-sensor fusion-based busbar plug-in box production monitoring method and system. This system collects multi-source operating parameters of the busbar plug-in box in real time using multiple sensors; performs adaptive gain control on the heterogeneous sensor data set; weights and fuses the gain dataset to generate a standardized feature vector; performs state identification and evaluation based on the standardized feature vector to obtain the state evaluation result; and performs production monitoring analysis based on the state evaluation result to generate monitoring commands for real-time control and early warning of the busbar plug-in box production process. This solves the technical problems in existing technologies, such as the inability to comprehensively capture the multi-dimensional operating status of the busbar plug-in box, low data utilization efficiency, inaccurate state evaluation, and untimely early warning and control, leading to insufficient production efficiency and reliability. It achieves the technical effect of accurate monitoring and real-time early warning of the busbar plug-in box production process, thereby improving production efficiency and reliability. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 This is a schematic diagram of the production monitoring method for busbar plug-in boxes based on multi-sensor fusion provided in an embodiment of this application.

[0018] Figure 2 This is a schematic diagram of the production monitoring system for busbar plug-in boxes based on multi-sensor fusion, provided in an embodiment of this application.

[0019] Explanation of reference numerals in the attached diagram: 10 for operation parameter acquisition module, 20 for gain control module, 30 for status identification and evaluation module, and 40 for production monitoring and analysis module. Detailed Implementation

[0020] To further illustrate the technical means and effects adopted by the present invention in order to achieve the intended purpose, the following detailed description is provided in conjunction with the accompanying drawings and preferred embodiments, based on the specific implementation methods, structures, features and effects of the present invention.

[0021] This application provides a method for monitoring the production of busbar connector boxes based on multi-sensor fusion, such as... Figure 1 As shown, the method includes: Step S100: Collect multi-source operating parameters of the bus plug-in box in real time through multiple sensors to construct a heterogeneous sensor data set.

[0022] Step S100 further includes step S110, traversing the busbar plug-in boxes to perform production impact analysis, and determining multiple key locations based on the impact analysis results; step S120, constructing a time synchronization mechanism based on the data acquisition sequence, and deploying a sensor array based on the multiple key locations according to the time synchronization mechanism; step S130, performing real-time data acquisition on the busbar plug-in boxes through the sensor array to obtain multi-source operating parameters; step S140, extracting data timestamp markers according to the time synchronization mechanism, performing structural analysis on the multi-source operating parameters according to the data timestamp markers, and determining a preset data structure; step S150, encapsulating the multi-source operating parameters according to the preset data structure to construct the heterogeneous sensor data set.

[0023] Preferably, a systematic examination and analysis of the entire structure, circuitry, and operating principle of the busbar junction box is conducted to identify the state parameters at various locations that directly determine or significantly affect the final product quality, production safety, or process stability. For example, the contact resistance of the connection points, the current-carrying capacity of the busbar, and the weak points of the insulation components are identified as key influencing factors. The analysis results are then used to determine multiple key locations, such as each connection interface, specific segment points of the busbar, and support structure points. Then, based on the data acquisition sequence, a time synchronization mechanism is constructed using a synchronous clock signal in the hardware to ensure that each data point acquired by all sensors has a unified, high-precision time reference, thereby achieving time alignment of data from different sensors.

[0024] Preferably, multiple temperature and current sensors are deployed at key locations according to a time synchronization mechanism to form a sensor array. The acquisition cycle and timing of all sensors are uniformly coordinated and controlled by this time synchronization mechanism. The sensor array is then activated to collect real-time operational data from the busbar junction box, obtaining multi-source operating parameters, including temperature values ​​from temperature sensors, current values ​​from current sensors, voltage values ​​from voltage sensors, and acceleration or spectral data from vibration sensors. Based on the time synchronization mechanism, a timestamp is extracted from each acquired operating parameter. The multi-source operating parameters are then structurally analyzed according to these timestamps, binding data collected by all different sensors at the same time or within the same time window to form logical data units, i.e., a preset data structure. Finally, the timestamp-aligned multi-source operating parameters are packaged and encapsulated according to the preset data structure to construct a heterogeneous sensor data set that is strictly synchronized in time and uniformly standardized in structure.

[0025] Furthermore, step S130 also includes step S131, where the sensor array includes a temperature sensor, a current sensor, a voltage sensor, and a vibration sensor; step S132, where the temperature sensor performs contact sensing on the conductive connection points of the busbar junction box to obtain contact temperature data; step S133, where the current sensor performs current sensing on the busbar conductors of the busbar junction box to obtain load current data; step S134, where the voltage sensor performs voltage sensing on the phase line and ground line between the busbar junction boxes to obtain voltage data; step S135, where the vibration sensor performs vibration sensing on the mechanical structure of the busbar junction box to obtain vibration spectrum data; and step S136, where the contact temperature data, the load current data, the voltage data, and the vibration spectrum data are integrated to construct the multi-source operating parameters.

[0026] Preferably, the sensor array includes a temperature sensor, a current sensor, a voltage sensor, and a vibration sensor. Specifically, the temperature sensor performs contact sensing on the conductive connection points of the busbar plug-in box. The conductive connection point refers to the physical interface where the busbar contacts the pin in the plug-in box. A thermocouple or platinum resistance contact sensor is used to conduct heat through physical contact with the busbar or pin surface. The temperature sensor directly measures the temperature value of the connection point to obtain contact temperature data, which is used to monitor the overheating state of the conductive connection point and assess whether the contact resistance has increased or the connection has become loose. The current sensor monitors the busbar of the busbar plug-in box... The conductors are used for current sensing. The busbar conductor is the main busbar inside the plug-in box that carries the current. Hall effect current sensors or current transformers are used to measure the magnetic field around the conductor in a non-contact or through-core manner to determine the current value flowing through the busbar and obtain load current data. This data is used to monitor the electrical load, calculate power, assess current carrying capacity, and correlate with temperature data for analysis. Voltage sensors are used to sense the voltage between the phase wires and the ground wires between the busbar plug-in boxes. That is, a resistive voltage divider or voltage transformer is used to measure the potential difference between the phase wire and the ground wire to obtain voltage data. This data is used to monitor power supply quality, insulation status, and ground faults.

[0027] Preferably, vibration sensors are used to sense the vibration of the mechanical structure of the busbar junction box. The mechanical structure refers to the box body, brackets, internal support components, etc. An accelerometer is used to measure the vibration signal, and then the vibration signal in the time domain is converted to the frequency domain by a fast Fourier transform to obtain vibration spectrum data. This data is used to reveal the distribution of vibration energy at different frequencies to diagnose the mechanical condition. For example, an increase in vibration at a specific frequency may indicate loose screws, resonance of internal components, or busbar vibration caused by electrodynamics. Finally, the contact temperature data, load current data, voltage data, and vibration spectrum data are integrated on the basis of time synchronization to construct multi-source operating parameters, so as to provide multi-dimensional data that can comprehensively reflect the health status of the equipment.

[0028] Step S200: Adaptive gain control is performed on the heterogeneous sensing data set, and weighted fusion is performed based on the gain dataset to generate a standardized feature vector.

[0029] Step S200 further includes step S210, traversing the heterogeneous sensing data set to perform signal quality assessment on the sensor array respectively, and calculating the instantaneous signal strength of the sensor array; step S220, performing gain analysis based on the instantaneous signal strength, setting gain coefficients, and applying the gain coefficients to perform real-time gain control on the sensor array to generate a gain dataset; step S230, performing quality analysis based on the gain dataset, setting data quality indicators, assigning weights to the gain dataset according to the data quality indicators, and calculating multiple weight coefficients; step S240, performing weighted fusion calculation on the heterogeneous sensing data set based on the multiple weight coefficients to generate a primary fusion feature; and step S250, normalizing the primary fusion feature to construct the standardized feature vector.

[0030] Preferably, the heterogeneous sensor data set is traversed, and signal quality assessment is performed on the sensor arrays individually. For slowly varying signals such as temperature, current, and voltage, the instantaneous amplitude or average value within a very short time window is calculated. For dynamic signals such as vibration, the root mean square value or peak value within a very short time window is calculated to represent the instantaneous vibration energy, thereby determining the instantaneous signal strength of the sensor array, which is used to measure the amplitude or energy level of the signal at the current moment. Then, gain analysis is performed based on the instantaneous signal strength, that is, the instantaneous signal strength of each sensor is compared with the preset ideal amplitude range. According to the comparison result, the amplification or reduction factor for each sensor is dynamically calculated and set as the gain coefficient. If the signal is below the lower limit, a gain coefficient greater than 1 is set for amplification; if the signal is above the upper limit, a gain coefficient less than 1 is set for attenuation; if the signal is within the ideal amplitude range, the gain coefficient is 1. The calculated gain coefficient is then applied in real time to the raw data stream of the corresponding sensor for real-time gain control, that is, to unify the amplitude of all sensor signals to approximately the same dynamic range, avoiding some signals from being submerged by noise due to being too weak, or saturated and distorted due to being too strong, thereby determining the gain dataset.

[0031] Preferably, the adjusted gain dataset is subjected to quality analysis, and data quality indicators are set to quantify signal reliability. These may include signal-to-noise ratio (SNR), signal stability, and deviation from the expected range. SNR is the power ratio of the signal to the background noise; a higher SNR indicates better quality. Signal stability is the degree of signal fluctuation over a short period; smaller fluctuations indicate greater stability. Deviation from the expected range refers to whether the signal remains within a reasonable physical range even after gain adjustment. Then, based on the quality indicator calculation results, each sensor data is weighted and multiple weight coefficients are determined. Higher quality sensor data is assigned a larger weight coefficient, while lower quality data is assigned a smaller weight coefficient. Next, multiple operating parameters from different sensors are weighted and fused according to their respective weight coefficients to generate comprehensive primary fusion features, so as to rely on the information provided by high-quality sensors and suppress the interference of low-quality data. Finally, Min-Max normalization or Z-Score standardization is used to normalize the primary fusion features, that is, the data is scaled to the interval [0, 1] or [-1, 1] to eliminate the influence of different physical unit data dimensions. The final output is a standardized feature vector with uniform scale, dimensionless and optimized for quality and signal-to-noise ratio.

[0032] Further, step S220 also includes step S221, performing intensity extreme value analysis on the sensor array based on the heterogeneous sensing data set to determine the maximum and minimum signal strength values; step S222, setting a target intensity range by using the maximum signal strength value as the upper limit of the range and the minimum signal strength value as the lower limit of the range; step S223, comparing the instantaneous signal strength with the preset target intensity range; step S224, if the instantaneous signal strength is lower than the lower limit of the range, generating an increase command, performing gain analysis based on the increase command, and setting a first gain coefficient; step S225, amplifying the sensor array based on the first gain coefficient to generate a first gain dataset; step S226, if the instantaneous signal strength is higher than the upper limit of the range, generating a decrease command, performing gain analysis based on the decrease command, and setting a second gain coefficient; step S227, attenuating the sensor array based on the second gain coefficient to generate a second gain dataset.

[0033] Preferably, a heterogeneous sensor dataset is used to perform intensity extremum analysis on the sensor array to identify the maximum and minimum signal strength values ​​of each sensor within a given time period. This defines the original dynamic range of the sensor in actual operation. The maximum signal strength value is used as the upper limit of the interval, and the minimum signal strength value is used as the lower limit of the interval. A target intensity interval is set, representing the actual operating range of the sensor, and serves as the benchmark for automatic gain control. Then, the instantaneous signal strength of each sensor at the current moment is compared with the preset target intensity interval. If the instantaneous signal strength is less than the minimum signal strength value, it is determined that the current signal is too weak. An adjustment command is generated, and gain analysis is performed. An amplification factor greater than 1 is calculated as the first gain factor, ensuring that the amplified signal reaches the lower limit of the target intensity interval. The calculated first gain factor is then applied to the original signal of the sensor channel to amplify the sensor array, increasing the level of the weak signal and making it stand out from the background noise, thereby improving the signal-to-noise ratio and discernibility, and generating a first gain dataset. If the instantaneous signal strength is higher than the maximum signal strength, it is determined that the current signal is too strong and there is a risk of saturation distortion or exceeding the range. In this case, a reduction command is generated and gain analysis is performed to calculate an attenuation coefficient less than 1 as a second gain coefficient to pull the signal back into the target range. The calculated second gain coefficient is applied to the original signal of the sensor channel to attenuate the sensor array, prevent signal clipping or saturation, protect the subsequent acquisition and processing circuits, and ensure that the signal amplitude is within a reasonable range of the standardized processing. The second gain dataset is then determined.

[0034] Step S300: Perform state recognition and evaluation based on the standardized feature vector to obtain the state evaluation result.

[0035] Step S300 further includes step S310, performing deep confidence analysis based on the standardized feature vector, extracting features based on the confidence level, and determining deep feature parameters; step S320, classifying the busbar plug-in box into operating states to obtain multiple state categories, calculating the distribution of the multiple state categories based on the deep feature parameters, and obtaining multiple state category probability distribution parameters; step S330, performing multi-timescale analysis based on the multiple state category probability distribution parameters to obtain multiple timescale features, including short-term transient features, medium-term trend features, and long-term degradation features; step S340, performing a comprehensive evaluation based on the short-term transient features, the medium-term trend features, and the long-term degradation features to obtain multi-timescale state evaluation results for abnormal risk identification and classifying multiple state abnormal risk levels; step S350, adding the multiple state abnormal levels, the confidence level, and the multi-timescale state evaluation results to the state evaluation results.

[0036] Preferably, the standardized feature vectors are subjected to state recognition and evaluation, namely, operation state classification and anomaly risk assessment. Specifically, deep learning models such as Deep Belief Network (DBN) or stacked autoencoders are used to perform deep belief analysis on the standardized feature vectors. Through a multi-layer nonlinear network structure, the input features are combined and abstracted in a high order, automatically learning and discovering complex and nonlinear patterns in the data. At the same time, the reliability or significance of the learned abstract patterns is evaluated as confidence level, and feature extraction is performed based on the confidence level, including screening high-confidence and stable patterns and determining deep feature parameters to characterize the operation state features. Then, the operation state of the bus plug-in box is classified, that is, the operation status of the bus plug-in box is divided into multiple discrete state categories, such as normal, attention, abnormal and fault. The deep feature parameters are then input into a classifier connected to the Softmax layer at the end of the deep learning model to calculate the distribution of multiple state categories and obtain multiple state category probability distribution parameters, which are used to quantify the probability that the bus plug-in box currently belongs to each predefined state category.

[0037] Preferably, multi-timescale analysis is performed based on multiple state category probability distribution parameters. This involves using multiple continuous state category probability distribution parameters from the time series as input for time series analysis. Different time windows are used to analyze and extract three different types of time-scale features from the probabilistic time series data, including short-term transient features, medium-term trend features, and long-term degradation features. Short-term transient features analyze probability changes at the second / minute level to capture sudden events, such as a sudden increase in abnormal probability caused by a momentary current surge followed by recovery, reflecting instantaneous stability. Medium-term trend features use moving averages or linear regression to analyze probability change trends at the hour / day level to identify continuous, slow changes, such as the probability of the "watch" state steadily increasing over several hours, reflecting the gradual change in operating conditions. Long-term degradation features analyze the evolution of probability distribution at the week / month level to assess long-term performance degradation, such as the average probability of the "normal" state gradually decreasing over time, reflecting the aging and wear of the busbar junction box.

[0038] Preferably, a comprehensive assessment is conducted based on short-term transient characteristics, medium-term trend characteristics, and long-term degradation characteristics. This involves integrating assessment information from three different time scales to jointly judge transient, trend, and degradation characteristics. For example, even if the current short-term state is normal, if the medium-term trend shows continuous deterioration and long-term degradation is significant, the comprehensive assessment result also points to high risk. This leads to a multi-time-scale state assessment result and abnormal risk identification. Based on the comprehensive assessment result, the risk is quantified into multiple state abnormality risk levels, such as low risk, medium risk, and high risk. Finally, multiple state abnormality levels, confidence levels, and multi-time-scale state assessment results are integrated to obtain a structured state assessment result. This enables the monitoring of instantaneous state to predict health throughout the entire life cycle, allowing for earlier, more accurate, and more comprehensive identification of potential problems and quantification of risk levels, thereby ensuring precise regulation and predictive maintenance.

[0039] Step S400: Based on the status assessment results, perform production monitoring analysis and generate monitoring instructions to conduct real-time control and early warning of the production process of the busbar plug-in box.

[0040] Step S400 further includes step S410, constructing a production control strategy library, using the status assessment result as an index to search and match the production control strategy library, and determining a control scheme based on the matching result; step S420, performing production monitoring analysis on the busbar plug-in box according to the control scheme, and generating monitoring instructions; step S430, issuing the monitoring instructions to the production line execution unit to perform real-time control feedback on the busbar plug-in box, and generating a control feedback parameter set; step S440, performing control analysis based on the control feedback parameter set to obtain real-time control effect data; step S450, performing dynamic iterative optimization by backtracking to the monitoring instructions according to the real-time control effect data, generating monitoring optimization instructions to perform closed-loop control on the production process of the busbar plug-in box, and constructing a control log; step S460, performing production control anomaly analysis on the busbar plug-in box according to the control log, extracting control anomaly data for real-time early warning.

[0041] Preferably, a production control strategy library is constructed based on historical busbar plug-in box operation status control data records. This library stores the mapping relationship between various known status categories and corresponding control actions. For example, it includes multiple preset rules containing different status conditions and corresponding execution actions. Status assessment results are used as keyword indexes and input into the production control strategy library for retrieval and matching. The preset rule that best matches the current operating status is identified and determined as the matching result, and a control scheme is determined. For example, if the status is medium risk and the cause is an upward temperature trend, the control scheme is executed, reducing the load current by 15% and notifying the inspection team. Production monitoring analysis is performed on the busbar plug-in box according to the control scheme. This involves converting the control scheme into operation instructions with specific parameters that the production line can understand and execute, determining the monitoring instructions. These monitoring instructions are then sent to the production line execution unit via PLC and industrial bus. The production line execution unit may be a programmable logic controller (PLC) for controlling circuit breakers, regulating valves, etc., a robot control unit for performing mechanical adjustments, or a manufacturing execution management unit (MEP) for generating work orders and triggering alarms. The production line execution unit of the busbar plug-in box receives the monitoring instructions and provides real-time control feedback, returning a confirmation signal or status report, and determining the control feedback parameter set, which includes the execution status of the monitoring instructions.

[0042] Preferably, control analysis is performed based on the control feedback parameter set. This involves monitoring the actual changes in the production of the busbar connection box after the execution of the monitoring command, correlating the control feedback parameter set with the re-acquired state assessment results, and generating real-time control effect data to quantify the effectiveness of the control actions. For example, if the load current reduction command has been successfully executed and the busbar connection point temperature drops by 8°C within the following 5 minutes, the trend changes from rising to stabilizing. Then, the real-time control effect data is used to backtrack to the monitoring command for dynamic iterative optimization. This involves correlating the control effect data with the monitoring command to form a command-effect causal chain, and optimizing decisions based on this causal chain to generate monitoring optimization commands. The production process of the busbar connection box is then subject to closed-loop control. If the effect is good, the applicability of the command in this state is strengthened; if the effect is poor or ineffective, the control scheme is adjusted or a new, parameter-optimized monitoring optimization command is generated and reissued.

[0043] Preferably, the original state assessment results, issued instructions, feedback parameters, effect data, and optimization process throughout the entire control lifecycle are recorded in a complete time series format to form a control log, which is used for tracing, auditing, and analyzing the complete cycle of each control event. Finally, based on the control log, anomaly analysis of bus plug-in box production control is performed, that is, monitoring whether the control behavior of the bus plug-in box production itself is normal, analyzing the control log to identify and determine abnormal patterns in the control process, such as the execution unit returning an instruction execution failure status, instruction execution response timeout, or the equipment status deteriorating after control, etc. Finally, control anomaly data is extracted, fault information about the control itself is generated, and real-time early warnings are issued. For example, an early warning is issued that the execution unit PLC_02 does not respond to the current adjustment instruction, please check the communication link.

[0044] Furthermore, step S410 also includes step S411, retrieving the historical production control dataset of the busbar plug-in box for state analysis, dividing it into multiple control state data for strategy matching, and determining the normal state strategy set, the early warning state strategy set, and the alarm state strategy set; step S412, integrating the normal state strategy set, the early warning state strategy set, and the alarm state strategy set to construct a three-level strategy library structure; step S413, mapping the historical production control dataset to the three-level strategy library structure to construct a production control strategy library; step S414, calculating the state features of the state assessment results based on the multiple state anomaly levels, the confidence level, and the multi-timescale state assessment results, determining the feature contribution, and constructing a multi-level index tree based on the feature contribution; step S415, traversing the production control strategy library based on the multi-level index tree to perform fuzzy matching, obtaining the scheme matching degree, and determining the control scheme based on the scheme matching degree.

[0045] Preferably, the historical production control dataset of the busbar plug-in box is retrieved, which includes past status assessment results and their corresponding, verified, and effective control schemes. Status analysis is performed on the historical production control dataset, including clustering or classification analysis of the status assessment results in the historical data, dividing it into multiple control status data, and then classifying all historical control cases into normal status strategy sets, early warning status strategy sets, and alarm status strategy sets according to their severity. The normal status strategy set is applicable to maintenance or preventive strategies taken when the status assessment is "normal" or "low risk", such as "keeping the current parameters running" and "recording operating data". The early warning status strategy set is applicable to intervention strategies taken when the status assessment is "medium risk" or "concerned", such as "issuing early warning notices", "suggesting load reduction", and "arranging routine inspections". The alarm status strategy set is applicable to emergency response strategies taken when the status assessment is "high risk" or "abnormal", such as "immediately disconnecting the circuit", "initiating emergency shutdown procedures", and "issuing the highest level alarm".

[0046] Preferably, the normal state strategy set, the early warning state strategy set, and the alarm state strategy set are integrated. That is, the strategy library is divided into three clear levels—normal, early warning, and alarm—according to the severity level of the state, constructing a hierarchical and prioritized knowledge base. This three-level strategy library structure ensures that during retrieval, the most relevant broad category to the current state severity can be located first, improving retrieval efficiency and accuracy. Then, each historical case in the historical production control dataset is mapped to the three-level strategy library structure according to its state category, constructing a production control strategy library containing multiple preset rules or cases populated by historical experience data. Based on multiple state anomaly levels, confidence levels, and multi-timescale state assessment results, the state characteristics of the state assessment results are calculated. This involves analyzing each dimension constituting the current state assessment result and evaluating the importance or feature contribution of each dimension to the final decision using decision trees or information gain algorithms. For example, "state anomaly level" has the highest contribution, followed by "long-term degradation features."

[0047] Preferably, a multi-level index tree is constructed based on the feature contribution. The first-level index / root node consists of the feature with the highest contribution, "state anomaly level". The second-level index / branch node consists of the second most important feature, "main cause". The third-level index / leaf node consists of a more detailed feature "confidence level" range or specific trend value. The multi-level index tree organizes the cases in the production control strategy library according to the same feature dimension. Then, using the current state assessment result, fuzzy matching is performed by traversing the production control strategy library based on a multi-level index tree. That is, the production control strategy library is searched along the multi-level index tree from the root node to the leaf node. Fuzzy matching is used to evaluate the similarity between the current state assessment result and the states of each historical case in the strategy library, such as calculating Euclidean distance or cosine similarity, to determine the scheme matching degree, which represents the percentage of matching between each historical case and the current situation. Finally, a judgment is made based on the scheme matching degree. That is, a matching degree threshold of 0.9 is set. If the matching degree of a scheme is higher than the matching degree threshold, the control scheme of that case is directly adopted. If multiple schemes have high matching degrees but none exceed the threshold, the corresponding cases are weighted and merged to generate a comprehensive update scheme. If the matching degrees of all schemes are very low, a safety policy may be triggered or manual intervention may be requested. Finally, a control scheme based on historical experience is output.

[0048] Furthermore, step S460 also includes step S461, performing time-series analysis on the control log to obtain multiple time-series indicator features, including instruction issuance timestamp, execution start timestamp, and execution completion timestamp; step S462, performing control response calculation based on the instruction issuance timestamp, execution start timestamp, and execution completion timestamp to obtain control response time parameters; step S463, performing execution calculation based on the control response time parameters according to parameter stability indicators to obtain execution deviation; step S464, identifying production control anomalies in the busbar plug-in box according to the execution deviation, and determining control anomaly data.

[0049] Preferably, the control log is read, and time-related fields are extracted from each log record and time-series parsing is performed to determine time data points with time sequence, obtaining multiple time-series indicator features, including instruction issuance timestamp, execution start timestamp, and execution completion timestamp. These represent the precise time when the command is issued to the production line execution unit, the precise time when the production line execution unit receives the instruction and drives the actuator to start execution, and the precise time when the actuator completes the instruction and sends back a success signal. Then, the control response is calculated based on the instruction issuance timestamp, execution start timestamp, and execution completion timestamp, i.e., the time difference is calculated for quantifying production. The delays at each stage of the control process are specifically calculated as follows: The difference between the execution start timestamp and the instruction issuance timestamp is used to determine the communication delay, which measures the network and interface delays required for the instruction to reach the execution unit and begin processing; the difference between the execution completion timestamp and the execution start timestamp is used to determine the actuator action time, which measures the time required for actuators to complete physical actions such as circuit breaker tripping and motor rotation; and the difference between the execution completion timestamp and the instruction issuance timestamp is used to determine the total response time. The communication delay, actuator action time, and total response time are then used as control response time parameters to measure the execution efficiency of production control.

[0050] Preferably, the calculation is performed based on the control response time parameter according to the parameter stability index. This involves analyzing the stability and consistency of the control response time parameter across multiple controls over a period of time. The parameter stability index is calculated based on historical data and is used to assess the fluctuation of the time parameter, such as historical average and standard deviation, moving average, and fluctuation range. Then, the current control response time parameter is compared with the parameter stability index to calculate... The deviation indicates the degree to which the current response time deviates from the historical normal level. A larger positive deviation indicates that the current control response is significantly slower than the historical normal response. Then, the production control anomaly of the bus plug-in box is marked according to the execution deviation. That is, an execution deviation threshold is set according to production requirements. When the execution deviation exceeds the threshold, it is determined that the execution process of the current production control has an anomaly, the anomaly of the production control execution link is determined, and the control anomaly data is output. This may clearly identify the delayed control command and control link, as well as the severity of the anomaly determined based on the execution deviation, thereby realizing accurate monitoring and real-time early warning of the bus plug-in box production process and improving production efficiency and reliability.

[0051] In the above text, refer to Figure 1 This paper describes in detail a production monitoring method for busbar junction boxes based on multi-sensor fusion according to embodiments of the present invention. Next, reference will be made to... Figure 2 This invention describes a busbar junction box production monitoring system based on multi-sensor fusion according to an embodiment of the present invention.

[0052] The busbar plug-in box production monitoring system based on multi-sensor fusion according to embodiments of the present invention addresses the technical problems in existing technologies, such as the inability to comprehensively capture the multi-dimensional operating status of busbar plug-in boxes, low data utilization efficiency, inaccurate status assessment, and untimely early warning and control, leading to insufficient production efficiency and reliability. It achieves the technical effect of realizing accurate monitoring and real-time early warning of the busbar plug-in box production process, thereby improving production efficiency and reliability. Figure 2 As shown, the busbar junction box production monitoring system based on multi-sensor fusion includes: an operating parameter acquisition module 10, a gain control module 20, a status identification and evaluation module 30, and a production monitoring and analysis module 40.

[0053] The operating parameter acquisition module 10 is used to acquire multi-source operating parameters of the busbar plug-in box in real time through multiple sensors and construct a heterogeneous sensor data set; the gain control module 20 is used to perform adaptive gain control on the heterogeneous sensor data set, perform weighted fusion based on the gain dataset, and generate a standardized feature vector; the state recognition and evaluation module 30 is used to perform state recognition and evaluation based on the standardized feature vector and obtain the state evaluation result; the production monitoring and analysis module 40 is used to perform production monitoring and analysis based on the state evaluation result, generate monitoring instructions to perform real-time regulation and early warning of the production process of the busbar plug-in box.

[0054] The specific configuration of the operation parameter acquisition module 10 will be described in detail below. The operation parameter acquisition module 10 further includes: traversing the busbar plug-in boxes to perform production impact analysis, and determining multiple key locations based on the impact analysis results; constructing a time synchronization mechanism based on the data acquisition sequence, and deploying a sensor array based on the multiple key locations according to the time synchronization mechanism; performing real-time operation acquisition of the busbar plug-in boxes through the sensor array to obtain multi-source operation parameters; extracting data timestamp markers according to the time synchronization mechanism, performing structural analysis on the multi-source operation parameters according to the data timestamp markers, and determining a preset data structure; and encapsulating the multi-source operation parameters according to the preset data structure to construct the heterogeneous sensor data set.

[0055] The specific configuration of the operating parameter acquisition module 10 will be described in detail below. The operating parameter acquisition module 10 further includes: a sensor array comprising a temperature sensor, a current sensor, a voltage sensor, and a vibration sensor; contact temperature data is obtained by sensing the conductive connection points of the busbar junction box using the temperature sensor; load current data is obtained by sensing the current of the busbar conductors of the busbar junction box using the current sensor; voltage data is obtained by sensing the voltage between the phase wires and ground wires between the busbar junction boxes using the voltage sensor; vibration spectrum data is obtained by sensing the vibration of the mechanical structure of the busbar junction box using the vibration sensor; and the contact temperature data, load current data, voltage data, and vibration spectrum data are integrated to construct the multi-source operating parameters.

[0056] The specific configuration of the gain control module 20 will be described in detail below. The gain control module 20 further includes: traversing the heterogeneous sensing data set to perform signal quality assessment on the sensor array respectively, and calculating the instantaneous signal strength of the sensor array; performing gain analysis based on the instantaneous signal strength, setting gain coefficients, and applying the gain coefficients to perform real-time gain control on the sensor array to generate a gain dataset; performing quality analysis based on the gain dataset, setting data quality indicators, and assigning weights to the gain dataset according to the data quality indicators to calculate multiple weight coefficients; performing weighted fusion calculation on the heterogeneous sensing data set based on the multiple weight coefficients to generate primary fusion features; and normalizing the primary fusion features to construct the standardized feature vector.

[0057] The specific configuration of the gain control module 20 will be described in detail below. The gain control module 20 further includes: performing intensity extreme value analysis on the sensor array based on the heterogeneous sensing data set to determine the maximum and minimum signal strength values; setting a target intensity range by using the maximum signal strength value as the upper limit of the range and the minimum signal strength value as the lower limit of the range; comparing the instantaneous signal strength with the preset target intensity range; if the instantaneous signal strength is lower than the lower limit of the range, generating an increase command, performing gain analysis based on the increase command, and setting a first gain coefficient; amplifying the sensor array based on the first gain coefficient to generate a first gain dataset; if the instantaneous signal strength is higher than the upper limit of the range, generating a decrease command, performing gain analysis based on the decrease command, and setting a second gain coefficient; and attenuating the sensor array based on the second gain coefficient to generate a second gain dataset.

[0058] The specific configuration of the state identification and assessment module 30 will be described in detail below. The state identification and assessment module 30 further includes: performing deep confidence analysis based on the standardized feature vector, extracting features based on the confidence level, and determining deep feature parameters; classifying the busbar plug-in box into operating states to obtain multiple state categories, calculating the distribution of the multiple state categories based on the deep feature parameters, and obtaining multiple state category probability distribution parameters; performing multi-timescale analysis based on the multiple state category probability distribution parameters to obtain multiple timescale features, including short-term transient features, medium-term trend features, and long-term degradation features; conducting a comprehensive assessment based on the short-term transient features, the medium-term trend features, and the long-term degradation features to obtain multi-timescale state assessment results for abnormal risk identification, and classifying multiple state abnormal risk levels; and adding the multiple state abnormal levels, the confidence level, and the multi-timescale state assessment results to the state assessment results.

[0059] The specific configuration of the production monitoring and analysis module 40 will be described in detail below. The production monitoring and analysis module 40 further includes: constructing a production control strategy library; using the status evaluation results as an index to search and match the production control strategy library; determining a control scheme based on the matching results; performing production monitoring and analysis on the busbar plug-in box according to the control scheme; generating monitoring instructions; issuing the monitoring instructions to the production line execution unit to provide real-time control feedback on the busbar plug-in box, generating a control feedback parameter set; performing control analysis based on the control feedback parameter set to obtain real-time control effect data; dynamically iterating and optimizing the monitoring instructions according to the real-time control effect data, generating monitoring optimization instructions to perform closed-loop control of the busbar plug-in box production process, and constructing a control log; performing production control anomaly analysis on the busbar plug-in box based on the control log, extracting control anomaly data for real-time early warning.

[0060] The specific configuration of the production monitoring and analysis module 40 will be described in detail below. The production monitoring and analysis module 40 further includes: retrieving historical production control datasets from the busbar plug-in box for status analysis; dividing the control status data into multiple sets for strategy matching to determine normal status strategy sets, early warning status strategy sets, and alarm status strategy sets; integrating the normal status strategy sets, early warning status strategy sets, and alarm status strategy sets to construct a three-level strategy library structure; mapping the historical production control datasets to the three-level strategy library structure to construct a production control strategy library; calculating the status features of the status assessment results based on the multiple status anomaly levels, the confidence level, and the multi-timescale status assessment results to determine the feature contribution; constructing a multi-level index tree based on the feature contribution; traversing the production control strategy library based on the multi-level index tree to perform fuzzy matching to obtain the scheme matching degree; and generating the control scheme based on the scheme matching degree.

[0061] The specific configuration of the production monitoring and analysis module 40 will be described in detail below. The production monitoring and analysis module 40 further includes: performing time-series analysis on the control log to obtain multiple time-series indicator features, including instruction issuance timestamp, execution start timestamp, and execution completion timestamp; performing control response calculation based on the instruction issuance timestamp, execution start timestamp, and execution completion timestamp to obtain control response time parameters; performing execution calculation based on the control response time parameters according to parameter stability indicators to obtain execution deviation; and identifying production control anomalies in the busbar plug-in box according to the execution deviation to determine abnormal control data.

[0062] The busbar plug-in box production monitoring system based on multi-sensor fusion provided in this embodiment of the invention can execute the busbar plug-in box production monitoring method based on multi-sensor fusion provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0063] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for production monitoring of busbar junction boxes based on multi-sensor fusion, characterized in that, The method includes: By collecting multi-source operating parameters of the bus plug-in box in real time through multiple sensors, a heterogeneous sensor data set is constructed. Adaptive gain control is applied to the heterogeneous sensing data set, and weighted fusion is performed based on the gain dataset to generate a standardized feature vector; State recognition and evaluation are performed based on the standardized feature vectors to obtain state evaluation results; Based on the status assessment results, production monitoring analysis is performed to generate monitoring instructions for real-time control and early warning of the busbar plug-in box production process.

2. The busbar connector box production monitoring method based on multi-sensor fusion as described in claim 1, characterized in that, The method involves real-time acquisition of multi-source operating parameters from the busbar junction box using multiple sensors to construct a heterogeneous sensor dataset. A production impact analysis was conducted by traversing all busbar junction boxes, and several key locations were identified based on the impact analysis results. A time synchronization mechanism is constructed based on the data acquisition sequence, and a sensor array is deployed according to the time synchronization mechanism based on the multiple key locations. The sensor array is used to collect real-time operational data of the busbar junction box to obtain multi-source operating parameters. Data timestamps are extracted based on the time synchronization mechanism, and the multi-source operating parameters are structurally analyzed according to the data timestamps to determine the preset data structure. The multi-source operating parameters are encapsulated according to the preset data structure to construct the heterogeneous sensing data set.

3. The busbar connector box production monitoring method based on multi-sensor fusion as described in claim 2, characterized in that, The method involves real-time data acquisition of multi-source operating parameters from the busbar junction box using the sensor array, and includes: The sensor array includes a temperature sensor, a current sensor, a voltage sensor, and a vibration sensor; The temperature sensor is used to sense the contact temperature of the conductive connection points of the busbar junction box. The current sensor is used to sense the current in the bus conductor of the bus junction box to obtain load current data. The voltage sensor is used to sense the voltage between the phase wire and the ground wire between the busbar junction boxes to obtain voltage data. The vibration sensor is used to sense the vibration of the mechanical structure of the busbar junction box and obtain vibration spectrum data. The contact temperature data, load current data, voltage data, and vibration spectrum data are integrated to construct the multi-source operating parameters.

4. The busbar connector box production monitoring method based on multi-sensor fusion as described in claim 2, characterized in that, Adaptive gain control is applied to the heterogeneous sensing data set, and weighted fusion is performed based on the gain dataset to generate a standardized feature vector. The method includes: The signal quality of the sensor array is evaluated by traversing the heterogeneous sensing data set, and the instantaneous signal strength of the sensor array is calculated. Gain analysis is performed based on the instantaneous signal strength, a gain coefficient is set, and the gain coefficient is used to perform real-time gain control on the sensor array to generate a gain dataset. Based on the gain dataset, a quality analysis is performed, data quality indicators are set, weights are allocated to the gain dataset according to the data quality indicators, and multiple weight coefficients are calculated. Based on the multiple weighting coefficients, a weighted fusion calculation is performed on the heterogeneous sensor data set to generate a primary fusion feature; The primary fusion features are normalized to construct the standardized feature vector.

5. The busbar connector box production monitoring method based on multi-sensor fusion as described in claim 4, characterized in that, Gain analysis is performed based on the instantaneous signal strength, a gain coefficient is set, and the gain coefficient is used to perform real-time gain control on the sensor array to generate a gain dataset. The method includes: Based on the heterogeneous sensing data set, intensity extremum analysis is performed on the sensor array to determine the maximum and minimum signal strength values. The target strength range is defined by taking the maximum value of the signal strength as the upper limit of the range and the minimum value of the signal strength as the lower limit of the range. The instantaneous signal strength is compared with a preset target strength range; If the instantaneous signal strength is lower than the lower limit of the interval, an increase command is generated, and gain analysis is performed based on the increase command to set a first gain coefficient. The sensor array is amplified based on the first gain coefficient to generate a first gain dataset. If the instantaneous signal strength is higher than the upper limit of the interval, a reduction instruction is generated, and gain analysis is performed based on the reduction instruction to set a second gain coefficient. The sensor array is attenuated based on the second gain coefficient to generate a second gain dataset.

6. The busbar connector box production monitoring method based on multi-sensor fusion as described in claim 1, characterized in that, The state identification and evaluation are performed based on the standardized feature vector to obtain the state evaluation result. The method includes: Deep confidence analysis is performed based on the standardized feature vectors, and feature extraction is performed based on the confidence level to determine the deep feature parameters. The busbar plug-in box is classified into multiple status categories based on its operating status. The distribution of these multiple status categories is calculated based on the deep feature parameters to obtain the probability distribution parameters of the multiple status categories. Multi-timescale analysis is performed based on the probability distribution parameters of the multiple state categories to obtain multiple timescale features, which include short-term transient features, medium-term trend features, and long-term degradation features. Based on the short-term transient characteristics, the medium-term trend characteristics, and the long-term degradation characteristics, a comprehensive evaluation is conducted to obtain multi-timescale state evaluation results for anomaly risk identification, and multiple state anomaly risk levels are classified. The multiple state anomaly levels, the confidence level, and the multi-timescale state assessment results are added to the state assessment results.

7. The busbar connector box production monitoring method based on multi-sensor fusion as described in claim 6, characterized in that, Based on the status assessment results, production monitoring analysis is performed to generate monitoring commands for real-time control and early warning of the busbar connector box production process. The method includes: A production control strategy library is constructed, and the status assessment results are used as an index to search and match the production control strategy library. The control scheme is determined based on the matching results. Based on the aforementioned control scheme, the production monitoring and analysis of the busbar plug-in box is performed to generate monitoring instructions. The monitoring command is sent to the production line execution unit to perform real-time control and feedback on the bus plug-in box, and a control and feedback parameter set is generated. Based on the aforementioned set of control feedback parameters, control analysis is performed to obtain real-time control effect data. Based on the real-time control effect data, the monitoring instructions are traced back to perform dynamic iterative optimization, generating monitoring optimization instructions to perform closed-loop control of the busbar plug-in box production process, and constructing a control log. Based on the aforementioned control logs, analyze any production control anomalies in the busbar connector box and extract the anomaly data for real-time early warning.

8. The busbar connector box production monitoring method based on multi-sensor fusion as described in claim 7, characterized in that, A production control strategy library is constructed, and the state assessment results are used as an index to search and match the production control strategy library. Control schemes are determined based on the matching results. The method includes: Retrieve historical production control datasets from busbar plug-in boxes for status analysis, divide multiple control status data for strategy matching, and determine normal status strategy sets, early warning status strategy sets, and alarm status strategy sets. Based on the normal state policy set, the early warning state policy set, and the alarm state policy set, a three-level policy library structure is constructed. The historical production control dataset is mapped to the three-level strategy library structure to construct the production control strategy library; Based on the multiple state anomaly levels, the confidence level, and the multi-timescale state assessment results, the state features of the state assessment results are calculated, the feature contribution is determined, and a multi-level index tree is constructed based on the feature contribution. Based on the multi-level index tree, the production control strategy library is traversed to perform fuzzy matching to obtain the scheme matching degree. The control scheme is then generated based on the scheme matching degree.

9. The busbar connector box production monitoring method based on multi-sensor fusion as described in claim 7, characterized in that, Based on the aforementioned control log, perform anomaly analysis on the production control of the busbar connector box, extract anomaly data for real-time early warning, and include the following methods: The control log is analyzed by time series to obtain multiple time series indicator features, which include instruction issuance timestamp, execution start timestamp, and execution completion timestamp. The control response is calculated based on the instruction issuance timestamp, the execution start timestamp, and the execution completion timestamp to obtain the control response time parameter; Based on the aforementioned control response time parameter, the execution deviation is calculated according to the parameter stability index to obtain the execution deviation. The abnormal control data is determined by identifying the abnormal control data based on the aforementioned execution deviation degree to mark the abnormal control of the busbar plug-in box production control.

10. A busbar connector box production monitoring system based on multi-sensor fusion, characterized in that, The system is used to implement the busbar plug-in box production monitoring method based on multi-sensor fusion as described in any one of claims 1 to 9, and the system includes: The operation parameter acquisition module is used to collect multi-source operation parameters of the bus plug-in box in real time through multiple sensors and construct a heterogeneous sensor data set. The gain control module is used to perform adaptive gain control on the heterogeneous sensing data set, and to perform weighted fusion based on the gain dataset to generate a standardized feature vector. The state recognition and evaluation module is used to perform state recognition and evaluation based on the standardized feature vector to obtain the state evaluation result. The production monitoring and analysis module is used to perform production monitoring and analysis based on the status assessment results, and generate monitoring instructions to conduct real-time control and early warning of the production process of the busbar plug-in box.

Images