Laminating machine remote parameter debugging and fault diagnosis system based on edge calculation

Abstract

The invention relates to the technical field of industrial equipment internet of things systems, and discloses an edge computing laminating machine remote parameter debugging and fault diagnosis system which can trace process root causes from equipment abnormities or look forward equipment risks during process optimization by synchronously collecting equipment and process data and utilizing a causal network to perform bidirectional reasoning so as to improve the efficiency of process optimization. Therefore, a collaborative strategy of deep integration of the equipment and the process is generated, transformation from passive maintenance to active collaborative intervention is realized, and the comprehensive efficiency and reliability of a production system are greatly improved. The system can automatically adapt to changes of a production environment through a closed-loop self-learning mechanism based on feedback data, and long-term accuracy is kept. And in combination with cloud edge collaboration, group knowledge can be quickly fused, and deployment and popularization of intelligent applications are accelerated. The remarkable comprehensive benefits of reducing non-planned shutdown, improving product quality and equipment reliability, reducing operation and maintenance cost and expert dependence and supporting intelligent scale deployment are achieved.

Description

Technical Field

[0001] This invention relates to the field of industrial equipment Internet of Things (IoT) systems, and in particular to a remote parameter adjustment and fault diagnosis system for edge computing bonding machines. Background Technology

[0002] In the coating and lamination production of industrial textile fabrics and films, ensuring the stability of the lamination process and the achievement of product quality standards is crucial. As the core production equipment, the laminating machine's operating status directly affects the coating uniformity, lamination strength, and yield. During the operation of the laminating machine, a large amount of process data is generated, including the equipment's own operating parameters, such as coating head pressure, substrate traction speed, heating roller temperature, and adhesive coating amount, as well as production environment and material condition data, such as workshop ambient temperature and humidity, substrate type and batch characteristics, and adhesive viscosity.

[0003] Currently, intelligent methods have been applied to the monitoring and diagnosis of laminating machines. For example, by deploying an edge computing gateway, operating parameters such as equipment vibration, temperature, and power can be collected in real time. The data can be filtered, threshold compared, and preliminary anomaly judgment can be performed locally. At the same time, the data can be uploaded to the cloud platform, historical data can be used to train a fault diagnosis model, and the optimized model can be sent to the edge device for execution. This enables real-time diagnosis and early warning of typical equipment faults such as bearing damage and motor overload.

[0004] The aforementioned and existing related technologies often have the following drawbacks: In real-world scenarios with long-term operation, the system cannot achieve closed-loop collaborative optimization between equipment health status and real-time process quality requirements. This means that the system cannot accurately trace the process cause when the equipment alarms, nor can it proactively avoid equipment risks when optimizing the process, making it difficult to eradicate intermittent and complex production problems. Summary of the Invention

[0005] The technical problem to be solved by this invention is that existing technologies have the drawback of being unable to achieve closed-loop collaborative optimization between equipment health status and real-time process quality requirements. To address this, we propose an edge computing-based remote parameter debugging and fault diagnosis system for bonding machines.

[0006] To achieve the above objectives, this application adopts the following technical solution: a remote parameter debugging and fault diagnosis system for industrial equipment based on edge computing. The system is deployed at the production site of the target industrial equipment (e.g., the workshop where precision manufacturing equipment such as laminating machines, coating machines, and extruders are located) and is communicatively connected to a remote cloud platform. The system includes: The equipment operation sensing module is used to collect multi-dimensional operating status data of the target industrial equipment in real time. The multi-dimensional operating status data includes pressure data of core actuators (e.g., coating head pressure of laminating machine, screw pressure of extruder), material transmission speed data (e.g. substrate traction speed of laminating machine, conveyor belt speed of production line), key process temperature data (e.g. heating roller temperature of laminating machine, barrel temperature of injection molding machine), and vibration spectrum data of rotating or reciprocating motion parts (e.g., coating roller vibration of laminating machine, vibration of machine tool spindle). The process status sensing module is used to collect process condition data and online quality data of the target production process in real time. The process condition data includes production environment temperature and humidity data (e.g., ambient temperature and humidity of the workshop), material batch identification data (e.g., QR code of substrate roll, raw material batch number), and process medium characteristic data (e.g., real-time viscosity of adhesive in laminating machine, melt flow rate of injection molding machine). The online quality data includes product surface or coating quality image data obtained by machine vision unit (e.g., coating uniformity image of laminating machine, appearance defect detection image of workpiece). The equipment health diagnosis module is connected to the equipment operation perception module and is used to execute equipment-side abnormal diagnosis rules based on the multi-dimensional operation status data to generate equipment health index and preliminary fault warning information. The process quality assessment module, connected to the process status sensing module, is used to execute process-side deviation assessment rules based on the process condition data and the online quality data, and generate process quality scores and process parameter adjustment suggestions. The collaborative decision-making central module is connected to the equipment health diagnosis module and the process quality assessment module respectively. It is used to construct and apply a two-way causal relationship network between process parameters and equipment status, receive the equipment health index, the preliminary fault warning information, the process quality score and the process parameter adjustment suggestions, perform collaborative analysis based on causal relationship, and generate collaborative debugging instructions or collaborative maintenance strategies. The closed-loop optimization module connects the collaborative decision-making center module, the equipment operation perception module, and the process status perception module. It is used to execute the collaborative debugging command and iteratively optimize the parameters of the equipment-side anomaly diagnosis rule, the process-side deviation evaluation rule, and the causal relationship network based on the multi-dimensional feedback data after the command execution. The edge communication gateway module connects the collaborative decision-making hub module, the closed-loop optimization module, and the remote cloud platform, and is used to realize data upload and receive model / rule update instructions between the system and the remote cloud platform.

[0007] The technical effects and advantages of this invention are as follows: In this invention, high-dimensional equipment status data and process quality data are simultaneously collected through the equipment operation sensing module and the process status sensing module, laying a data foundation for in-depth analysis. The equipment health diagnosis module and the process quality assessment module operate in parallel, independently generating equipment health indices, fault warnings, process quality scores, and parameter adjustment suggestions, achieving accurate and independent assessment of equipment and process status. The core innovation of the system lies in the collaborative decision-making central module, which constructs and dynamically maintains a bidirectional causal relationship network between process and equipment. When an abnormal or optimization signal is received in any dimension, it can perform reverse root cause tracing or forward risk prediction along the network, thereby generating a collaborative strategy integrating equipment maintenance and process adjustment, or optimization instructions with attached equipment safety constraints, truly breaking down the decision-making silos between equipment and process. The closed-loop optimization module executes the decision instructions and, based on multi-dimensional feedback data, automatically iteratively adjusts the diagnostic rules, evaluation models, and causal network parameters using built-in optimization algorithms, enabling the system to continuously evolve. Through the synergistic effect of the above technical solutions, the following comprehensive effects are achieved: Elevating fault diagnosis from reactive maintenance to proactive early warning and root cause collaborative handling significantly reduces unplanned downtime and effectively prevents hidden equipment damage induced by processes. By transforming process parameter tuning from single-point trial and error to global collaborative optimization within the safety boundaries of equipment health status, the reliability of long-term equipment operation is ensured while improving product quality and production efficiency. The system has strong self-learning and self-adaptive capabilities, and can automatically adjust itself to follow dynamic factors such as equipment aging and material batch changes, maintaining high precision over a long period of time, reducing the continuous dependence on external expert experience and the system's own maintenance costs. Fourth, relying on the cloud-edge collaborative architecture, a single device can quickly absorb the collective wisdom of similar devices, accelerate model maturity, and realize the rapid replication and promotion of diagnostic and optimization knowledge, which greatly improves the efficiency of large-scale deployment and the overall intelligence level of the production system. Attached Figure Description

[0008] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts: Figure 1 This is a flowchart illustrating the overall collaborative workflow of the system according to the present invention; Figure 2 This is a flowchart of the collaborative decision-making causal reasoning process of the present invention; Figure 3 This is a schematic diagram of the closed-loop self-optimization principle of the present invention. Detailed Implementation

[0009] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0010] In real-world scenarios with long-term operation, the system cannot achieve closed-loop collaborative optimization between equipment health status and real-time process quality requirements. Specifically, when analyzing abnormalities such as coating head pressure, the diagnostic logic on the edge side only tracks the sensor data of the hydraulic system itself in isolation, completely severing the process status information such as substrate tension and adhesive temperature collected at the same time. It ignores the fact that fluctuations in process conditions are often the root cause of abnormal equipment parameters. Conversely, when the process system needs to adjust the heating temperature setting due to substrate replacement, the parameter adjustment command on the edge side is unidirectional and open-loop, failing to consider the cumulative fatigue state of the current heater elements. This disconnect between diagnosis and adjustment makes it impossible for the system to accurately trace the process cause when the equipment alarms, nor can it proactively avoid equipment risks when optimizing the process, making it difficult to eradicate intermittent and complex production problems.

[0011] To solve the above problems, refer to Figure 1 As shown, the present invention provides a technical solution: a remote parameter debugging and fault diagnosis system for edge computing bonding machines, specifically including the following: In this embodiment, a fabric or film lamination production line of a certain electromechanical equipment company is used as an application scenario. The company's laminating machine is used to precisely laminate substrates and films of different materials using adhesives. The production environment has extremely high requirements for temperature, humidity, and process stability. The system is deployed at the laminating machine production site and connects to the laminating machine controller, various sensors, and actuators via industrial Ethernet and fieldbus, forming an edge computing node. It also maintains communication with a remote cloud platform deployed in the company's data center through an edge communication gateway module, forming a cloud-edge collaborative architecture.

[0012] The system comprises an equipment operation sensing module, a process status sensing module, an equipment health diagnosis module, a process quality assessment module, a collaborative decision-making center module, a closed-loop optimization module, an edge communication gateway module, and a human-machine interface module. These modules work collaboratively to form a complete intelligent system for sensing, diagnosis, decision-making, and optimization, enabling real-time monitoring of the bonding machine's operating status, online assessment of process quality, fault diagnosis and parameter adjustment for equipment and process collaboration, and continuous optimization based on feedback.

[0013] The equipment operation sensing module is used to collect multi-dimensional operating status data of the laminating machine body in real time. In this fabric or film lamination scenario, the multi-dimensional operating status data specifically includes: coating head pressure data, which is collected by a high-precision pressure sensor installed in the hydraulic circuit of the coating head pressure roller, reflecting the pressure state of the coating unit on the substrate. Unstable pressure will lead to uneven adhesive coating; substrate traction speed data, which is obtained by the encoder collecting the real-time rotation speed of the main traction roller, and is a core parameter that determines the production cycle; heating roller temperature data, which is collected by a thermocouple installed on the surface of the composite heating roller, and the temperature directly affects the activation of the adhesive and the bonding strength; and coating roller vibration spectrum data, which is collected by a vibration acceleration sensor installed on the bearing seat of the coating roller and processed by fast Fourier transform to obtain spectrum information, used to identify wear or imbalance of mechanical parts in the early stage.

[0014] The process status sensing module is used to collect real-time process condition data and online quality data during the lamination production process. In this scenario, the process condition data includes workshop ambient temperature and humidity data, collected by temperature and humidity sensors deployed near the laminating machine; excessively high ambient humidity can affect the adhesive curing speed. It also includes substrate batch identification data, obtained by scanning the QR code on the substrate roll using a barcode scanner, used to associate the specific process formula of that batch of substrate. Real-time adhesive viscosity data is obtained through continuous measurement in the adhesive tank circulation pipeline using an online viscometer; viscosity deviation is one of the main process factors leading to coating defects. The online quality data includes coating uniformity image data acquired by a machine vision unit. This machine vision unit consists of a high-resolution industrial camera and a backlight source, installed after the coating process and before the lamination process, used to capture the translucent image of the coated adhesive layer. Image analysis enables online non-destructive testing of product quality.

[0015] The equipment health diagnosis module connects to the equipment operation sensing module and is used to execute equipment-side anomaly diagnosis rules based on multi-dimensional operational status data, generating equipment health indices and preliminary fault warning information. This module includes a first data preprocessing module, an equipment status assessment module, and a fault reasoning module.

[0016] The first data preprocessing module is used to preprocess the collected raw data, filter and smooth the coating head pressure data and heating roller temperature data to eliminate measurement noise and instantaneous interference, and extract characteristic frequency components from the coating roller vibration spectrum data, such as extracting characteristic amplitudes related to the bearing ball passing frequency and the natural frequency of the roller surface.

[0017] The equipment status assessment module connects to the first data preprocessing module to input preprocessed data into a pre-set equipment health assessment model. This model is trained based on historical normal operation data to generate dynamic benchmark ranges for various operating parameters. For example, the model learns the normal fluctuation range of coating head pressure at different production speeds, or the stable control range of heating roller temperature under different ambient temperatures. The equipment status assessment module compares real-time data with the corresponding dynamic benchmark range. If the data continuously deviates from the benchmark range, it calculates an equipment health index based on the degree and duration of deviation. This index is a comprehensive scalar; a decrease in the value indicates a deterioration in the overall health status of the equipment.

[0018] The fault reasoning module connects to the equipment status assessment module. When the equipment health index falls below a first preset threshold, it activates the fault reasoning engine, which contains a built-in fault mode library. This library defines typical faults such as wear of the inner ring of the coating roller bearing, partial failure of the heating rod, and jamming of the hydraulic proportional valve, and associates each fault with corresponding abnormal data pattern combinations, such as increased vibration amplitude at a specific frequency band accompanied by intensified pressure fluctuations. The fault reasoning module matches the current abnormal parameter combination with the fault mode library to generate preliminary fault warning information, including the suspected fault type, possible fault location, and confidence level.

[0019] The process quality assessment module connects to the process status awareness module. Based on process condition data and online quality data, it executes process-side deviation assessment rules to generate process quality scores and process parameter adjustment suggestions. This module includes a second data preprocessing module, a process parameter compliance module, a quality correlation analysis module, and a parameter tuning suggestion module.

[0020] The second data preprocessing module is used to perform grayscale analysis and edge detection on the coating uniformity image data, calculate the coating thickness distribution variance as a uniformity quantification value, the smaller the variance, the more uniform the coating is, and perform unit unification and validity verification on the real-time viscosity data of the adhesive, and remove outliers caused by instantaneous contamination of the sensor.

[0021] The process parameter compliance module connects to the second data preprocessing module. Based on the substrate batch identification data, it retrieves the standard process parameter range corresponding to the batch of substrate from a pre-set process parameter standard library. This standard library is pre-set by process engineers according to material characteristics and includes recommended parameter windows for temperature, pressure, speed, viscosity, etc. The process parameter compliance module compares relevant parameters in the real-time collected process condition data with the standard process parameter range, calculates a parameter compliance score, and the score reflects the degree of conformity between the current process settings and the standard formulation.

[0022] The quality correlation analysis module connects the second data preprocessing module and the process parameter compliance module to establish a mapping relationship between uniform quantified values, parameter compliance scores, and process quality scores. This mapping relationship is typically implemented by a machine learning model that learns from historical data and integrates objective detection indicators with process compliance to output a score that reflects the final composite quality expectation.

[0023] The parameter tuning suggestion module connects to the quality correlation analysis module. When the process quality score falls below a second preset threshold, it analyzes the weight of each process parameter on the score and generates targeted process parameter adjustment suggestions based on the weight ranking. For example, if the analysis finds that the current quality score decline is mainly caused by poor coating uniformity, and uniformity is most sensitive to adhesive viscosity, followed by the temperature of the heating roller, then a viscosity adjustment suggestion will be generated first, along with a temperature compensation suggestion.

[0024] The collaborative decision-making central module connects to both the equipment health diagnosis module and the process quality assessment module. It constructs and applies a bidirectional causal relationship network for the process equipment, receiving equipment health indices, preliminary fault warnings, process quality scores, and process parameter adjustment suggestions. It then performs collaborative analysis based on causal relationships, generating collaborative debugging instructions or collaborative maintenance strategies. The core innovation of this module lies in breaking down the silos between equipment diagnosis and process optimization in traditional systems. By constructing a dynamic causal relationship network, it deeply correlates equipment anomalies with process parameter deviations at the root cause level, achieving cross-domain collaborative decision-making. For a detailed causal reasoning process, please refer to [link to relevant documentation]. Figure 2 The unexpected technical benefits are that the system can not only handle surface faults, but also identify and resolve hidden equipment damage induced by process factors, or process optimization bottlenecks constrained by equipment status. This fundamentally improves the overall robustness and comprehensive efficiency of the production system, something that single-dimensional diagnostic or optimization systems cannot achieve. This module includes a causal correlation mapping module and a collaborative adjudication module.

[0025] The causal association mapping module is used to maintain and dynamically update a bidirectional causal association network for a process device. This network interconnects nodes with abnormal device status with nodes with deviations in process parameters. Each association edge records the causal direction, influence strength coefficient, and historical trigger frequency. For example, there might be a causal edge in the network that leads from a process node with high adhesive viscosity to a device node with increased vibration of the coating roller; the influence strength coefficient indicates the strength of the former's effect on the latter.

[0026] The causal association mapping module performs bidirectional reasoning at runtime. When it receives an initial fault warning from the equipment side, it traces backward along the associated edges to one or more process root cause nodes that may have caused the equipment malfunction, generating a deep diagnostic report containing process triggers. For example, after receiving a warning about wear on the coating roller bearing, network backtracking reveals a strong correlation with several historical process events where incomplete adhesive filtration led to impurity particles entering the bearing. When it receives a process parameter adjustment suggestion from the process side, it predicts forward along the associated edges the equipment status risk nodes that might be triggered by implementing the adjustment, generating a forward-looking equipment risk assessment report. For example, when planning to increase production speed to improve efficiency, the network predicts that this may exacerbate the risk of overheating and wear on the traction motor.

[0027] The collaborative adjudication module connects to the causal correlation mapping module for comprehensive adjudication based on in-depth diagnostic reports or equipment risk prospective assessment reports. If the adjudication addresses equipment malfunction, it generates a collaborative maintenance strategy that incorporates collaborative corrections to process parameters. This strategy includes not only equipment maintenance recommendations but also preventative adjustment instructions for related process parameters. For example, when deciding to replace a worn coating roller bearing, the strategy might simultaneously recommend increasing the adhesive preheating temperature by a certain value for a period before and after maintenance to reduce its rheological resistance and lessen the initial load on the new bearing. If the adjudication focuses on process optimization, it generates collaborative commissioning instructions that incorporate equipment state constraints. These instructions, in addition to process parameter adjustment recommendations, add safety margin limits based on the equipment health index. For example, the instruction might allow increasing the heating roller temperature by five degrees Celsius to improve composite strength, but with the condition that the heating roller drive current must be monitored simultaneously. If the current consistently exceeds the safety threshold allowed by the current equipment health index, the heating process is automatically terminated.

[0028] The causal association mapping module maintains a bidirectional causal association network for process equipment. Its construction and dynamic update mechanism includes an initial network generation unit, a data-driven learning unit, a feedback reinforcement unit, and a cross-machine knowledge fusion unit.

[0029] The initial network generation unit is used to derive the fundamental theoretical relationships between equipment parameters and process parameters based on the equipment physical mechanism model and the process chemical model, forming the network skeleton. For example, the theoretical relationship between coating head pressure and adhesive viscosity is derived based on fluid mechanics, and the theoretical relationship between heating roller temperature and ambient temperature is derived based on the heat conduction model. These relationships are then transformed into the initial nodes and edges of the network.

[0030] The data-driven learning unit continuously analyzes the time-series data streams reported by the equipment operation sensing module and the process status sensing module. It employs an algorithm based on a causal correlation strength calculation function to mine statistical causal sequences between equipment and process variables from the data, transforming or strengthening strongly correlated sequences into or enhancing the correlation edges in the network. The core invention lies in the fact that the system does not rely on human experience to fully construct complex causal graphs; instead, it automatically discovers and quantifies unknown or difficult-to-describe implicit causal relationships through data-driven algorithms. The unexpected technical effect is that the system can adaptively learn and solidify production line-specific operating experiences. For example, it may discover a unique causal chain where a specific brand of substrate, under certain humidity conditions, leads to increased electrostatic adsorption, resulting in slight slippage of the traction roller. This gives the diagnostic and decision-making model personalized and continuously evolving capabilities, far exceeding the limitations of static knowledge bases.

[0031] The feedback reinforcement unit connects to the closed-loop optimization module to obtain comprehensive feedback on the effects of each collaborative decision. If the decision successfully solves the problem, the weight of the key causal path on which the decision was based is increased; if the decision fails to meet expectations, a re-evaluation and correction of that path is triggered. This enables the network to learn from practical experience, and its knowledge continuously evolves over time.

[0032] The cross-device knowledge fusion unit connects to the edge communication gateway module to receive common causal patterns extracted from data generated by multiple devices of the same model from a remote cloud platform, and then securely integrates them with the local network. This allows individual devices to benefit from collective experience, quickly gaining knowledge about rare faults or complex coupling relationships, thus improving the completeness and accuracy of the network.

[0033] The causal relationship strength calculation function used in the data-driven learning unit is as follows:

[0034] In this bonding machine scenario, the parameters of this function are explained as follows: : Indicates the strength of causal relationship, i.e., the order of maximum time lag. Within, a specific process parameter For a specific device parameter The causal influence strength. The calculation result is a value between zero and one. The closer the value is to one, the stronger the causal influence of the process parameter on the equipment parameter.

[0035] and : These represent points in time. The specific equipment parameter values ​​and process parameter values ​​collected. For example, It could be the amplitude of the coating roller vibration spectrum at a certain characteristic frequency. It can be the real-time viscosity value of the adhesive.

[0036] : This is the preset maximum time lag order, representing the maximum time range within which changes in process parameters may affect equipment parameters. For example, setting... A value of ten indicates that the potential impact of process parameter values ​​over the past ten sampling periods on the current equipment parameter values ​​is being examined.

[0037] : Indicates time Equipment parameters The estimated value is based on its own historical data. Or, based on both its own history and the history of process parameters... To make predictions.

[0038] : This represents variance calculation, used to measure the uncertainty of prediction. The core idea of ​​the ratio part in the formula is that if incorporating historical information about process parameters significantly reduces the prediction variance of the current value of the equipment parameters, it indicates that the process parameters contain additional causal information.

[0039] : Represents a random variable and Standardized mutual information between them is used to quantify the time delay. The statistical dependence strength between historical values ​​of process parameters and current values ​​of equipment parameters is used as the prior weight of causal strength.

[0040] The calculation logic of the formula can be understood as: at multiple different time lags The system assesses the improvement in equipment parameter prediction accuracy after incorporating historical process parameter information, multiplies this improvement by the statistical dependence of the two parameters at that time delay, and finally averages the results for all time delays to obtain a robust causal strength estimate. This helps the system automatically discover implicit causal relationships from massive amounts of operational data, such as the fact that an abnormal increase in adhesive viscosity (process parameter) usually lags behind aggravated fluctuations in coating head pressure (equipment parameter).

[0041] The closed-loop optimization module connects the collaborative decision-making central module, the equipment operation sensing module, and the process status sensing module. It executes collaborative debugging commands and iteratively optimizes the parameters of equipment-side anomaly diagnosis rules, process-side deviation assessment rules, and causal relationship networks based on multi-dimensional feedback data after command execution. The core invention of this module lies in establishing a complete self-learning closed loop that spans sensing, decision-making, execution, and feedback. This transforms the system from a static rule executor into an intelligent agent capable of learning from each intervention result and dynamically adjusting its judgment logic. For its working principle, please refer to [link to relevant documentation]. Figure 3 The unexpected technical benefits are that the system can automatically adapt to dynamic factors such as equipment aging, batch-to-batch material differences, and slow environmental changes, ensuring that diagnostic thresholds, evaluation standards, and causal understanding always maintain optimal matching with the current actual production status. This sustains high-precision diagnostic and decision-making capabilities over the long term, significantly reducing the risk of false alarms, missed alarms, or decision failures caused by system rigidity. This module includes an instruction issuance and execution tracking module, a multi-dimensional feedback data acquisition module, a rule parameter optimization module, and an optimization parameter synchronization module.

[0042] The instruction issuance and execution tracking module sends collaborative debugging instructions to the laminating machine controller and monitors changes in key execution parameters to confirm that the instructions have been executed. For example, after issuing a heating instruction, it continuously reads the feedback from the temperature controller to confirm that the heating roller temperature has started to rise as required by the instruction.

[0043] The multi-dimensional feedback data acquisition module connects to the command issuance and execution tracking module. Within a preset observation window after command execution, it collects changes in multi-dimensional operational status data and online quality data through the equipment operation sensing module and process status sensing module, serving as feedback data. The length of the observation window must be sufficient to assess the short-term effects of the command, such as continuously monitoring production data for the following hour after adjustments.

[0044] The rule parameter optimization module connects to the multi-dimensional feedback data acquisition module to analyze the feedback data, calculate the optimization amount based on the rule parameter update function, and iteratively adjust the parameters of the equipment-side anomaly diagnosis rules, process-side deviation assessment rules, and causal relationship networks. This is the core calculation step of the system's adaptive optimization.

[0045] The rule parameter optimization module uses the following rule parameter update function:

[0046] In this bonding machine scenario, the parameters of this function are explained as follows: This indicates the amount of change in the rule parameter vector that needs to be adjusted in this iteration. Rule parameters can include anomaly detection thresholds in equipment health diagnosis models, weight coefficients in process quality assessment models, edge weights in causal relationship networks, etc.

[0047] : is the learning rate, a constant greater than zero, used to control the step size of each parameter adjustment to prevent over-updating.

[0048] and These represent the actual change in the equipment health index and the change originally expected by the collaborative decision-making center module within the preset observation window after executing the collaborative debugging command. For example, the health index was expected to increase by fifteen points after equipment maintenance, but it only increased by ten points in reality.

[0049] and These represent the actual and expected changes in the process quality score within the preset observation window after executing the collaborative debugging command. For example, the expected improvement in the quality score after process adjustment was eight points, but the actual improvement was nine points.

[0050] and These are the normalization coefficients for the deviations in equipment health and process quality, used to map the difference between the actual and expected values ​​to a suitable numerical scale, and are usually obtained based on historical data statistics.

[0051] : is a hyperbolic tangent function, its function is to compress the deviation value of the equipment health index to between negative one and one, so that the adjustment amount This deviation exhibits a saturated response, preventing drastic parameter oscillations caused by excessively large single deviations.

[0052] : This is a penalty term based on a Gaussian function. Its meaning is that when the actual change in the process quality score deviates significantly from expectations (whether better or worse than expected), the value of this penalty term decreases, thus inhibiting parameter updates. This prompts the system to prioritize robust optimization paths that balance the improvement of equipment health and process quality, avoiding sacrificing one for the other.

[0053] : This is the feature vector associated with the optimized rule. It indicates which specific rule parameters the current feedback data is associated with, ensuring that the optimization is applied to the correct parameters.

[0054] The formula's calculation logic embodies the system's optimization philosophy: the direction of parameter adjustment is driven by the deviation in equipment health improvement (tanh term), while the magnitude of adjustment is simultaneously constrained by whether process quality improvement meets expectations (exp penalty term). This enables the system to intelligently balance the two major goals of equipment maintenance and process optimization, adaptively adjusting its internal rules towards maximizing overall benefits.

[0055] The optimized parameter synchronization module connects to the rule parameter optimization module, enabling real-time synchronization of optimized rule parameters to the equipment health diagnosis module, process quality assessment module, and collaborative decision-making center module. This closed-loop learning process allows the system to make more accurate judgments and decisions when faced with similar situations in the future.

[0056] The edge communication gateway module connects the collaborative decision-making center module, the closed-loop optimization module, and the remote cloud platform, enabling data uploading and the reception of model and rule update commands between the system and the remote cloud platform. Its interaction includes a data uploading unit, a cloud model receiving unit, and a local model update unit.

[0057] The data upload unit is used to anonymize system operation data and collaborative analysis process data, and then packages the anonymized data, locally optimized rule parameters, and system operation logs to upload to the remote cloud platform. The uploaded data provides raw materials for the cloud platform to perform global model training and knowledge mining.

[0058] The cloud model receiving unit receives global optimization resources from a remote cloud platform, including an enhanced equipment health assessment model, an optimized library of general process parameters, and an updated causal relationship knowledge graph. These resources integrate the operational wisdom of numerous similar devices, surpassing the knowledge of a single device.

[0059] The local model update unit connects to the cloud model receiving unit and is used to securely merge and incrementally update the received global optimization resources with the corresponding local model. For example, it can merge new fault modes issued by the cloud platform with the fault mode library of the local fault inference engine, but compatibility verification must be performed before merging to avoid conflicts.

[0060] The human-machine interface module connects the collaborative decision-making center module and the closed-loop optimization module. It provides on-site operators or remote engineers with a visual representation of equipment health indices, process quality scores, collaborative debugging instructions, or collaborative maintenance strategies, as well as optimization logs from the closed-loop optimization module. This interface is typically presented as an industrial dashboard or remote desktop, offering intuitive displays such as charts, curves, and diagnostic trees. The human-machine interface module receives and forwards manual confirmation instructions, parameter fine-tuning instructions, or emergency intervention instructions to the collaborative decision-making center module or the closed-loop optimization module. This ensures an organic combination of AI decision-making and human expert experience, retaining necessary human oversight and veto power while maintaining automation.

[0061] Through the coordinated operation of the above modules, this system achieves a complete intelligent process on the fabric or film lamination production line, from data perception and independent diagnostic assessment to cross-domain collaborative decision-making and closed-loop execution optimization. When an alarm for uneven coating quality occurs on the production line, the system not only suggests parameter adjustments from a process perspective but also analyzes the equipment status to determine whether abnormal vibration of the coating roller is causing the process problem and provides collaborative maintenance strategies. Conversely, when planning to increase production, the system proactively assesses the risks to equipment lifespan and provides speed-up solutions with constraints. Simultaneously, the effects of each decision execution are fed back to optimize the system's own judgment rules and knowledge network. This system effectively improves the operational reliability, process stability, and maintenance and debugging efficiency of the laminating machine, making it a key link in realizing intelligent manufacturing.

[0062] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. A fitting machine remote parameter debugging and fault diagnosis system of edge computing, characterized in that, The method comprises the following steps: A device operation perception module is used to collect multi-dimensional operation state data of a target industrial device in real time; A process state perception module is used to collect process condition data and online quality data of a target production process in real time; A device health diagnosis module is used to execute device-side abnormal diagnosis rules based on the multi-dimensional operation state data, and generate a device health index and preliminary fault warning information; A process quality evaluation module is used to execute process-side deviation evaluation rules based on the process condition data and the online quality data, and generate a process quality score and process parameter adjustment suggestions; A collaborative decision hub module is used to build and apply a bidirectional causal correlation network between process parameters and device states, receive the device health index, the preliminary fault warning information, the process quality score and the process parameter adjustment suggestions, execute collaborative analysis based on causal correlation, and generate collaborative debugging instructions or collaborative maintenance strategies; A closed-loop optimization module is used to execute the collaborative debugging instructions, and based on multi-dimensional feedback data after instruction execution, iteratively optimize the device-side abnormal diagnosis rules, the process-side deviation evaluation rules and parameters of the causal correlation network; An edge communication gateway module is used to realize data uploading between the system and a remote cloud platform connected by the system, and receiving model / rule update instructions.

2. The edge-computing splicing machine remote parameter debugging and fault diagnosis system according to claim 1, characterized in that: The device health diagnosis module comprises: A first data preprocessing module is used to filter and smooth the pressure data and process temperature data, and extract characteristic frequency components from the vibration frequency spectrum data; A device state evaluation module is used to input the preprocessed data into a preset device health evaluation model, and the device health evaluation model is trained based on historical normal operation data to generate dynamic baseline ranges for each operation parameter; The device state evaluation module compares real-time data with the corresponding dynamic baseline ranges, and if the data continuously deviates from the baseline ranges, the device health index is calculated according to the deviation degree and the duration; A fault reasoning module is used to activate a fault reasoning engine when the device health index is lower than a first preset threshold, and the fault reasoning engine has a built-in fault mode library; The fault reasoning module matches the current abnormal parameter combination with the fault mode library to generate preliminary fault warning information containing suspected fault types, possible fault locations and confidence levels.

3. The edge-computing splicing machine remote parameter debugging and fault diagnosis system according to claim 1, characterized in that: The process quality evaluation module comprises: A second data preprocessing module is used to perform image feature analysis and extraction on the quality image data, calculate a quantitative value representing quality uniformity, and perform unit unification and validity verification on the process medium characteristic data; A process parameter compliance module is used to retrieve standard process parameter ranges corresponding to the batch of materials from a preset process parameter standard library according to the material batch identification data; The process parameter compliance module compares related parameters in the real-time collected process condition data with the standard process parameter ranges to calculate a parameter compliance score; A quality correlation analysis module is used to establish a mapping relationship between the quality uniformity quantitative value, the parameter compliance score and the process quality score. The parameter tuning suggestion module is configured to, when the process quality score is lower than a second preset threshold, analyze an influence weight of each process parameter on the score, and generate a targeted process parameter adjustment suggestion according to a weight order.

4. The edge-computing splicing machine remote parameter debugging and fault diagnosis system according to claim 1, characterized in that: The collaborative decision hub module comprises: The causal correlation mapping module is configured to maintain and dynamically update a bidirectional causal correlation network between process parameters and equipment states; the network interconnects equipment state abnormal nodes and process parameter deviation nodes, and each correlation edge records a causal direction, an influence intensity coefficient and a historical triggering frequency; The causal correlation mapping module performs bidirectional reasoning in runtime: when the preliminary fault early warning information on the equipment side is received, the correlation edge is reversely traced to one or more process root cause nodes that may cause the equipment abnormality, and a deep diagnosis report containing process causes is generated; When the process parameter adjustment suggestion on the process side is received, the correlation edge is forwardly predicted to execute the adjustment to generate an equipment risk forward-looking evaluation report; The collaborative decision module is configured to make a comprehensive decision based on the deep diagnosis report or the equipment risk forward-looking evaluation report; If the decision is to handle the equipment abnormality, the collaborative maintenance strategy that integrates process parameter collaborative correction is generated, and the strategy not only contains equipment maintenance suggestions, but also contains preventive adjustment instructions for associated process parameters; If the decision is to execute process optimization, the collaborative debugging instruction that integrates equipment state constraints is generated, and the instruction adds a safety margin limit based on the equipment health index to the process parameter adjustment suggestion.

5. The edge-computing splicing machine remote parameter debugging and fault diagnosis system according to claim 4, characterized in that: The bidirectional causal correlation network between process parameters and equipment states maintained by the causal correlation mapping module comprises: An initial network generation unit is configured to deduce basic theoretical correlations between equipment parameters and process parameters based on equipment physical mechanism models and process models, and form a network skeleton; A data-driven learning unit is configured to continuously analyze time series data streams reported by the equipment operation perception module and the process state perception module, use an algorithm based on a causal correlation intensity calculation function to mine statistical causal sequences between equipment variables and process variables from the data, and convert or enhance correlation edges in the network from strong correlation sequences; A feedback reinforcement unit is configured to obtain a comprehensive effect feedback after each collaborative decision is executed; if the decision successfully solves a problem, the weight of a key causal path on which the decision is based is enhanced; if the decision does not achieve an expected result, the key causal path is reevaluated and corrected; A cross-machine knowledge fusion unit is configured to receive common causal patterns extracted from running data of multiple devices of the same type from the remote cloud platform, and safely fuse the common causal patterns with the local network.

6. The edge-computing splicing machine remote parameter commissioning and fault diagnosis system according to claim 6, characterized in that: The causal correlation intensity calculation function used by the data-driven learning unit is as follows: ， wherein, : denotes the strength of the causal relationship, i.e. the causal influence of a certain process parameter on a certain device parameter within a maximum time lag order , and : denote the specific device parameter value and process parameter value, respectively, collected at time point , : is the pre-defined maximum time lag order, : denotes the estimated value of the device parameter at time , which is predicted based on its own historical data or simultaneously based on its own historical and process parameter historical data, : denotes the variance calculation, which measures the uncertainty of the prediction, : denotes the normalized mutual information between the random variable and .

7. The edge-computing splicing machine remote parameter commissioning and fault diagnosis system according to claim 1, characterized in that: The closed-loop optimization module comprises: An instruction issuing and execution tracking module is configured to send the collaborative debugging instruction to a controller of a target industrial equipment, and monitor changes in key execution parameters to confirm that the instruction is executed. A multi-dimensional feedback data collection module is configured to collect changes in the multi-dimensional running state data and the online quality data as feedback data through the device running perception module and the process state perception module within a preset observation window after execution of the instruction; A rule parameter optimization module is configured to analyze the feedback data, calculate an optimization amount according to a rule parameter update function, and iteratively adjust parameters of the device-side abnormal diagnosis rule, the process-side deviation evaluation rule, and the causal correlation network; An optimized parameter synchronization module is configured to synchronize the optimized rule parameters to the device health diagnosis module, the process quality evaluation module, and the collaborative decision hub module in real time.

8. The edge-computing splicing machine remote parameter commissioning and fault diagnosis system according to claim 7, characterized in that: The rule parameter update function used by the rule parameter optimization module is as follows: ， wherein, : represents the variation of the rule parameter vector that needs to be adjusted in this iteration, : is the learning rate, and : respectively represent the actual variation of the equipment health index and the variation expected by the collaborative decision hub module in the preset observation window after executing the collaborative debugging instruction, and : respectively represent the actual variation of the process quality score and the expected variation in the preset observation window after executing the collaborative debugging instruction, and : are respectively the effect bias normalization coefficients of the equipment health and the process quality, : is the hyperbolic tangent function, : is the penalty term based on the Gaussian function, : is the feature vector related to the optimized rule.

9. The edge-computing splicer remote parameter commissioning and fault diagnosis system of claim 1, wherein: The interaction between the edge communication gateway module and the remote cloud platform includes: A data uploading unit is configured to desensitize system running data and collaborative analysis process data, and package and upload the desensitized data, locally optimized rule parameters, and system running logs to the remote cloud platform; A cloud model receiving unit is configured to receive global optimization resources issued by the remote cloud platform, the global optimization resources including an enhanced device health evaluation model, an optimized general process parameter standard library, and an updated version of a causal correlation knowledge graph; A local model updating unit is configured to securely fuse and incrementally update the received global optimization resources with corresponding local models.

10. The edge-computing in-line machine remote parameter tuning and fault diagnosis system according to any one of claims 1-9, characterized in that: The system further includes: A human-computer interaction interface module is configured to visually display the device health index, the process quality score, the collaborative debugging instruction or the collaborative maintenance strategy, and the optimization log of the closed-loop optimization module to a field operator or a remote engineer; The human-computer interaction interface module receives and forwards a manual confirmation instruction, a parameter fine-tuning instruction, or an emergency intervention instruction to the collaborative decision hub module or the closed-loop optimization module.

Images