Safety monitoring and chain protection system of production line

Abstract

The invention provides a safety monitoring and chain protection system for a production line, and the system comprises the steps: collecting and updating all data in a production process in real time through each node sensor of the production line, and integrating all data into a production data set; a production data set is used for training an IF model, distance measurement generated by the IF model is used as distance measurement of DBSCAN, anomaly detection of all parts of a production line is completed, and a detection result is fed back to a fault prediction module and a corresponding protection module; based on a production data set corresponding to the feedback result, performing fault prediction on each part of the production line through a built-in time sequence neural network; and executing corresponding fault processing on corresponding parts of the production line based on a chain protection mechanism. The method has the beneficial effects that the reliability, the safety and the production efficiency of equipment are effectively improved through anomaly detection of multi-model fusion, accurate time sequence prediction and rapid interlocking protection disposal.

Description

Technical Field

[0001] The present invention belongs to the field of production line control, and in particular relates to a safety monitoring and chain protection system for a production line. Background Art

[0002] In the context of Industry 4.0, intelligent manufacturing systems are increasingly adopting collaborative control and self-diagnosis technologies, leveraging sensors, data analysis, and artificial intelligence to achieve automated and intelligent management of production lines. These systems typically involve the collaboration of multiple modules, including data collection, predictive analysis, fault diagnosis, anomaly detection, and protective response.

[0003] But the defects of the prior art are as follows: Traditional production line safety monitoring systems mostly rely on simple threshold settings or conventional rule engines for anomaly detection. This approach is insensitive to complex production data sets and struggles to detect subtle anomalies. Existing systems lack effective data mining and machine learning technology, resulting in limited detection capabilities for potential failures and an inability to provide accurate predictions and early warnings.

[0004] Many traditional monitoring systems rely on separate anomaly detection and fault prediction modules, lacking synergy. This design prevents the system from leveraging early signals from anomaly detection to predict failures. It also lacks comprehensive monitoring of anomaly trends, increasing the risk of the system failing to provide timely warnings of failures.

[0005] Traditional protection and response mechanisms rely heavily on manual judgment and pre-set rules, lacking flexibility and intelligence. When faced with complex production environments and unexpected failures, protective measures can lag behind, preventing accurate, real-time processing. This can lead to prolonged production line downtime or equipment damage.

[0006] Many existing technologies lack real-time data collection and fault detection, preventing production process failures from being discovered promptly and preventing major accidents. This not only impacts production line safety but can also lead to lower product quality and inefficient production.

[0007] Existing safety monitoring systems are often closed and unable to effectively integrate with other intelligent manufacturing systems or factory management systems. This leads to severe data silos, making global optimization and information sharing difficult. Furthermore, traditional systems lack scalability, making them difficult to adapt to future technological developments and new production demands. Summary of the Invention

[0008] In view of this, the present invention aims to provide a safety monitoring and interlocking protection system for a production line, in order to solve at least one of the above-mentioned technical problems.

[0009] To achieve the above object, the technical solution of the present invention is realized as follows: The first aspect of the present invention provides a safety monitoring and interlock protection system for a production line, including: A data acquisition module, which is configured to collect and update in real time various data in the production process through sensors at each node of the production line, and integrate various data into a production data set; An anomaly detection module, which is configured to train an IF model using the production data set, use the distance metric generated by the IF model as the distance metric of DBSCAN, complete the anomaly detection of each part of the production line, and feed the detection results back to the fault prediction module and the protection response module; A fault prediction module, which responds to the feedback of the anomaly detection module and, based on the production data set corresponding to the feedback result, performs fault prediction on each part of the production line through a built-in time series neural network; A protection response module, which responds to the feedback of the anomaly detection module and the fault prediction module and, based on the interlock protection mechanism, performs corresponding fault handling on the corresponding part of the production line.

[0010] Further, before integrating various data into a production data set, preprocessing operations are performed on various data, including: Using the forward filling method to process missing values in the data; reducing data noise by extracting frequency domain features for high-frequency vibration signals; summarizing the data of each node sensor based on the finest-grained time step, and aligning various data according to the time step.

[0011] Further, the working process of the anomaly detection module includes: Offline training of the IF model based on the historical production data set, using the IF model to infer the current production data set and calculate the anomaly score, and transforming and mapping the anomaly score into a distance metric between samples.

[0012] Further, the working process of the anomaly detection module further includes: Based on the mapped distance metric, using the DBSCAN clustering algorithm to cluster the production data set, identifying and determining the anomaly clusters and isolated noise points in the production data set, and obtaining anomaly information.

[0013] Further, the working process of the fault prediction module includes: Synchronizing the production data set with the anomaly detection result, and constructing an input sequence within a time window; Using the time series neural network to perform forward inference on the input sequence, and outputting the prediction of key indicators for several future time steps; Denormalizing the result of the key indicator prediction, and calculating the confidence interval according to the uncertainty estimate output by the model; By comparing with the preset soft and hard thresholds and confidence interval probabilities, the future fault risks of each node are classified and evaluated, and corresponding warning labels are generated.

[0014] Further, the working process of the protection response module includes: Receiving feedback content and dividing the current and future states into four levels of risks: normal, warning, fault, and emergency according to the preset hard threshold and soft threshold; if it is at the warning level, trigger a warning through sound and light alarms and operator notifications; if it is at the fault level, reduce the load to relieve stress; if it is at the emergency level, immediately cut off the power supply and shut down slowly.

[0015] Further, the system is built-in with a redundancy mechanism, including: Real-time record and update key configurations and operation logs. After confirming that the fault is repaired, the system is automatically restored to the normal working state according to the record at the moment before the fault handling is executed.

[0016] The second aspect of the present invention provides an electronic device, including a processor and a memory communicatively connected to the processor and used for storing instructions executable by the processor, and the processor is used to execute the system described in the first aspect above.

[0017] The third aspect of the present invention provides a server, including at least one processor and a memory communicatively connected to the processor, and the memory stores instructions executable by the at least one processor, and the instructions are executed by the processor to enable the at least one processor to execute the system as described in the first aspect.

[0018] The fourth aspect of the present invention provides a computer-readable storage medium, storing a computer program, and when the computer program is executed by a processor, it implements the system described in the first aspect.

[0019] Compared with the prior art, the safety monitoring and interlock protection system for a production line described in the present invention has the following beneficial effects: Through the real-time capture and summary of the sensor data of each node on the production line by the data acquisition module, it can completely and timely reflect the changes of various indicators in the production process, providing a reliable and high-quality data basis for subsequent anomaly detection and fault prediction.

[0020] The anomaly detection module first uses the Isolation Forest (IF) model to model the production data set, and then introduces its "distance metric" into the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm, effectively combining the advantages of the Isolation Forest and DBSCAN. This composite detection strategy significantly improves the detection accuracy and recall rate, and effectively reduces the false negative and false positive rates.

[0021] Based on the anomaly detection results, the fault prediction module selects the corresponding historical and real-time data and models them through built-in time series neural networks (such as LSTM, GRU, etc.). It can not only alert when an anomaly occurs, but also predict potential faults in advance, giving production managers more sufficient response time, thereby further reducing the risk of production line downtime.

[0022] Based on the dual feedback of anomaly detection and fault prediction, the protection response module can take targeted actions on relevant components or subsystems of the production line relying on the interlock protection mechanism. Brief Description of the Drawings

[0023] The drawings forming a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings: Figure 1 It is a schematic structural diagram of a safety monitoring and interlock protection system for a production line according to an embodiment of the present invention. Detailed Embodiments

[0024] It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0025] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "transverse", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present invention. In addition, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first", "second", etc. may explicitly or implicitly include one or more of such features. In the description of the present invention, unless otherwise stated, the meaning of "plurality" is two or more.

[0026] In the description of the present invention, it should be noted that unless otherwise clearly defined and limited, the terms "installed", "connected", "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood through specific situations.

[0027] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0028] A safety monitoring and interlock protection system for a production line, comprising: A data acquisition module, which is configured to collect and update in real time various data during the production process through sensors at each node of the production line, and integrate the various data into a production data set; An anomaly detection module, which is configured to train an IF model using the production data set, use the distance metric generated by the IF model as the distance metric of DBSCAN, complete the anomaly detection of each part of the production line, and feed the detection results back to the fault prediction module and the protection response module; A fault prediction module, which responds to the feedback of the anomaly detection module and, based on the production data set corresponding to the feedback result, performs fault prediction on each part of the production line through a built-in time series neural network; A protection response module, which responds to the feedback of the anomaly detection module and the fault prediction module and, based on the interlock protection mechanism, performs corresponding fault handling on the corresponding part of the production line.

[0029] Before integrating the various data into a production data set, preprocessing operations are performed on the various data, including: Using the forward filling method to process missing values in the data; reducing data noise for high-frequency vibration signals by extracting their frequency domain features; summarizing the data of each node sensor based on the finest-grained time step, and aligning the various data according to the time step.

[0030] The working process of the anomaly detection module includes: Offline training of the IF model based on the historical production data set, using the IF model to infer the current production data set and calculate the anomaly score, and transforming and mapping the anomaly score into a distance metric between samples.

[0031] Based on the distance metric obtained by the mapping, using the DBSCAN clustering algorithm to cluster the production data set, identifying and determining the abnormal clusters and isolated noise points in the production data set, and obtaining the anomaly information.

[0032] In some embodiments, the specific execution process of the above anomaly detection module is as follows: The working idea of the anomaly detection module is: When training IsolationForest, each tree splits the samples into different leaf nodes; For any two samples \(x_i, x_j\), their "similarity" or "path distance" can be defined as: Similarity: Count the number of times the two samples fall into the same leaf node in the same tree / total number of trees → high similarity; Path distance: The mean of the differences in leaf depths of two samples on each tree.

[0033] Use the above similarity (or path distance) matrix directly as the input distance metric for DBSCAN (metric = "precomputed"). In this way, when clustering, the closeness of samples is defined not simply based on Euclidean distance, but on the "separability of points from the perspective of the Isolation Forest".

[0034] The Isolation Forest is naturally sensitive to high-dimensional sparse anomalies, while DBSCAN is more sensitive to multi-density clusters. Combining the two can take both into account. By adjusting the eps and min_samples of DBSCAN, both fault clusters and outlier points (noise) can be identified in the similarity space derived from the IF.

[0035] 1. Data collection and preprocessing: The real-time sensor stream from the data collection module includes various types such as temperature, pressure, vibration signals, current, voltage, and rotational speed, corresponding to different nodes and different sampling frequencies. It is necessary to unify the timestamps and align them (sliding window alignment or interpolation processing can be used).

[0036] Feature extraction Time-domain features: Such as the average value, standard deviation, maximum / minimum value, skewness / kurtosis, and interquartile range within the window.

[0037] Frequency-domain features: Perform FFT on signals such as vibration and sound, and extract features such as spectral energy distribution, main frequency, and harmonic components.

[0038] Time-series features: Such as the trend slope, moving average, and exponentially weighted average of short-term or long-term sliding windows, which can help capture gradual anomalies.

[0039] Statistical / aggregation features: Combined features across sensors (such as the difference or ratio of multiple sensors of the same device), which help capture collaborative anomalies.

[0040] If the feature dimension is too high, perform dimensionality reduction (such as PCA, Autoencoder embedding, etc.) before IF training to improve the efficiency and effect of the IF model. If sensitive to real-time computational overhead, the most discriminative features can be pre-evaluated and only the core features are retained.

[0041] 2. Offline training of the IsolationForest (IF) model: Training purpose: The IsolationForest is good at "isolating" outlier samples through a random tree structure in high-dimensional data to give an anomaly score. In the training stage, data representing the "normal operation" state should be used as much as possible.

[0042] Data Selection: Collect historical data for a sufficiently long time and try to exclude known faulty / abnormal periods as much as possible to form a training set; if there are already labeled abnormal points, they can be excluded before training or used for subsequent evaluation; if there is no labeling, assume that most data is normal.

[0043] Training Process: Input the preprocessed feature data into the IF training interface to obtain a trained model object; at the same time, save the normalization / standardization parameters, dimensionality reduction model (if any), and IF model serialization parameters for online inference.

[0044] Model Update Strategy: As the production environment changes (new equipment, new processes, new sensors), the IF model may become invalid. Adopt a periodic retraining (such as weekly / monthly) or incremental training mechanism, collect the most recent normal state data, and retrain in combination with the old model or directly; at the same time, retain a part of the historical data to prevent the model from forgetting early patterns.

[0045] 3. Construct an IF-driven similarity / distance matrix: 31. IsolationForest Model Training: For the historical production data set , we train a tree-based anomaly detection model f IF (), and the anomaly score s i output by this model can be represented in the following way: s i =f IF (x i )_for_i = 1, …, N train 32. Multiple Transformations and Normalization of Anomaly Scores: Normalize the anomaly scores. First, calculate the minimum and maximum values of all anomaly scores: ; Then, perform Min - Max normalization: ; Introduce a non - linear transformation and a weighting factor to enhance the influence of the anomaly scores on the final distance and improve the way of mapping the anomaly scores to the distance metric: ; where: w i is the weight of the sample, which can be set according to the importance of the sample or other prior information; γ is a tuning factor that controls the degree of non - linear transformation of the scores; p is a hyperparameter that controls the power of the mapping, making the samples with a high degree of anomaly more distant from the normal samples.

[0046] In this way, the weight sum of the outlier points and their influence on the distance are enhanced, making the outlier points more easily recognizable in subsequent clustering.

[0047] 33. Construction of the distance matrix between samples: Introduce a weighted distance metric and calculate the distance matrix between samples through weighted calculation: ; where: ∣d i −d j ∣ is the difference between the outlier scores (distance metrics); α is a hyperparameter that controls the influence of the outlier score difference; ∥x i −x j ∥ 2 is the Euclidean distance in the sample feature space, measuring the similarity between samples; σ controls the influence range of the distance in the feature space and is usually adjusted through cross-validation.

[0048] This distance matrix fuses both the outlier scores and the similarity in the sample feature space, enabling clustering to consider not only the outlier degree but also the potential similarity between samples.

[0049] 34. Enhancement of DBSCAN clustering: In DBSCAN clustering, introduce a dynamic threshold and a density-based multi-scale clustering strategy for more refined outlier detection.

[0050] The neighborhood radius ε i and the minimum number of samples minPts of DBSCAN are key hyperparameters, which are dynamically adjusted according to the outlier degree of the samples: ε i =ε base +λ·D ij ; where, ε base is the base neighborhood radius and λ is the adjustment factor.

[0051] For clusters of different scales, process them through the multi-layer DBSCAN algorithm and dynamically adjust the density threshold according to the density change of the data: c i =DBSCAN(D,ε i ,minPts); This means that the higher the outlier degree of a sample, the larger its ε i , and DBSCAN will cluster these samples more loosely, thus being able to identify relatively "loose" outlier clusters.

[0052] 35. Outlier point identification: Noise points (outlier points) are defined as samples that do not belong to any cluster. We use: ; Define the size of the cluster: ; Then, select the clusters smaller than the threshold τ as the abnormal clusters: ; Finally, merge these abnormal clusters into the abnormal point set: .

[0053] 36. Final abnormal point set: Merge the noise points and small cluster points to obtain the final abnormal point set: .

[0054] The working process of the fault prediction module includes: Synchronize the production data set with the anomaly detection results and construct the input sequence within the time window; Use the time series neural network to perform forward inference on the input sequence and output the prediction of key indicators for several future time steps; Denormalize the results of the key indicator prediction and calculate the confidence interval according to the uncertainty estimation output by the model; By comparing with the preset soft and hard thresholds and the confidence interval probability, classify and evaluate the future fault risks of each node and generate corresponding warning labels.

[0055] In some embodiments, the specific execution process of the above fault prediction module is as follows: 1. Data synchronization and input sequence construction: Synchronize production data with anomaly detection: Synchronize the real-time data collected during the production process with the results of the anomaly detection module. The anomaly detection results may include whether there are faults in the equipment and whether the sensors are abnormal; Construct the input sequence within the time window: Based on the synchronized data, divide it according to a fixed time window and organize the data into a time series input. The input data within each time window is used for training and prediction, including data for multiple time steps (e.g., data for the past few hours or days).

[0056] 2. Time series neural network inference: Transfer the input data to the time series neural network: Input the constructed time series into the time series neural network (such as LSTM, GRU or Transformer) to learn the data. The model will infer the future change trend based on the historical data; Forward inference: The neural network performs forward inference and outputs the prediction of key indicators for several future time steps. The key indicators include equipment operation parameters and production efficiency, which are closely related to the health status of the equipment.

[0057] 3. Denormalization and uncertainty estimation: Inverse normalization operation: Convert the predicted results output by the neural network from the standardized or normalized state back to the original dimension, facilitating actual analysis and application. Uncertainty estimation: Estimate the confidence level of the predicted results through the output uncertainty of the neural network (such as through Bayesian neural network or Monte Carlo dropout method). This uncertainty value can help evaluate the reliability of the model prediction.

[0058] 4. Confidence interval calculation: Calculate the confidence interval: Calculate the confidence interval of the predicted results based on the uncertainty estimation of the model output; the confidence interval is used to represent the range of variation of the predicted values, thus providing a more guiding reference for future risk prediction.

[0059] 5. Fault risk classification and warning: Compare with thresholds: Compare the predicted key indicators with the preset soft and hard thresholds. The soft threshold is usually used to send warning signals to alert the occurrence of faults, while the hard threshold is used to trigger mandatory measures to prevent serious faults in the system.

[0060] Compare the probability of the confidence interval: Compare the probability value of the confidence interval with the preset threshold probability to further improve the accuracy and reliability of the prediction.

[0061] Fault risk assessment: Based on the above comparison results, classify and evaluate the fault risks of each node. The fault risks can be divided into low risk, medium risk, and high risk levels, facilitating the adoption of different treatment measures.

[0062] Generate warning labels: Generate corresponding warning labels according to the risk assessment results. The warning labels can help operators or system managers take measures in advance to prevent equipment or system failures.

[0063] Furthermore, the working process of the protection response module includes: Receive feedback content and divide the current and future states into four levels of risk: normal, warning, fault, and emergency according to the preset hard and soft thresholds; if it is at the warning level, trigger a warning through audible and visual alarms and operator notifications; if it is at the fault level, reduce the load to relieve the pressure; if it is at the emergency level, immediately cut off the power supply and shut down slowly.

[0064] In some embodiments, the specific execution process of the above protection response module is as follows: 1. Input and preprocessing: Collect feedback data: The source is the real-time status reports from sensors, control units, or upper-level monitoring systems, and the content is key indicators such as voltage, current, temperature, and vibration.

[0065] Data verification: Check data integrity and timestamp validity. If packet loss or timeout is detected, trigger the "data anomaly" flag and enter the fault diagnosis sub - process, which includes: If data is lost or verification fails, enter the diagnostic mode; Check the communication link, sensor self - test, and module self - test in sequence; If the self - test passes, resume the normal process; otherwise, trigger the fault level response, perform load reduction, and notify maintenance.

[0066] 2. Risk level classification: Preset threshold definition: Hard threshold (Hard), an uncrossable safety boundary, once touched, it is regarded as urgent; Soft threshold (Soft), the trigger point for warning and load reduction, used to identify warning or fault states.

[0067] Compare the currently collected value and the predicted future trend value (based on a sliding window or model) with the hard / soft thresholds respectively.

[0068] 3. Hierarchical response strategy: Warning level: The trigger method is that the soft threshold is about to be breached; The response actions are audible and visual alarms and operator notifications. The buzzer / warning light of the audible and visual alarm flashes, and the frequency is proportional to the degree of over - limit; The operator notification pushes messages to the SCADA / mobile terminal, along with the exceeded parameters and trend curves.

[0069] Continue to detect the trend. If the over - limit risk decreases and remains within the normal range continuously, the warning is automatically cleared; otherwise, enter the next level.

[0070] Fault level: The trigger method is that the soft threshold is actually touched, but the hard threshold has not been touched; The response actions include dynamic load reduction, logging, and work order generation; The dynamic load reduction reduces the system load according to a preset strategy (such as stepped power reduction or priority - based load transfer); The logging records the fault occurrence time, parameters, and the load reduction amplitude that has been executed in the local and remote logs; The work order generation automatically creates a work order in the maintenance system and notifies the maintenance personnel to inspect the equipment on - site.

[0071] If the indicators return to normal after load reduction, reset the status and notify the recovery; If it continues to deteriorate, upgrade to the emergency state.

[0072] Emergency level: The trigger method is that the hard threshold is touched or predicted to be touched; The response actions include immediately cutting off the power supply, slow shutdown, and multi - channel alarm; The immediate power cut performs a global power-off through a hardware quick breaker or a software emergency stop command; the slow shutdown is to start a cooling / slowing-down program to ensure a smooth transition of the mechanical and thermal systems and avoid secondary damage; the multi-channel alarm synchronously triggers various means such as audible and visual alarms, text messages / phone calls, and emails to notify all relevant personnel.

[0073] The system enters the standby safety mode and performs an autonomous or manual reset process; records a complete event log and generates a detailed accident report.

[0074] 4. Common functions and interfaces: Status write-back: After each risk determination and action execution, the new status is written back to the monitoring platform for upper-layer scheduling decisions.

[0075] Heartbeat detection: At the normal / warning level, send an "alive" signal at regular intervals (such as 1 s) to ensure the health of the module; Configurability: Thresholds, load reduction strategies, alarm level mappings, etc. are dynamically issued through configuration files or remote instructions.

[0076] Safety isolation: The emergency cut-off path is physically isolated from the daily control path to prevent accidental mis-triggering.

[0077] The system is built-in with a redundancy mechanism, including: Real-time record and update key configurations and operation logs. After confirming that the fault is repaired, according to the records at the moment before the fault handling is executed, the system is automatically restored to the normal working state.

[0078] An electronic device includes a processor and a memory communicatively connected to the processor and used for storing instructions executable by the processor, and the processor is used to execute the above-mentioned safety monitoring and interlocking protection system for a production line.

[0079] A server includes at least one processor and a memory communicatively connected to the processor, and the memory stores instructions executable by the at least one processor. The instructions are executed by the processor to enable the at least one processor to execute a safety monitoring and interlocking protection system for a production line.

[0080] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements a safety monitoring and interlocking protection system for a production line.

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the present embodiments, and they should all be covered by the scope of the claims and the description of the present invention.

[0082] The foregoing is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A safety monitoring and interlocking protection system for a production line, characterized in that, Including: A data acquisition module, which is configured to collect and real-time update various data in the production process through sensors at each node of the production line, and integrate the various data into a production data set; An anomaly detection module, which is configured to train an IF model using the production data set, use the result generated by the IF model as an adjustment parameter for DBSCAN, complete the anomaly detection of each part of the production line, and feed the detection result back to the fault prediction module and the protection response module; A fault prediction module, which responds to the feedback of the anomaly detection module and, based on the production data set corresponding to the feedback result, performs fault prediction on each part of the production line through a built-in time series neural network; A protection response module, which responds to the feedback of the anomaly detection module and the fault prediction module and, based on the interlock protection mechanism, performs corresponding fault handling on the corresponding part of the production line.

2. The safety monitoring and interlocking protection system for a production line according to claim 1, wherein Before integrating the various data into a production data set, preprocessing operations are performed on the various data, including: Using the forward filling method to handle missing values in the data; reducing data noise for high-frequency vibration signals by extracting their frequency domain features; summarizing the data of each node sensor based on the finest-grained time step, and aligning the various data according to the time step.

3. The safety monitoring and interlock protection system for a production line according to claim 1, characterized in that, The working process of the anomaly detection module includes: Offline training of the IF model based on the historical production data set, using the IF model to infer and calculate the anomaly score for the current production data set, transforming and mapping the anomaly score into a distance metric between samples, and calculating the distance matrix based on the weighted distance metric.

4. The safety monitoring and interlocking protection system for a production line according to claim 3, characterized in that, The working process of the anomaly detection module further includes: Based on the weighted calculated distance matrix, using the DBSCAN clustering algorithm to cluster the production data set, identifying and determining the anomaly clusters and isolated noise points in the production data set, and obtaining the anomaly information.

5. The safety monitoring and interlock protection system for a production line according to claim 1, characterized in that, The working process of the fault prediction module includes: Synchronizing the production data set with the anomaly detection result, and constructing an input sequence within a time window; Using the time series neural network to perform forward inference on the input sequence, and outputting the prediction of key indicators for several future time steps; Denormalizing the result of the key indicator prediction, and calculating the confidence interval according to the uncertainty estimate output by the model; Through comparison with preset soft and hard thresholds and the confidence interval probability, grading and evaluating the future fault risks of each node and generating corresponding warning labels.

6. The safety monitoring and interlock protection system for a production line according to claim 1, wherein, The working process of the protection response module includes: Receiving the feedback content and, according to the preset hard and soft thresholds, dividing the current and future states into four levels of risk: normal, warning, fault, and emergency; if it is at the warning level, triggering a warning through audible and visual alarms and operator notifications; if it is at the fault level, reducing the load to relieve the pressure; if it is at the emergency level, immediately cutting off the power supply and shutting down slowly.

7. The safety monitoring and interlock protection system for a production line according to claim 1, characterized in that The system is built-in with a redundancy mechanism, including: Real-time recording and updating of key configurations and operation logs, and after confirming the fault repair, automatically restoring the system to the normal working state according to the record at the moment before the fault handling is executed.

8. An electronic device, comprising a processor and a memory communicatively connected to the processor and configured to store executable instructions of the processor, wherein: The processor is used to execute the system according to any one of claims 1-7 above.

9. A server, characterized in that: Comprising at least one processor and a memory communicatively connected to the processor, the memory storing instructions executable by the at least one processor, the instructions being executed by the processor to cause the at least one processor to execute the system according to any one of claims 1-7.

10. A computer-readable storage medium stores a computer program, characterized in that: When the computer program is executed by a processor, it implements the system according to any one of claims 1-7.

Images