Troubleshooting system of multi-station automatic production line

Abstract

The invention relates to the technical field of troubleshooting and early warning, in particular to a troubleshooting system for a multi-station automatic production line, which comprises the following modules: intelligent sensor network modules which are deployed at stations and adopt various sensors to monitor key parameters and equipment states of the production line in real time; the fault diagnosis module is used for analyzing the received key parameters and the equipment state, identifying a fault which occurs at present, and determining a fault type and a fault station; the fault source tracing and influence analysis module is used for positioning a fault source by analyzing a propagation path and an influence range of a fault signal after determining a fault type; and the self-adaptive alarm adjusting module is used for dynamically adjusting an alarm strategy according to the emergency degree and property of the fault. According to the invention, automatic troubleshooting is realized, the efficiency is high, the downtime of a production line is reduced, and the fault processing efficiency of a maintenance team is also improved.

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis and early warning technology, and in particular to a fault diagnosis system for a multi-station automated production line. Background Technology

[0002] In today's multi-station automated production lines, timely detection and handling of faults are key factors in ensuring production efficiency and product quality. These production lines often involve complex mechanical equipment, sophisticated electronic control systems, and highly integrated software management systems. Because each station in a multi-station production line is interdependent, a small fault can spread rapidly, affecting the entire production process and leading to production stoppages or even safety accidents.

[0003] Traditional fault diagnosis methods mainly rely on manual step-by-step troubleshooting, which is not only inefficient but also makes it difficult to accurately and quickly locate the source of the fault and assess its impact. As the level of automation and intelligence of production lines continues to improve, single fault diagnosis methods can no longer meet the needs of production lines. There is an urgent need for an intelligent fault diagnosis system that can monitor in real time, diagnose quickly, accurately locate the source of the fault, and assess its impact.

[0004] Furthermore, due to the close interconnectivity of each workstation in the production line, a failure at one workstation may affect multiple workstations downstream or upstream, making the assessment of the impact of the failure more complex. Therefore, an efficient fault diagnosis system not only needs to be able to accurately and quickly diagnose individual faults, but also needs to have the ability to analyze the fault propagation path and scope of impact in order to take timely measures to minimize the impact of the fault on production. Summary of the Invention

[0005] To achieve the above objectives, the present invention provides a fault diagnosis system for multi-station automated production lines.

[0006] A fault diagnosis system for a multi-station automated production line includes the following modules:

[0007] Intelligent sensor network module: Deployed at each workstation, it uses a variety of sensors (such as temperature, pressure, vibration, sound and optical sensors) to monitor key parameters and equipment status of the production line in real time;

[0008] Fault diagnosis module: Analyzes received key parameters and equipment status to identify ongoing faults and determine the fault type and fault location;

[0009] Fault source tracing and impact analysis module: After determining the fault type, the fault source is located by analyzing the propagation path and impact range of the fault signal;

[0010] Adaptive alarm adjustment module: Dynamically adjusts alarm strategies based on the urgency and nature of the fault to ensure that critical faults can be responded to in a timely manner.

[0011] Furthermore, the intelligent sensor network module includes a sensor node deployed at each workstation, and each sensor node includes a temperature sensor, a pressure sensor, a vibration sensor, a sound sensor, and an optical sensor.

[0012] Furthermore, the fault diagnosis module specifically includes:

[0013] Feature extraction: Extracting model features from key parameters and equipment status. These features reflect the operating status and abnormal modes of the equipment. Feature extraction methods include statistical analysis, frequency domain analysis, and time series analysis.

[0014] Pattern recognition: Machine learning algorithms are used to analyze the extracted features and identify patterns that differ from the normal operating state. These different patterns correspond to the corresponding types of faults.

[0015] Fault type determination: By comparing with historical fault data, the identified abnormal patterns are matched with known fault types to determine the specific type of fault that is occurring. For unknown patterns, anomaly detection technology is used for identification and they are marked as new fault types. Based on the acquisition of abnormal data from sensors, the faulty workstation is determined.

[0016] Furthermore, the machine learning algorithm in the pattern recognition is based on the SVM (Support Vector Machine) model, specifically including:

[0017] Data labeling: The extracted feature data is labeled into two categories: normal operation and abnormal operation. This step requires knowledge from domain experts or historical failure data.

[0018] Training the SVM model: The SVM model is trained using a labeled training dataset. The training process includes a kernel function and determining the optimal hyperplane based on an optimization problem, which is expressed as: Where w represents the normal vector of the hyperplane, b is the bias term, and ξ i is the slack variable for the i-th data point, used to handle cases where the data is not perfectly linearly separable. C is the regularization parameter, used to control the balance between the width of the interval and the number of misclassified points.

[0019] Set constraints: y i (w·x i +b)≥1-ξ i ,ξ i ≥0,i=1,…,n, where, y ix is the class label of the i-th data point, which is +1 for the normal state and -1 for the abnormal state. i It is the feature vector of the i-th data point. After the SVM model is trained, it is used to identify whether the new observation data belongs to the normal operation state or the abnormal state. The new data point is classified by the decision function of the SVM model: f(x) = w·x + b. If f(x) ≥ 0, then x is classified as normal state; otherwise, it is classified as abnormal state.

[0020] Furthermore, the fault source tracing and impact analysis module specifically includes:

[0021] Fault signal propagation analysis: Based on the fault type determined by the fault diagnosis module, analyze the propagation path of the signal corresponding to that fault type in the production line;

[0022] Fault impact assessment: Continue to assess the impact range caused by the propagation of the fault signal, including the number of affected workstations and partial shutdown of the production line;

[0023] Fault source location: Combining the analysis of the propagation path and impact range of the fault signal, a path algorithm is used to trace and locate the fault source.

[0024] Furthermore, the fault signal propagation analysis includes:

[0025] Construct a graph model of the production line: Based on the topology and process flow of the production line, construct a graph model G = (V, E), where V is the set of workstations and E is the set of edges connecting the workstations. Each edge is assigned a weight w(e), which represents the strength of the connection or the ease of signal propagation.

[0026] Identify the fault source node: Based on the fault type and corresponding fault station identified by the fault diagnosis module, mark the station as the fault source node in the diagram;

[0027] Analyze the signal propagation path: Use a graph search algorithm to traverse the graph starting from the fault source node to simulate the propagation path of the fault signal.

[0028] Furthermore, the assessment of the fault impact range is based on the signal propagation path, including the number of affected nodes and the degree of impact, summing the shortest path lengths from the fault source node to other nodes, and considering the edge weights:

[0029] D(s, v) = min p∈P(s,v) ∑ e∈p w(e), where D(s,v) is the shortest path distance from the fault source node s to node v, P(s,v) is the set of all paths from s to v, and e is an edge in path p.

[0030] Furthermore, the path algorithm used in the fault source localization employs Dijkstra's algorithm. Dijkstra's algorithm is used to find the shortest path from a node in the graph to all other nodes. Starting from the source node, it gradually expands to the node with the shortest distance until the target node or all nodes in the graph are reached. For each node, the shortest path to that node and the total weight of the path are calculated, as follows:

[0031] Initialize the distances of all nodes to infinity, except for the starting node, whose distance is 0.

[0032] Mark all nodes as unvisited and set the starting node as the current node;

[0033] For the current node, consider all unvisited neighbor nodes and calculate the distance from the starting node to these neighbor nodes. If a shorter path is found, update the distance to the neighbor nodes.

[0034] The current node is marked as visited, and the node with the shortest distance among the unvisited nodes becomes the new current node;

[0035] Repeat the above steps until all nodes have been visited;

[0036] Tracing back from the affected nodes (or multiple nodes) to the fault source node, for each affected node, find the shortest path from it to all other nodes. The fault source is the common starting point or intersection of the shortest paths.

[0037] Furthermore, the specific calculation steps of the Dijkstra algorithm are expressed as follows:

[0038] Set the shortest path estimate of all nodes to infinity, except for the source node, whose value is set to 0, because the shortest distance from the source node to itself is always 0. Create an empty set S to store the nodes that have been processed, since the shortest path of the processed nodes has been determined. Create a priority queue Q containing all nodes, and sort the priority queue according to the shortest path estimate of the nodes.

[0039] The main loop of the algorithm is as follows:

[0040] When the priority queue Q is not empty, take the node i with the shortest distance from Q (initially this will be the source node);

[0041] Add node u to set S to indicate that the shortest path of u has been determined;

[0042] Update all neighboring nodes v of node u. If the path from u to v is shorter than the known path, update the shortest path estimate for v, expressed as:

[0043] dist(v) = mindist(v), dist(u) + weight(u, v), where dist(v) is the current shortest path estimate of node v, dist(u) is the shortest path value of node u (which has been determined), and weight(u, v) is the weight of the edge from u to v.

[0044] Repeat the above process until the priority queue Q is empty, meaning the shortest path for all nodes has been determined.

[0045] Furthermore, the adaptive alarm adjustment module specifically includes:

[0046] Urgency assessment: Assess the urgency of the failure based on the speed and scope of its impact on the production line;

[0047] Alarm priority setting: Each fault is assigned an alarm priority based on its urgency and nature. Higher priority alarms correspond to higher urgency. Priority settings are based on predefined rules.

[0048] The beneficial effects of this invention are:

[0049] This invention utilizes graph theory to analyze the propagation path of fault signals in a multi-station automated production line. Through detailed investigation, the source of the fault can be accurately traced, thereby shortening the time for fault diagnosis and location. This precise location method not only reduces production line downtime but also improves the efficiency of the maintenance team in handling faults, thus ensuring the stable operation of the production line and maximizing production capacity.

[0050] This invention, by comprehensively investigating the path and scope of fault propagation, can fully assess the impact of a single fault event on the entire production line. This comprehensive impact assessment provides production line managers with crucial information, enabling them to optimize the allocation of maintenance resources and manpower based on the urgency and scope of the fault. Especially when resources are limited, it can prioritize handling faults that have the greatest impact on production, thereby minimizing economic losses and maintaining production continuity.

[0051] This invention can dynamically adjust alarm strategies according to the nature and urgency of the fault, ensuring that critical faults can be responded to in a timely manner. This dynamic alarm system not only improves the efficiency and timeliness of fault response, but also reduces false alarms and unnecessary emergency responses through customized alarm information and priority management, thereby increasing the attention and processing speed of production line personnel to truly important faults. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 This is a schematic diagram of the functional modules of the investigation system according to an embodiment of the present invention;

[0054] Figure 2 This is a schematic diagram of the fault source tracing and impact analysis module in an embodiment of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0056] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0057] like Figure 1-2 As shown, a fault diagnosis system for a multi-station automated production line includes the following modules:

[0058] Intelligent sensor network module: Deployed at each workstation, it uses a variety of sensors (such as temperature, pressure, vibration, sound and optical sensors) to monitor key parameters and equipment status of the production line in real time;

[0059] Fault diagnosis module: Analyzes received key parameters and equipment status to identify ongoing faults and determine the fault type and fault location;

[0060] Fault source tracing and impact analysis module: After determining the fault type, the fault source is located by analyzing the propagation path and impact range of the fault signal;

[0061] Adaptive alarm adjustment module: Dynamically adjusts alarm strategies based on the urgency and nature of the fault to ensure that critical faults can be responded to in a timely manner.

[0062] The intelligent sensor network module includes a sensor node deployed at each workstation. Each sensor node comprises a temperature sensor, pressure sensor, vibration sensor, sound sensor, and optical sensor. These sensors are individually assigned and configured to meet the specific monitoring needs of their respective workstations. For example, vibration and sound sensors are used to detect equipment malfunctions, temperature and pressure sensors are used to monitor whether process parameters are within normal ranges, and optical sensors are used to detect abnormal conditions in product quality and on equipment surfaces. This allows for real-time monitoring of key parameters and equipment status across the entire production line. Furthermore, sensor nodes can be customized and optimized according to the characteristics of the production line and monitoring requirements to ensure efficient and accurate data acquisition.

[0063] Key parameters include:

[0064] Temperature: For production lines that require heating or cooling processes, temperature is a critical parameter that can affect product quality and the stability of the production process.

[0065] Pressure: In workstations involving the handling of liquids or gases, pressure monitoring is crucial to ensuring smooth and safe production processes.

[0066] Vibration: Abnormal vibration of equipment is often an early sign of mechanical failure, which is especially important for rotating machinery.

[0067] Sound: Similar to vibration, abnormal sounds can be an early warning of equipment malfunctions, and sound analysis can help diagnose specific types of problems.

[0068] Optical properties: On some production lines, the optical properties of a product (such as color and brightness) are important indicators for quality control.

[0069] Current / Voltage: The current and voltage parameters of electrical equipment can reflect the operating status of the equipment. Abnormal current or voltage indicates equipment failure.

[0070] Flow rate: In liquid processing production lines, flow rate is one of the key parameters, as it affects the manufacturing process and quality of the product.

[0071] Equipment status includes:

[0072] Operating time: The cumulative operating time of equipment plays an important role in predicting maintenance needs and evaluating equipment performance.

[0073] On / Off Status: The on / off status of the equipment can help monitor the operation of the production line and detect downtime or malfunctions in a timely manner.

[0074] Error codes: Modern production equipment typically generates error codes when malfunctions occur. These codes help to quickly diagnose the problem.

[0075] Production count: The number of production tasks completed by the equipment, which is crucial for monitoring production efficiency and equipment performance.

[0076] Maintenance records: Equipment maintenance and troubleshooting records can provide important information about its reliability and potential problems.

[0077] The fault diagnosis module specifically includes:

[0078] Feature extraction: Extracting model features from key parameters and equipment status. These features reflect the operating status and abnormal modes of the equipment. Feature extraction methods include statistical analysis, frequency domain analysis, and time series analysis.

[0079] Pattern recognition: Machine learning algorithms are used to analyze the extracted features and identify patterns that differ from the normal operating state. These different patterns correspond to the corresponding types of faults.

[0080] Fault type determination: By comparing with historical fault data, the identified abnormal patterns are matched with known fault types to determine the specific type of the fault that is occurring. For unknown patterns, anomaly detection technology is used to identify them and mark them as new fault types. Based on the acquisition of abnormal data from sensors, the faulty workstation is determined.

[0081] The diagnostic results will be fed back to the corresponding alarm response knowledge base and updated to continuously optimize the accuracy and efficiency of fault diagnosis.

[0082] Statistical analysis mainly involves calculating the basic statistical properties of data in order to capture the central trend, distribution pattern and fluctuation characteristics of the data. Statistical features include mean, standard deviation, skewness and kurtosis.

[0083] 1. Mean: The average value of all data points in the dataset, calculated using the following formula:

[0084] Where x i is the i-th observation, N is the total number of observations, and the average value can reflect the central location of the data.

[0085] 2. Standard deviation: Measures the dispersion of values ​​in a dataset. The formula is:

[0086] The larger the standard deviation, the greater the volatility of the data.

[0087] 3. Skewness: Describes the asymmetry in the distribution of data. The calculation formula is:

[0088] A positive skewness indicates that the data distribution is right-skewed, while a negative skewness indicates left-skewed.

[0089] 4. Kurtosis: Reflects the sharpness of the data distribution; the formula is as follows: The larger the kurtosis, the sharper the distribution.

[0090] In this invention, statistical analysis of data from various sensors can extract statistical features that reflect the equipment status and operating characteristics, which can be used for preliminary fault detection and classification.

[0091] Frequency domain analysis transforms a time-domain signal into the frequency domain to analyze its frequency components, using the Fourier transform (FFT).

[0092] 1. Fast Fourier Transform (FFT): An algorithm that converts a time-domain signal into a frequency-domain signal. The calculation formula is as follows: Where x(n) is the time-domain signal and X(k) is the frequency-domain signal.

[0093] In this invention, by performing FFT on sensor data, the main frequency components and their amplitudes during equipment operation can be identified. This information is crucial for identifying specific types of faults (mechanical vibration, bearing damage).

[0094] Time series analysis is used to analyze data characteristics that change over time, focusing on time dependencies and periodic patterns in the data. This is particularly important for detecting periodic failures or time-based abnormal behavior in production lines.

[0095] 1. Autocorrelation Analysis: Autocorrelation analysis helps identify recurring patterns in data (such as periodic variations), which is very useful for identifying specific types of periodic faults. The calculation formula is:

[0096] Where R(k) is the autocorrelation coefficient with a delay of k, n is the total number of data points, and σ 2 It is the variance of the data, x t It is the observation value at time t. It is the average value of the data.

[0097] 2. Time Series Decomposition: Time series data can be decomposed into trend, seasonal, and random components. For multi-station automated production lines, this decomposition helps identify failure modes associated with specific stages of the production process. For example, the seasonal component may reflect the performance of the production line at different shifts or during specific time periods.

[0098] 3. Time series forecasting models: such as the Autoregressive Moving Average (ARIMA) model, can be used to predict the system performance over a future period, thereby identifying trends that may deviate from normal operating modes and providing a basis for fault warning.

[0099] In conjunction with this invention, time-series analysis can be used to capture the dynamic characteristics of the operating status of equipment at each workstation on a production line as a function of time, identifying failure modes that may be caused by equipment wear, insufficient periodic maintenance, or other reasons. By monitoring the changes in these time-dependent characteristics in real time, the fault diagnosis module can promptly detect and warn of potential problems that may affect the normal operation of the production line, providing maintenance personnel with accurate fault diagnosis information, thereby ensuring the efficient and stable operation of the production line.

[0100] Machine learning algorithms in pattern recognition are based on the SVM (Support Vector Machine) model, specifically including:

[0101] Data labeling: The extracted feature data is labeled into two categories: normal operation and abnormal operation. This step requires knowledge from domain experts or historical failure data.

[0102] Training the SVM model: The SVM model is trained using a labeled training dataset. The training process includes a kernel function and determining the optimal hyperplane based on an optimization problem, which is expressed as: Where w represents the normal vector of the hyperplane, b is the bias term, and ξ i is the slack variable for the i-th data point, used to handle cases where the data is not perfectly linearly separable. C is the regularization parameter, used to control the balance between the width of the interval and the number of misclassified points.

[0103] Set constraints: y i (w·x i +b)≥1-ξ i ,ξ i ≥0,i=1,…,n, where, y i x is the class label of the i-th data point, which is +1 for the normal state and -1 for the abnormal state. i It is the feature vector of the i-th data point. After the SVM model is trained, it is used to identify whether the new observation data belongs to the normal operation state or the abnormal state. The new data point is classified by the decision function of the SVM model: f(x) = w·x + b. If f(x) ≥ 0, then x is classified as normal state; otherwise, it is classified as abnormal state.

[0104] The fault source tracing and impact analysis module specifically includes:

[0105] Fault signal propagation analysis: Based on the fault type determined by the fault diagnosis module, analyze the propagation path of the signal corresponding to that fault type in the production line;

[0106] Fault impact assessment: Continue to assess the impact range caused by the propagation of the fault signal, including the number of affected workstations and partial shutdown of the production line;

[0107] Fault source location: Combining the analysis of the propagation path and impact range of the fault signal, a path algorithm is used to trace and locate the fault source;

[0108] Once the source of the fault is located, the system will provide maintenance personnel with detailed fault information, including the fault type, location, suggested handling measures and prevention strategies. At the same time, the fault handling results and experience will be fed back into the system's knowledge base to optimize future fault diagnosis and prevention.

[0109] Fault signal propagation analysis includes:

[0110] Construct a graph model of the production line: Based on the topology and process flow of the production line, construct a graph model G = (V, E), where V is the set of workstations and E is the set of edges connecting the workstations. Each edge is assigned a weight w(e), which represents the strength of the connection or the ease of signal propagation.

[0111] Identify the fault source node: Based on the fault type and corresponding fault station identified by the fault diagnosis module, mark the station as the fault source node in the diagram;

[0112] Analyze the signal propagation path: Use a graph search algorithm to traverse the graph starting from the fault source node to simulate the propagation path of the fault signal. In this process, the physical distance between nodes, logical dependencies, and possible shielding or interference factors can be considered.

[0113] The assessment of the fault impact range is based on the signal propagation path, including the number of affected nodes and the degree of impact. It involves summing the shortest path lengths from the fault source node to other nodes, taking edge weights into account.

[0114] D(s, v) = min p∈P(s,v) ∑ e∈p w(e), where D(s,v) is the shortest path distance from the fault source node s to node v, P(s,v) is the set of all paths from s to v, and e is an edge in path p.

[0115] Impact Scope Assessment:

[0116] Number of affected workstations: Through fault propagation analysis, the number of all affected nodes (workstations) is counted, showing the widespread impact of the fault on the production line.

[0117] Partial production line downtime: For each affected node, assess its role and importance in the production line, and the specific impact of the downtime on production. Factors to consider include:

[0118] Criticality of workstations: Some workstations may be critical to the production process, and their failure may cause the entire production line to stop.

[0119] Failure duration: Assess the time required for fault handling and repair, and the operational status of the production line during this period.

[0120] Alternative solutions and response strategies: Consider whether there are backup systems or temporary solutions that can maintain some production during the repair of a faulty workstation.

[0121] In the multi-station automated production line fault diagnosis system of the present invention, by constructing a graphical model of the production line and analyzing the propagation path of fault signals, the system can accurately identify the scope of influence and potential points of influence of the fault. This analysis not only depends on the physical layout of the production line, but also considers the logical dependencies between workstations and the propagation characteristics of fault signals in these connections. Using this method, maintenance personnel can be provided with a detailed view of fault propagation, helping them to locate the fault source more quickly and take corresponding corrective measures, ultimately reducing production interruptions and improving the stability of the production line.

[0122] The path algorithm used in fault source localization is Dijkstra's algorithm. Dijkstra's algorithm is used to find the shortest path from a node in a graph to all other nodes. Starting from the source node, it gradually expands to the node with the shortest distance until the target node or all nodes in the graph are reached. For each node, the shortest path to that node and the total weight of the path are calculated, as follows:

[0123] Initialize the distances of all nodes to infinity, except for the starting node, whose distance is 0.

[0124] Mark all nodes as unvisited and set the starting node as the current node;

[0125] For the current node, consider all unvisited neighbor nodes and calculate the distance from the starting node to these neighbor nodes. If a shorter path is found, update the distance to the neighbor nodes.

[0126] The current node is marked as visited, and the node with the shortest distance among the unvisited nodes becomes the new current node;

[0127] Repeat the above steps until all nodes have been visited;

[0128] Tracing back from the affected nodes (or multiple nodes) to the fault source node, for each affected node, find the shortest path from it to all other nodes. The fault source is the common starting point or intersection of the shortest paths.

[0129] The specific calculation steps of Dijkstra's algorithm are as follows:

[0130] Set the shortest path estimate of all nodes to infinity, except for the source node, whose value is set to 0, because the shortest distance from the source node to itself is always 0. Create an empty set S to store the nodes that have been processed, since the shortest path of the processed nodes has been determined. Create a priority queue Q containing all nodes, and sort the priority queue according to the shortest path estimate of the nodes.

[0131] The main loop of the algorithm is as follows:

[0132] When the priority queue Q is not empty, take the node u with the shortest distance from Q (initially this will be the source node);

[0133] Add node u to set S to indicate that the shortest path of u has been determined;

[0134] Update all neighboring nodes v of node u. If the path from u to v is shorter than the known path, update the shortest path estimate for v, expressed as:

[0135] dist(v) = min(dist(v), dist(u) + weight(u,v)), where dist(v) is the current shortest path estimate of node v, dist(u) is the shortest path value of node u (which has been determined), and weight(u,v) is the weight of the edge from u to v.

[0136] Repeat the above process until the priority queue Q is empty, meaning the shortest path for all nodes has been determined.

[0137] Dijkstra's algorithm starts from the source node and gradually expands to the shortest distance node in a "greedy" manner, ensuring that each step is a local optimum. In each iteration, the algorithm selects an unprocessed shortest distance node u and considers the path from u to its neighbor node v. If a shorter path is found, the shortest path estimate to v is updated. The algorithm uses a priority queue to efficiently select the next shortest distance node to process, and the priority queue is sorted according to the shortest path estimate of the nodes.

[0138] For each affected node, Dijkstra's algorithm is used to calculate the shortest path from it to all other nodes in the graph (including other affected nodes and unaffected nodes). This process will reveal the possible propagation paths and directions of the fault signal.

[0139] If only one node is affected, then we can start from that node and use Dijkstra's algorithm to find the shortest path to that node in reverse. The source of the failure is the starting point of these paths or an edge with an abnormally high weight in the path.

[0140] When multiple nodes are affected, Dijkstra's algorithm is executed on each affected node. The shortest path results for each node are then compared to find common paths or nodes. These commonalities represent the source or propagation path of the fault. Specifically, the following can be looked for:

[0141] The intersection of the shortest path trees of all affected nodes (generated by Dijkstra's algorithm).

[0142] Sudden increases in weight or path forks in the shortest path are where failures occur.

[0143] The adaptive alarm adjustment module specifically includes:

[0144] Urgency assessment: Assess the urgency of a failure based on the speed and extent of its impact on the production line. For example, a failure whose impact spreads rapidly to multiple workstations is more urgent than one that only affects a single workstation.

[0145] Alarm priority setting: Each fault is assigned an alarm priority based on its urgency and nature. Higher priority alarms correspond to higher urgency. Priority settings are based on predefined rules.

[0146] Establish an alarm feedback mechanism that allows recipients to respond to and provide feedback on alarms. This can be achieved through a response system, such as confirming receipt of an alarm and reporting the status of fault handling. The alarm system should be dynamically adjusted based on feedback. For example, if the fault is not responded to within the expected time, the alarm can be escalated or higher-level personnel can be notified.

[0147] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the invention is limited to these examples; within the framework of the invention, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of the different aspects of the invention as described above, which are not provided in detail for the sake of brevity.

[0148] This invention is intended to cover all such substitutions, modifications, and variations falling within the broad scope of the claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A fault diagnosis system for a multi-station automated production line, characterized in that, Includes the following modules: Intelligent sensor network module: Deployed at each workstation, it uses multiple sensors to monitor key parameters and equipment status of the production line in real time; Fault diagnosis module: Analyzes received key parameters and equipment status to identify ongoing faults and determine the fault type and fault location; Fault source tracing and impact analysis module: After determining the fault type, the fault source is located by analyzing the propagation path and impact range of the fault signal; Adaptive alarm adjustment module: Dynamically adjusts alarm strategies based on the urgency and nature of the fault to ensure that critical faults can be responded to in a timely manner.

2. The fault diagnosis system for a multi-station automated production line according to claim 1, characterized in that, The intelligent sensor network module includes a sensor node deployed at each workstation, and each sensor node includes a temperature sensor, a pressure sensor, a vibration sensor, a sound sensor, and an optical sensor.

3. The fault diagnosis system for a multi-station automated production line according to claim 2, characterized in that, The fault diagnosis module specifically includes: Feature extraction: Extracting model features from key parameters and equipment status. These features reflect the operating status and abnormal modes of the equipment. Feature extraction methods include statistical analysis, frequency domain analysis, and time series analysis. Pattern recognition: Machine learning algorithms are used to analyze the extracted features and identify patterns that differ from the normal operating state. These different patterns correspond to the corresponding types of faults. Fault type determination: By comparing with historical fault data, the identified abnormal patterns are matched with known fault types to determine the specific type of fault that is occurring. For unknown patterns, anomaly detection technology is used for identification and they are marked as new fault types. Based on the acquisition of abnormal data from sensors, the faulty workstation is determined.

4. The fault diagnosis system for a multi-station automated production line according to claim 3, characterized in that, The machine learning algorithm in the pattern recognition is based on the SVM model and specifically includes: Data labeling: The extracted feature data is labeled into two categories: normal operating status and abnormal status; Training the SVM model: The SVM model is trained using a labeled training dataset. The training process includes a kernel function and determining the optimal hyperplane based on an optimization problem, which is expressed as: Where w represents the normal vector of the hyperplane, b is the bias term, and ξ i is the slack variable for the i-th data point, used to handle the case of incomplete linear separability, and C is the regularization parameter, used to control the balance between the width of the interval and the number of misclassified points; Set constraints: y i (w·x i +b)≥1-ξ i ,ξ i ≥0,i=1,…,n, where, y i x is the class label of the i-th data point, which is +1 for the normal state and -1 for the abnormal state. i It is the feature vector of the i-th data point. After the SVM model is trained, it is used to identify whether the new observation data belongs to the normal operation state or the abnormal state. The new data point is classified by the decision function of the SVM model: f(x) = w·x + b. If f(x) ≥ 0, then x is classified as normal state; otherwise, it is classified as abnormal state.

5. The fault diagnosis system for a multi-station automated production line according to claim 1, characterized in that, The fault source tracing and impact analysis module specifically includes: Fault signal propagation analysis: Based on the fault type determined by the fault diagnosis module, analyze the propagation path of the signal corresponding to that fault type in the production line; Fault impact assessment: Continue to assess the impact range caused by the propagation of the fault signal, including the number of affected workstations and partial shutdown of the production line; Fault source location: Combining the analysis of the propagation path and impact range of the fault signal, a path algorithm is used to trace and locate the fault source.

6. The fault diagnosis system for a multi-station automated production line according to claim 5, characterized in that, The fault signal propagation analysis includes: Construct a graph model of the production line: Based on the topology and process flow of the production line, construct a graph model G = (V, E), where V is the set of workstations and E is the set of edges connecting the workstations. Each edge is assigned a weight w(e), which represents the strength of the connection or the ease of signal propagation. Identify the fault source node: Based on the fault type and corresponding fault station identified by the fault diagnosis module, mark the station as the fault source node in the diagram; Analyze the signal propagation path: Use a graph search algorithm to traverse the graph starting from the fault source node to simulate the propagation path of the fault signal.

7. The fault diagnosis system for a multi-station automated production line according to claim 6, characterized in that, The assessment of the fault impact range is based on the signal propagation path, including the number of affected nodes and the degree of impact, summing the shortest path lengths from the fault source node to other nodes, and considering the edge weights: D(s, v) = min p∈P(s，v) ∑ e∈p w(e), where D(s,v) is the shortest path distance from the fault source node s to node v, P(s,) is the set of all paths from s to v, and e is an edge in path p.

8. The fault diagnosis system for a multi-station automated production line according to claim 7, characterized in that, The path algorithm used in fault source localization employs Dijkstra's algorithm. Dijkstra's algorithm is used to find the shortest path from a node in a graph to all other nodes. Starting from the source node, it gradually expands to the node with the shortest distance until the target node or all nodes in the graph are reached. For each node, the shortest path to that node and the total weight of the path are calculated, as follows: Initialize the distance of all nodes to infinity, except for the starting node, whose distance is 0; Mark all nodes as unvisited and set the starting node as the current node; For the current node, consider all unvisited neighbor nodes and calculate the distance from the starting node to these neighbor nodes. If a shorter path is found, update the distance to the neighbor nodes. The current node is marked as visited, and the node with the shortest distance among the unvisited nodes becomes the new current node; Repeat the above steps until all nodes have been visited; Tracing back from the affected nodes to the fault source node, for each affected node, find the shortest path from it to all other nodes. The fault source is the common starting point or intersection of the shortest paths.

9. The fault diagnosis system for a multi-station automated production line according to claim 8, characterized in that, The specific calculation steps of the Dijkstra algorithm are as follows: Set the shortest path estimate of all nodes to infinity, except for the source node, whose value is set to 0, because the shortest distance from the source node to itself is always 0. Create an empty set S to store the nodes that have been processed, since the shortest path of the processed nodes has been determined. Create a priority queue Q containing all nodes, and sort the priority queue according to the shortest path estimate of the nodes. The main loop of the algorithm is as follows: When priority queue Q is not empty, take the node u with the shortest distance from Q; Add node u to set S to indicate that the shortest path of u has been determined; Update all neighboring nodes v of node u. If the path from u to v is shorter than the known path, update the shortest path estimate for v, expressed as: dist(v) = min(dist(v), dist(u) + weight(u, v)), where dist(v) is the current shortest path estimate of node v, dist(u) is the shortest path value of node u, and weight(u, v) is the weight of the edge from u to v. Repeat the above process until the priority queue Q is empty, meaning the shortest path for all nodes has been determined.

10. The fault diagnosis system for a multi-station automated production line according to claim 8, characterized in that, The adaptive alarm adjustment module specifically includes: Urgency assessment: Assess the urgency of the failure based on the speed and scope of its impact on the production line; Alarm priority setting: Each fault is assigned an alarm priority based on its urgency and nature. Higher priority alarms correspond to higher urgency. Priority settings are based on predefined rules.

Images