A method and system for detecting temperature anomalies in key equipment of belt conveyors
By constructing a measurement point association graph and feature fusion, and utilizing one-dimensional convolutional neural networks and graph convolutional networks, the problem of early identification of equipment failures in traditional belt conveyor temperature monitoring methods has been solved, achieving early and accurate temperature anomaly warning.
Patent Information
- Application Number
- CN202611133490.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-29
- Publication Date
- 2026-11-17
AI Technical Summary
Traditional temperature monitoring methods for belt conveyors are difficult to identify equipment faults in the early stages. Temperature exceeding the limit at a single measuring point often occurs in the late stages of a fault. The changes in the synergistic relationship between multiple measuring points are not fully utilized, and there is a lack of a benchmark for the normal thermal state of the equipment under fault-free conditions.
By constructing a measurement point association map, performing dual-path feature extraction and fusion, and utilizing one-dimensional convolutional neural networks and graph convolutional networks to extract temperature features, the error threshold discrimination is reconstructed to achieve early and accurate temperature anomaly warning.
By establishing a normal thermal state benchmark under fault-free conditions, sensing local temperature rise deviations and imbalances in the coordination relationship of measurement points, false alarms and missed alarms are reduced, and early and accurate temperature anomaly warnings are provided.
Smart Images

Figure CN123111504A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of equipment condition monitoring technology, and more specifically, to a method and system for detecting temperature anomalies in key equipment of belt conveyors. Background Technology
[0002] Belt conveyors are the core equipment for open-pit coal and mineral material transportation. Their key components (such as drive motors, reducers, hydraulic couplers, drive drums, redirecting drums, idler rollers, and tensioning devices) operate under long-term heavy loads, high dust levels, and large temperature differences. Abnormal temperatures are an important precursor to equipment failure.
[0003] Traditional temperature monitoring methods mainly rely on fixed threshold alarms at single measuring points, which have the following drawbacks: First, temperature exceedances at a single measuring point often occur in the late stage of a fault, making it difficult to identify early temperature rise anomalies in a timely manner; second, the changes in the synergistic relationship between multiple measuring points (such as the inherent correlation of temperature change trends at various measuring points under normal operating conditions) are not fully utilized, leading to missed alarms and delayed alarms under complex operating conditions; and third, there is a lack of effective means to establish a baseline for the normal thermal state of equipment without a large number of fault tags. Summary of the Invention
[0004] This invention aims to overcome the shortcomings of the prior art and provide a method and system for detecting temperature anomalies in key equipment of belt conveyors. By constructing a measurement point association map, extracting and fusing dual-path features, and reconstructing error threshold discrimination, it achieves early and accurate temperature anomaly warning.
[0005] In a first aspect, the present invention provides a method for detecting temperature anomalies in key equipment of a belt conveyor, the method comprising: Time-series data from multiple temperature measurement points of key equipment in a belt conveyor are collected, and the time-series data is preprocessed. Based on the preprocessed time series data, a temperature sample set containing temperature values from multiple temperature measurement points is constructed, and a correlation graph representing the relationship between measurement points is constructed based on the correlation of temperature change trends between different temperature measurement points. Extract the single-point temperature change features of each temperature measuring point in the temperature sample set, and extract the collaborative relationship features between multiple measuring points in the correlation graph; The single-point temperature change features are fused with the cooperative relationship features, and the normal thermal state pattern is reconstructed based on the fused features to obtain a reconstructed sample. Based on the reconstruction error between the original temperature sample and the reconstructed sample, an anomaly score is determined, and an anomaly discrimination threshold is established based on the reconstruction error distribution under normal operating conditions. Based on the comparison between the anomaly score and the anomaly discrimination threshold, the thermal state anomalies of key equipment are identified and an early warning is output.
[0006] Preferably, the construction includes a temperature sample set containing temperature values from multiple temperature measuring points, and based on the correlation of temperature change trends among different temperature measuring points, a correlation diagram characterizing the relationship between the measuring points is constructed, including: Based on the sliding time window, a multi-point temperature sample is constructed from the preprocessed time series data, and each temperature point is used as a graph node. Within the sliding time window, calculate the Pearson correlation coefficient of the temperature change sequence between any two temperature measurement points; When the Pearson correlation coefficient is greater than a preset similarity threshold, an edge is established between the graph nodes corresponding to the two temperature measurement points to form the association graph.
[0007] Preferably, the extraction of single-point temperature change characteristics of each temperature measuring point in the temperature sample set includes: The temperature sample set is input into a one-dimensional convolutional neural network to extract the local temperature rise change pattern of each temperature measurement point along the time dimension. The local temperature rise change pattern is time-dependently encoded by a gated loop unit, and the temperature change characteristics of the single measurement point are output.
[0008] Preferably, the extraction of collaborative relationship features among multiple measurement points in the association graph includes: The adjacency matrix of the association graph is input into the temperature sample set and then convolved into a network. The graph convolutional network aggregates the temperature state information of adjacent temperature measurement points along the graph structure and outputs the cooperative relationship feature.
[0009] Preferably, the step of fusing the single-point temperature change features with the cooperative relationship features, and reconstructing the normal thermal state pattern based on the fused features to obtain the reconstructed sample includes: The single-point temperature change feature and the collaborative relationship feature are concatenated along the feature dimension to form a fused feature vector; The fused feature vector is input into a decoding and reconstruction network consisting of a fully connected layer and an upsampling layer to generate a decoding and reconstruction sample with the same dimension as the original temperature sample, which is used as the reconstruction sample. Using a temperature sample set under normal operating conditions as training data, the mean square error between the original temperature sample and the decoded and reconstructed sample is minimized, so that the decoding and reconstruction network converges to the normal thermal state mode of the key equipment.
[0010] Preferably, the step of determining anomaly scores based on the reconstruction error between the original temperature sample and the reconstructed sample, and establishing anomaly discrimination thresholds based on the reconstruction error distribution under normal operating conditions, includes: Calculate the absolute error of each original temperature sample and its corresponding reconstructed sample at each temperature measurement point, and sum or calculate the root mean square of the absolute errors of all measurement points as the anomaly score of the sample. Collect abnormal scores from multiple samples under normal operating conditions and calculate their mean and standard deviation; The anomaly detection threshold is set to the mean plus three times the standard deviation.
[0011] Preferably, it further includes: The real-time temperature time series data collected from multiple measurement points is processed sequentially through preprocessing, temperature sample set construction, correlation graph construction, single measurement point temperature change feature extraction, collaborative relationship feature extraction, feature fusion and reconstruction to obtain a real-time anomaly score; When the real-time anomaly score exceeds the anomaly discrimination threshold, it is determined that the key equipment has a thermal anomaly, and an early warning result containing the location information of the anomaly measurement point and the direction of temperature rise offset is output.
[0012] Secondly, the present invention provides a temperature anomaly detection system for key equipment of belt conveyors, comprising: The data preprocessing module is used to collect time-series data from multiple temperature measurement points of key equipment of the belt conveyor and to preprocess the time-series data. The correlation graph construction module is used to construct a temperature sample set containing temperature values from multiple temperature measurement points based on the preprocessed time series data, and to construct a correlation graph representing the correlation between measurement points based on the correlation of temperature change trends between different temperature measurement points. The feature extraction module is used to extract the single-point temperature change features of each temperature measuring point in the temperature sample set, and to extract the collaborative relationship features between multiple measuring points in the correlation graph. The reconstruction module is used to fuse the temperature change features of the single measurement point with the cooperative relationship features, and reconstruct the normal thermal state mode based on the fused features to obtain a reconstructed sample. The anomaly detection threshold establishment module is used to determine the anomaly score based on the reconstruction error between the original temperature sample and the reconstructed sample, and to establish the anomaly detection threshold based on the reconstruction error distribution under normal operating conditions. The identification module is used to identify thermal state anomalies of key equipment and output early warnings based on the comparison result between the anomaly score and the anomaly discrimination threshold.
[0013] Thirdly, the present invention provides a readable medium including executable instructions, which, when executed by a processor of an electronic device, cause the electronic device to perform any of the methods described in the first aspect.
[0014] Fourthly, the present invention provides an electronic device including a processor and a memory storing execution instructions, wherein when the processor executes the execution instructions stored in the memory, the processor performs the method as described in any of the first aspects.
[0015] This invention provides a method and system for detecting temperature anomalies in key equipment of belt conveyors. It elevates multi-point temperature monitoring from judging single-point exceedances to identifying overall thermal anomalies through the perception of interconnected measurement points. The method employs a technical approach of "temperature sample set construction—association graph construction—dual-path feature extraction and fusion—reconstruction error threshold discrimination," simultaneously sensing local temperature rise deviations and imbalances in the collaborative relationships between measurement points. A sliding time window and Pearson correlation coefficient are used to dynamically construct the association graph, enabling adaptation to the thermal coupling structures of different key equipment. A one-dimensional convolutional neural network and gated recurrent units are used to extract temperature rise change features at single measurement points, while a graph convolutional network is used to extract collaborative relationship features across multiple measurement points. After fusion, the normal thermal state pattern is learned through a decoding and reconstruction network, establishing a baseline without requiring numerous fault labels. A threshold is set based on the mean and three standard deviations of the reconstruction error, effectively distinguishing normal temperature fluctuations from true anomaly deviations, reducing false alarms and missed alarms. Ultimately, in real-time detection, it can output abnormal measurement point location information and temperature rise offset direction, providing early, accurate, and deployable temperature anomaly warnings for key equipment of belt conveyors in open-pit coal mines, significantly improving the intelligence level and operational reliability of equipment condition monitoring.
[0016] The further effects of the aforementioned non-conventional preferred method will be explained below in conjunction with specific embodiments. Attached Figure Description
[0017] To more clearly illustrate the embodiments of the present invention or the existing technical solutions, 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 some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This is a schematic diagram of a temperature anomaly detection method for key equipment of a belt conveyor provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of another method for detecting temperature anomalies in key equipment of a belt conveyor, provided by an embodiment of the present invention. Figure 3 This is a schematic diagram of the composition of a temperature anomaly detection system for key equipment of a belt conveyor, provided in an embodiment of the present invention. Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation
[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.
[0020] See Figure 1 The image shows a specific embodiment of a temperature anomaly detection method for key equipment in a belt conveyor provided by the present invention. In this embodiment, the temperature anomaly detection method for key equipment in a belt conveyor includes:
[0021] Step 101: Collect time-series data from multiple temperature measurement points of key equipment of the belt conveyor and preprocess the time-series data; Specifically, the key equipment of the belt conveyor in this embodiment includes a drive motor, a reducer, a hydraulic coupler, a drive roller, a redirecting roller, idler rollers, and a tensioning device. Multiple temperature measurement points are pre-arranged on each key piece of equipment; for example, PT100 platinum resistance temperature sensors are installed on the drive motor's housing, front and rear bearings, and windings. Time-series data refers to a time series composed of temperature values from various measurement points collected continuously at a fixed sampling period (e.g., 1 minute). Preprocessing includes data integrity checks (detecting whether data packets are missing), outlier correction (removing values exceeding the sensor's range or physical limits, such as above 150℃ or below -50℃), missing value completion (using linear interpolation or forward filling), time alignment (ensuring all measurement point data corresponds to the same timestamp), and normalization (mapping each measurement point's temperature value to the [0,1] interval through a maximum-minimum transformation to eliminate dimensional interference). Preprocessing provides a clean, aligned, and scale-uniform data foundation for subsequent sample construction, avoiding noise and missing features in the original data during feature extraction.
[0022] Step 102: Based on the preprocessed time series data, construct a temperature sample set containing temperature values from multiple temperature measuring points, and construct a correlation diagram representing the correlation between measuring points based on the correlation of temperature change trends between different temperature measuring points. Furthermore, a temperature sample set is constructed based on a sliding time window: the window width is set to W time points (e.g., W=10, corresponding to 10 minutes), and the step size is 1 time point. The temperature values of all measuring points within each time window are organized into a two-dimensional matrix, with the number of rows equal to the number of measuring points N and the number of columns equal to the window width W. This matrix is a temperature sample. The sample sets obtained from all sliding windows constitute the temperature sample set. Then, an association graph is constructed: each temperature measuring point is regarded as a node in the graph, with a total of N nodes. For the samples within each window, the Pearson correlation coefficient between any two measuring points and their corresponding temperature change sequences (two vectors of length W) is calculated. The Pearson correlation coefficient measures the degree of linear correlation between two sequences, with a value range of [-1, 1]. The closer the absolute value is to 1, the more consistent the trend of change. A similarity threshold (e.g., 0.7) is set. When the Pearson correlation coefficient between two measuring points is greater than this threshold, it is considered that the temperature change trends of the two measuring points within the corresponding window are strongly correlated, and an undirected edge is established between the two graph nodes. Repeat the above operation for all windows to eventually form a dynamic relationship graph (different time windows may yield different graph structures). This step transforms the multi-point temperature monitoring problem into a graph structure learning problem, automatically capturing the thermal coupling relationships between different parts of key equipment, and providing topological constraints for subsequent extraction of collaborative relationships.
[0023] Step 103: Extract the single-point temperature change features of each temperature measuring point in the temperature sample set, and extract the collaborative relationship features between multiple measuring points in the correlation graph; Furthermore, the extraction method for single-point temperature change features is as follows: each sample (i.e., an N×W matrix) in the temperature sample set is input into a one-dimensional convolutional neural network (1D-CNN). This 1D-CNN architecture includes two convolutional layers, each containing 16 convolutional kernels with a kernel size of 3 and a stride of 1. The activation function is ReLU. The convolution operation slides along the time dimension (window width direction) to extract the local temperature rise pattern of each temperature measurement point over time, such as the temperature rise rate and local fluctuation amplitude. Subsequently, the feature sequence output from the convolutional layers is input into a gated recurrent unit (GRU) network. GRU is a lightweight recurrent neural network containing update and reset gates, effectively capturing temporal dependencies. In this embodiment, the GRU has 32 hidden units. After the input features are encoded by GRU, the output is the hidden state vector of each measurement point. The hidden states of all measurement points are concatenated to form a single-point temperature change feature matrix with dimensions N×32.
[0024] The collaborative relationship feature extraction method involves simultaneously inputting the adjacency matrix of the association graph (an N×N 0-1 matrix representing whether there are edges between nodes) and the temperature sample set into a Graph Convolutional Network (GCN). The GCN architecture consists of two layers of graph convolution. The first layer has an input dimension of N×W (the original temperature samples) and an output dimension of N×16; the second layer has an input dimension of N×16 and an output dimension of N×16. The graph convolution operation, according to the graph structure defined by the adjacency matrix, weights and aggregates the features of each node based on the features of its neighboring nodes, ensuring that the output features reflect the spatial coupling relationship between the measurement points. For example, when the winding temperature of the drive motor increases, the temperature of adjacent bearing measurement points will also be encoded by the GCN. The final output collaborative relationship feature matrix has a dimension of N×16.
[0025] In this step, 1D-CNN, GRU, and GCN were all trained offline. All training samples were temperature data collected over seven consecutive days from critical equipment operating under normal conditions (no faults, stable load). The training objective was not classification or regression, but rather to enable these feature extraction networks to output intermediate features with representational capabilities to serve subsequent reconstruction tasks. Training was conducted jointly with the decoding and reconstruction network in step 104, using an end-to-end approach to minimize reconstruction errors. This step, by separating and extracting local temporal features and global topological features, preserves the independent temperature rise evolution information of each measuring point while integrating the collaborative change patterns between measuring points, providing rich features for a comprehensive characterization of normal thermal states.
[0026] Step 104: Fuse the temperature change features of a single measurement point with the features of the cooperative relationship, and reconstruct the normal thermal state pattern based on the fused features to obtain the reconstructed sample; Furthermore, the fusion operation in this embodiment employs feature concatenation: the single-point temperature change feature matrix (N×32) obtained in step 103 and the cooperative relationship feature matrix (N×16) are concatenated along the feature dimension (the last dimension) to form a fused feature vector matrix with a dimension of N×48. This fused feature vector simultaneously contains the temporal evolution pattern of each measurement point and the topological coupling information between measurement points. The fused feature vector is then input into the decoding and reconstruction network. The decoding and reconstruction network consists of fully connected layers and upsampling layers: first, it passes through two fully connected layers, with 256 neurons in the first layer and 128 in the second, both using ReLU activation functions; then, it passes through another fully connected layer to map the dimension to N×W (the same dimension as the original temperature sample); finally, it is connected to an upsampling layer (using transposed convolution or simple repeated expansion if necessary) to ensure that the output dimension accurately matches the N×W of the original sample. The output of this decoding and reconstruction network is the reconstructed sample. During training, the temperature sample set under normal operating conditions is used as training data, and the loss function is the mean squared error (MSE) between the original temperature sample and the decoded and reconstructed sample. The optimizer used was Adam, with a learning rate of 0.001, a batch size of 32, and 100 training epochs until the loss function converged. At this point, the decoding and reconstruction network learned the temperature pattern of the key equipment under normal thermal conditions; that is, given an input temperature sample, the network could output a reconstruction result that matched the normal pattern. This step, through feature fusion and reconstruction, compressed the normal thermal state pattern into a low-dimensional manifold, providing a benchmark for subsequent anomaly detection.
[0027] Step 105: Determine the anomaly score based on the reconstruction error between the original temperature sample and the reconstructed sample, and establish anomaly discrimination threshold based on the reconstruction error distribution under normal operating conditions. Further, for each normal temperature sample used for training, the reconstruction error between it and the corresponding reconstructed sample is calculated. The reconstruction error is calculated as follows: first, the absolute error between the original value and the reconstructed value at each temperature measurement point is calculated, resulting in N absolute error values; then, the absolute errors of all measurement points are summed (or the root mean square is calculated) to obtain a scalar value, which serves as the anomaly score for that sample. The larger the anomaly score, the greater the degree to which the sample deviates from the normal thermal state pattern. After collecting the anomaly scores of all normal training samples, an anomaly score set is obtained. The mean μ and standard deviation σ of this set are calculated. Based on the three Sigma criterion in statistics (assuming that the reconstruction error under normal operating conditions follows a normal or approximately normal distribution), the anomaly discrimination threshold is set to μ+3σ. This threshold can theoretically cover 99.7% of normal fluctuations, and samples exceeding this threshold are considered anomalies. This step adaptively determines the threshold using the statistical distribution of the normal samples themselves, without the need for manual setting or reliance on fault samples, and is suitable for the actual situation of open-pit coal mines where there is no large amount of historical fault data.
[0028] Step 106: Based on the comparison results of the anomaly score and the anomaly discrimination threshold, identify the thermal state anomalies of key equipment and output an early warning.
[0029] Furthermore, for the temperature samples collected in real time, steps 101 to 104 are first executed to obtain the corresponding reconstructed samples and anomaly scores. Then, the anomaly score is compared with the anomaly discrimination threshold established in step 105. If the anomaly score is less than or equal to the threshold, the thermal state of the critical equipment is determined to be normal; if the anomaly score is greater than the threshold, an anomaly in the thermal state is determined, and a warning signal is output. The warning signal may include the time of the anomaly occurrence, the magnitude of exceeding the threshold, and the anomaly score value, used to prompt on-site maintenance personnel to pay attention to the equipment status. This step automates the entire process from data acquisition to anomaly determination, outputting clear and actionable warning information.
[0030] As can be seen from the above technical solutions, the beneficial effects of this embodiment are: by constructing a measurement point association map through a sliding window, extracting and fusing dual-path features, and reconstructing error statistical threshold discrimination, a normal thermal state benchmark can be established under fault-free label conditions, while simultaneously sensing local temperature rise shifts and imbalances in the collaborative relationship of measurement points, thus achieving early and accurate temperature anomaly warnings.
[0031] Figure 1 The embodiments shown are merely basic examples of the method of the present invention. Other preferred embodiments of the method can be obtained by making certain optimizations and extensions based on them.
[0032] like Figure 2 The image shows another specific embodiment of a temperature anomaly detection method for key equipment in a belt conveyor according to the present invention. This embodiment further describes the method based on the foregoing embodiments, and includes the following steps:
[0033] Step 201: After the real-time multi-point temperature time series data is preprocessed, temperature sample set is constructed, correlation graph is constructed, single-point temperature change feature is extracted, collaborative relationship feature is extracted, feature fusion and reconstruction are performed, and a real-time anomaly score is obtained. During the online monitoring phase, as the belt conveyor operates, the data acquisition system continuously acquires real-time time-series data from multiple temperature measurement points of key equipment at the same sampling period (e.g., 1 minute) as during the training phase. First, the same preprocessing operations as in step 101 are performed, including integrity checks, outlier correction, missing value completion, time alignment, and normalization, ensuring the format and quality of the real-time data are consistent with the training data. Then, following the method in step 102, a real-time temperature sample set is constructed based on a sliding time window: taking W time points backward from the current time as the window endpoint to form a temperature sample matrix. Subsequently, an association graph within the current window is constructed, the Pearson correlation coefficient is calculated, and compared with a preset similarity threshold to obtain the real-time association graph. Next, using the one-dimensional convolutional neural network, gated recurrent unit, and graph convolutional network trained in step 103, single-point temperature change features and collaborative relationship features are extracted, respectively. The two types of features are then concatenated and input into the decoding and reconstruction network trained in step 104 to obtain real-time reconstructed samples. Finally, following the method in step 105, the reconstruction error between the real-time original temperature samples and the real-time reconstructed samples is calculated to obtain a real-time anomaly score. Throughout the process, the parameters of the feature extraction network and the decoding and reconstruction network remain frozen and are not updated online. This step directly applies the offline-trained normal hot-state model to the real-time data stream, resulting in high computational efficiency and meeting the real-time requirements of online monitoring.
[0034] Step 202: When the real-time anomaly score exceeds the anomaly judgment threshold, it is determined that the key equipment has a thermal anomaly, and an early warning result containing the location information of the anomaly measurement point and the direction of temperature rise offset is output.
[0035] Specifically, the real-time anomaly score obtained in step 201 is compared with the anomaly discrimination threshold (mean plus three standard deviations) established in step 105. If the real-time anomaly score is greater than the threshold, an anomaly alarm is triggered. To provide maintenance personnel with more specific fault clues, this embodiment further calculates the individual reconstruction error for each temperature measuring point: for samples that trigger alarms, the absolute error between the original temperature value and the reconstructed temperature value at each measuring point is output, and the absolute errors are sorted from largest to smallest. The measuring points ranked first (e.g., the first 3) are marked as potential anomaly measuring points, serving as anomaly measuring point location information. At the same time, based on the difference between the original temperature value and the reconstructed temperature value at the measuring point (i.e., the reconstruction error component of the measuring point), if the difference is positive and exceeds a preset small threshold (e.g., 0.5℃), the temperature rise offset direction is determined to be positive (temperature rise); if the difference is negative and the absolute value exceeds the threshold, the temperature rise offset direction is determined to be negative (temperature fall). If the absolute value of the difference does not exceed the threshold, the measuring point is not determined to have a significant temperature rise offset. The final output of the warning result includes: the time of anomaly occurrence, the anomaly score, the anomaly measurement point identifier (e.g., "front bearing of drive motor"), and the direction and amount of temperature rise deviation for each anomaly measurement point. This warning result can be sent to the central control center or the field HMI via industrial communication protocols (e.g., OPC UA, Modbus TCP), and supports audible and visual alarms. This step provides interpretable diagnostic information simultaneously with the detection of the anomaly, helping maintenance personnel quickly locate the fault location and determine the nature of the fault, reducing downtime for troubleshooting.
[0036] As can be seen from the above technical solutions, the beneficial effects of this embodiment are: on the basis of the previous embodiment, real-time anomaly scoring calculation and fine early warning output including measurement point location and temperature rise direction are added, which significantly improves the practicality of field application and the efficiency of fault diagnosis.
[0037] This invention also provides a temperature anomaly detection system for key equipment in belt conveyors. See also Figure 3 The image shows a specific embodiment of a temperature anomaly detection system for key equipment in a belt conveyor provided by the present invention. This embodiment of the system is used to perform... Figures 1-2 The physical apparatus of the method. Its technical solution is essentially the same as the embodiments described above, and the corresponding descriptions in the embodiments above also apply to this embodiment. The system includes:
[0038] The data preprocessing module 301 is configured to collect time-series data from multiple temperature measurement points of key equipment of the belt conveyor and preprocess the time-series data. The correlation graph construction module 302 is configured to construct a temperature sample set containing temperature values of multiple temperature measuring points based on the preprocessed time series data, and construct a correlation graph representing the correlation between measuring points based on the correlation of temperature change trends between different temperature measuring points. The feature extraction module 303 is configured to extract the single-point temperature change features of each temperature measuring point in the temperature sample set, and to extract the collaborative relationship features between multiple measuring points in the correlation graph. The reconstruction module 304 is configured to fuse the temperature change features of a single measurement point with the features of the cooperative relationship, and reconstruct the normal thermal state mode based on the fused features to obtain the reconstructed sample. The anomaly discrimination threshold establishment module 305 is configured to determine anomaly scores based on the reconstruction error between the original temperature sample and the reconstructed sample, and to establish anomaly discrimination thresholds based on the reconstruction error distribution under normal operating conditions. The identification module 306 is configured to identify thermal state anomalies of key equipment and output an early warning based on the comparison result of anomaly score and anomaly discrimination threshold.
[0039] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. At the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and a memory. The memory may include main memory, such as high-speed random-access memory (RAM), or it may also include non-volatile memory, such as at least one disk storage device. Of course, the electronic device may also include other hardware required for other services.
[0040] The processor, network interface, and memory can be interconnected via an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. Buses can be categorized as address buses, data buses, and other types. For ease of representation, Figure 4 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.
[0041] Memory is used to store instructions for execution. Specifically, instructions for execution are computer programs that can be executed. Memory can include main memory and non-volatile memory, and it provides the processor with execution instructions and data.
[0042] In one possible implementation, the processor reads the corresponding execution instructions from non-volatile memory into main memory and then executes them. Alternatively, it may obtain the corresponding execution instructions from other devices to logically form a temperature anomaly detection device for critical equipment of a belt conveyor. The processor executes the execution instructions stored in the memory to implement the temperature anomaly detection method for critical equipment of a belt conveyor provided in any embodiment of the present invention.
[0043] The above is as described in the present invention. Figure 3 The method for temperature anomaly detection system of key equipment in a belt conveyor provided in the illustrated embodiment can be applied to or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor.
[0044] The steps of the method disclosed in the embodiments of this invention can be directly manifested as being executed by a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.
[0045] This invention also proposes a readable medium storing execution instructions. When these instructions are executed by a processor of an electronic device, the device can perform a temperature anomaly detection method for key equipment of a belt conveyor provided in any embodiment of this invention, specifically for performing actions such as... Figure 1 , Figure 2 The method shown.
[0046] The electronic devices in the foregoing embodiments may be computers.
[0047] Those skilled in the art will understand that embodiments of the present invention can be provided as methods or computer program products. Therefore, the present invention can be implemented in a completely hardware embodiment, a completely software embodiment, or a combination of software and hardware.
[0048] The various embodiments in this invention are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the apparatus embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.
[0049] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0050] The above are merely embodiments of the present invention and are not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.
Claims
1. A method for detecting temperature anomalies in key equipment of a belt conveyor, characterized in that, The method includes: Time-series data from multiple temperature measurement points of key equipment in a belt conveyor are collected, and the time-series data is preprocessed. Based on the preprocessed time series data, a temperature sample set containing temperature values from multiple temperature measurement points is constructed, and a correlation graph representing the relationship between measurement points is constructed based on the correlation of temperature change trends between different temperature measurement points. Extract the single-point temperature change features of each temperature measuring point in the temperature sample set, and extract the collaborative relationship features between multiple measuring points in the correlation graph; The single-point temperature change features are fused with the cooperative relationship features, and the normal thermal state pattern is reconstructed based on the fused features to obtain a reconstructed sample. Based on the reconstruction error between the original temperature sample and the reconstructed sample, an anomaly score is determined, and an anomaly discrimination threshold is established based on the reconstruction error distribution under normal operating conditions. Based on the comparison between the anomaly score and the anomaly discrimination threshold, the thermal state anomalies of key equipment are identified and an early warning is output.
2. The method according to claim 1, characterized in that, The construction includes a temperature sample set containing temperature values from multiple temperature measuring points, and based on the correlation of temperature change trends between different temperature measuring points, a correlation graph characterizing the relationship between the measuring points is constructed, including: Based on the sliding time window, a multi-point temperature sample is constructed from the preprocessed time series data, and each temperature point is used as a graph node. Within the sliding time window, calculate the Pearson correlation coefficient of the temperature change sequence between any two temperature measurement points; When the Pearson correlation coefficient is greater than a preset similarity threshold, an edge is established between the graph nodes corresponding to the two temperature measurement points to form the association graph.
3. The method according to claim 1, characterized in that, The extraction of single-point temperature change characteristics from each temperature measuring point in the temperature sample set includes: The temperature sample set is input into a one-dimensional convolutional neural network to extract the local temperature rise change pattern of each temperature measurement point along the time dimension. The local temperature rise change pattern is time-dependently encoded by a gated loop unit, and the temperature change characteristics of the single measurement point are output.
4. The method according to claim 1, characterized in that, The extraction of collaborative relationship features among multiple measurement points in the association graph includes: The adjacency matrix of the association graph is input into the temperature sample set and then convolved into a network. The graph convolutional network aggregates the temperature state information of adjacent temperature measurement points along the graph structure and outputs the cooperative relationship feature.
5. The method according to claim 1, characterized in that, The process involves fusing the single-point temperature change features with the cooperative relationship features, and reconstructing the normal thermal state pattern based on the fused features to obtain reconstructed samples, including: The single-point temperature change feature and the collaborative relationship feature are concatenated along the feature dimension to form a fused feature vector; The fused feature vector is input into a decoding and reconstruction network consisting of a fully connected layer and an upsampling layer to generate a decoding and reconstruction sample with the same dimension as the original temperature sample, which is used as the reconstruction sample. Using a temperature sample set under normal operating conditions as training data, the mean square error between the original temperature sample and the decoded and reconstructed sample is minimized, so that the decoding and reconstruction network converges to the normal thermal state mode of the key equipment.
6. The method according to claim 1, characterized in that, The process of determining anomaly scores based on the reconstruction error between the original temperature sample and the reconstructed sample, and establishing anomaly discrimination thresholds based on the reconstruction error distribution under normal operating conditions, includes: Calculate the absolute error of each original temperature sample and its corresponding reconstructed sample at each temperature measurement point, and sum or calculate the root mean square of the absolute errors of all measurement points as the anomaly score of the sample. Collect abnormal scores from multiple samples under normal operating conditions and calculate their mean and standard deviation; The anomaly detection threshold is set to the mean plus three times the standard deviation.
7. The method according to any one of claims 1 to 6, characterized in that, Also includes: The real-time temperature time series data collected from multiple measurement points is processed sequentially through preprocessing, temperature sample set construction, correlation graph construction, single measurement point temperature change feature extraction, collaborative relationship feature extraction, feature fusion and reconstruction to obtain a real-time anomaly score; When the real-time anomaly score exceeds the anomaly discrimination threshold, it is determined that the key equipment has a thermal anomaly, and an early warning result containing the location information of the anomaly measurement point and the direction of temperature rise offset is output.
8. A temperature anomaly detection system for key equipment in belt conveyors, characterized in that, include: The data preprocessing module is used to collect time-series data from multiple temperature measurement points of key equipment of the belt conveyor and to preprocess the time-series data. The correlation graph construction module is used to construct a temperature sample set containing temperature values from multiple temperature measurement points based on the preprocessed time series data, and to construct a correlation graph representing the correlation between measurement points based on the correlation of temperature change trends between different temperature measurement points. The feature extraction module is used to extract the single-point temperature change features of each temperature measuring point in the temperature sample set, and to extract the collaborative relationship features between multiple measuring points in the correlation graph. The reconstruction module is used to fuse the temperature change features of the single measurement point with the cooperative relationship features, and reconstruct the normal thermal state mode based on the fused features to obtain a reconstructed sample. The anomaly detection threshold establishment module is used to determine the anomaly score based on the reconstruction error between the original temperature sample and the reconstructed sample, and to establish the anomaly detection threshold based on the reconstruction error distribution under normal operating conditions. The identification module is used to identify thermal state anomalies of key equipment and output early warnings based on the comparison result between the anomaly score and the anomaly discrimination threshold.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein the program, when executed, performs the method of any one of claims 1 to 7.
10. An electronic device, characterized in that, The electronic device includes: processor; Memory used to store the processor's executable instructions; The processor is configured to read the executable instructions from the memory and execute the instructions to implement the method described in any one of claims 1 to 7.
