A method, device and computer storage medium for intelligent maintenance of highway electromechanical equipment

Abstract

This invention discloses a smart maintenance method, equipment, and computer storage medium for highway electromechanical equipment, relating to the field of highway electromechanical equipment maintenance technology. The method includes: collecting equipment operation data, cleaning and extracting features from the data to obtain a standardized multimodal feature set; calculating the current health index of the equipment based on the multimodal feature set through weighted fusion, and classifying the equipment health level according to the health index; when the health level reaches a preset trigger condition, inputting real-time and historical multimodal feature sequences into a pre-trained long short-term memory network model, and outputting a predicted value of the equipment's remaining service life; automatically generating maintenance work orders based on the health index and the predicted value of the remaining service life, and dynamically scheduling maintenance resources in conjunction with maintenance personnel location, spare parts inventory, and traffic flow information. This invention solves the technical problems of slow response, incomplete status perception, and high maintenance costs existing in the prior art.

Description

Technical Field

[0001] This invention relates to the field of highway electromechanical equipment maintenance technology, and more specifically to a smart maintenance method, equipment, and computer storage medium for highway electromechanical equipment. Background Technology

[0002] Highway electromechanical equipment is a core infrastructure ensuring road traffic safety and operational efficiency. With the continuous expansion of highway transportation networks and the improvement of their intelligence level, the number and complexity of electromechanical equipment are increasing, making its maintenance work increasingly important.

[0003] Currently, the maintenance of highway electromechanical equipment mainly relies on two methods: manual inspection and regular maintenance. However, these methods have the following technical drawbacks: First, fault response is delayed. Traditional maintenance methods typically only carry out repairs after equipment malfunctions and stops, leading to unplanned downtime that can cause traffic congestion or even safety accidents, disrupting road traffic flow.

[0004] Second, the inspection coverage is limited. Manual inspections cannot achieve continuous, real-time status perception of all types of equipment across the entire line, and cannot promptly detect potential deterioration trends in equipment.

[0005] Third, maintenance costs remain high. Due to a lack of accurate assessment of the true health status of equipment, over-maintenance strategies are often adopted in practice to reduce the risk of failure, or maintenance is neglected due to insufficient monitoring, both of which result in a waste of human and spare parts resources.

[0006] While existing technologies have developed remote monitoring solutions based on the Internet of Things (IoT), they are mostly limited to the collection of single-type data and threshold alarms. They lack the ability to deeply integrate and analyze multi-source heterogeneous data, cannot quantitatively assess the health status of equipment, and are even less capable of predicting faults in advance and proactively optimizing maintenance strategies. As a result, they fail to meet the technical requirements of predictive maintenance for equipment in the construction of smart highways.

[0007] Therefore, how to provide a smart maintenance method, equipment, and computer storage medium for highway electromechanical equipment is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0008] In view of this, the present invention provides a method, equipment and computer storage medium for intelligent maintenance of highway electromechanical equipment, aiming to solve the technical problems of slow response, incomplete status perception and high maintenance costs in the existing highway electromechanical equipment maintenance technology.

[0009] To achieve the above objectives, the present invention adopts the following technical solution: A smart maintenance method for highway electromechanical equipment includes the following steps: S1: Collect equipment operation data, and perform data cleaning and feature extraction on the equipment operation data to obtain a standardized multimodal feature set; S2: Based on the multimodal feature set, calculate the current health index of the device through weighted fusion, and classify the device health level according to the health index; S3: When the health level reaches the preset trigger condition, the real-time and historical multimodal feature sequences are input into the pre-trained long short-term memory network model, and the remaining service life prediction value of the device is output. S4: Based on the health index and the predicted remaining service life, automatically generate maintenance work orders and dynamically schedule maintenance resources by combining the location of maintenance personnel, spare parts inventory and traffic flow information.

[0010] Furthermore, it also includes: S5: Record the maintenance execution results and update the historical database. Use the updated data to retrain the Long Short-Term Memory network model and optimize the weights of the health index.

[0011] Furthermore, the data cleaning includes using the isolated forest algorithm to identify and remove sensor outliers.

[0012] Furthermore, the isolated forest algorithm constructs multiple isolated trees to score data points for anomalies, and marks points with scores exceeding a preset threshold as outliers and removes them.

[0013] Furthermore, the health index is calculated using a weighted fusion method, with the weights determined by a random forest model based on the importance scores of each feature to the equipment failure type.

[0014] Furthermore, the calculation of the health index also includes a correction coefficient based on digital twin simulation; the digital twin simulation constructs a three-dimensional geometric model that maps to the physical device, and continuously calibrates the simulation parameters using the multimodal feature set, performs accelerated degradation simulation in a virtual environment, and superimposes the degradation trend obtained from the simulation as a correction coefficient into the weighted fusion calculation result.

[0015] Furthermore, the health index is also corrected based on the equipment association risk, which is calculated by a graph neural network. The graph neural network uses each electromechanical device as a node and the physical distance and functional dependency relationship between devices as edges. It uses the multimodal features of each node as the initial embedding, and outputs the conditional risk probability of adjacent device failure on the target device after message transmission. This probability is then added to the health index as a subtraction term.

[0016] Furthermore, the Long Short-Term Memory network model has a two-layer structure, with each layer containing 64 neurons. The output layer is a fully connected layer, and the activation function is ReLU. The mean squared error is used as the loss function when training the model, and the Adam optimizer is used.

[0017] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a smart maintenance method for highway electromechanical equipment.

[0018] A computer storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of a smart maintenance method for highway electromechanical equipment.

[0019] As can be seen from the above technical solution, compared with the prior art, the present invention provides a method, equipment, and computer storage medium for intelligent maintenance of highway electromechanical equipment, which has the following beneficial effects: (1) By collecting data from multiple types of sensors such as vibration, temperature, current, and images, and combining them with the isolated forest algorithm to clean outliers, a comprehensive and accurate perception of the equipment's operating status can be achieved.

[0020] (2) The random forest model is used to determine the feature weights, the digital twin simulation correction coefficient is introduced, and the correlation risk between equipment is considered to construct a multi-dimensional and high-precision equipment health index assessment system, which can truly reflect the equipment deterioration trend.

[0021] (3) Use the long short-term memory network model to predict the remaining service life of the equipment and start the prediction when the health level reaches the trigger condition, so as to provide a scientific basis for maintenance decision-making.

[0022] (4) Taking into account multiple factors such as health index, remaining service life, personnel location, spare parts inventory and traffic flow, maintenance work orders are automatically generated and resources are optimized to achieve a balance between maintenance costs and road traffic efficiency.

[0023] (5) By recording maintenance results and retraining the model, an adaptive closed loop is formed, so that the maintenance strategy can continue to evolve with the accumulation of data. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present 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 embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0025] Figure 1This is a schematic diagram of the method flow provided by the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Example 1: See Figure 1 This invention discloses a smart maintenance method for highway electromechanical equipment, comprising the following steps: S1: Collect equipment operation data, and perform data cleaning and feature extraction on the equipment operation data to obtain a standardized multimodal feature set; S2: Based on the multimodal feature set, the current health index of the device is calculated through weighted fusion, and the health level of the device is classified according to the health index; S3: When the health level reaches the preset trigger condition, the real-time and historical multimodal feature sequences are input into the pre-trained long short-term memory network model, and the remaining service life prediction value of the device is output. S4: Based on the health index and the predicted remaining service life, automatically generate maintenance work orders and dynamically schedule maintenance resources by combining the location of maintenance personnel, spare parts inventory and traffic flow information.

[0028] As a further technical solution, the method also includes: S5: Record and maintain execution results and update the historical database. Use the updated data to retrain the long short-term memory network model and optimize the weights of the health index.

[0029] As a further technical solution, data cleaning includes using the isolated forest algorithm to identify and remove sensor outliers.

[0030] As a further technical solution, the isolated forest algorithm constructs multiple isolated trees to score data points for anomalies, and marks points with scores exceeding a preset threshold as outliers and removes them.

[0031] As a further technical solution, the health index is calculated using a weighted fusion method, with the weights determined by a random forest model based on the importance scores of each feature to the equipment failure type.

[0032] As a further technical solution, the calculation of the health index also includes a correction coefficient based on digital twin simulation. Digital twin simulation constructs a three-dimensional geometric model that maps to the physical device and continuously calibrates the simulation parameters using a multimodal feature set. It performs accelerated degradation simulation in a virtual environment and uses the degradation trend obtained from the simulation as a correction coefficient to be superimposed on the weighted fusion calculation result.

[0033] As a further technical solution, the health index is also adjusted based on the associated risks of the equipment.

[0034] As a further technical solution, the equipment association risk is calculated through a graph neural network. The graph neural network uses each electromechanical device as a node and the physical distance and functional dependence between devices as edges. The multimodal features of each node are used as the initial embedding. After message transmission, the conditional risk probability of adjacent device failure on the target device is output, and this probability is added to the health index as a subtraction term.

[0035] As a further technical solution, the Long Short-Term Memory Network model has a two-layer structure, with each layer containing 64 neurons. The output layer is a fully connected layer with ReLU activation function. The mean squared error is used as the loss function when training the model, and the Adam optimizer is used.

[0036] This invention also discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-mentioned technical solutions.

[0037] This invention also discloses a computer storage medium storing a computer program, which, when executed by a processor, implements the steps of a smart maintenance method for highway electromechanical equipment according to any of the above technical solutions.

[0038] Example 2: This embodiment provides a smart maintenance method for highway electromechanical equipment, including the following steps S1 to S5.

[0039] Specifically, S1: Multimodal data acquisition and preprocessing: Deploy multiple types of sensors on highway electromechanical equipment.

[0040] Specifically, vibration sensors (sampling frequency not less than 1kHz), temperature sensors (accuracy ±0.5℃), current transformers, and high-definition industrial cameras (resolution not less than 1080P) are installed on key equipment such as ETC lane controllers, tunnel fans, and variable message signs to collect vibration spectrum, temperature value, current value, and images of the equipment surface during equipment operation.

[0041] After data collection, the Isolation Forest algorithm was used to identify and remove outliers from the original data. The Isolation Forest algorithm scores data points as outliers by constructing multiple isolated trees: for each data point, its average path length across all isolated trees is calculated; the shorter the path length, the more likely the point is to be isolated, and the higher the outlier score. Points with scores exceeding a preset threshold (e.g., 0.6) are marked as outliers and removed. For missing data after removal, a sliding window averaging method is used for imputation.

[0042] Feature extraction includes: extracting time-domain features (mean, variance, peak value) and frequency-domain features (extracting energy of each frequency band through wavelet transform) from vibration signals; extracting features such as mean and volatility from current signals; and extracting features such as crack length and area from images using edge detection algorithms. All extracted features are then normalized using Min-Max to obtain a standardized multimodal feature set.

[0043] S2: Equipment Health Status Assessment Based on the multimodal feature set obtained from S1, the current health index of the device is calculated through weighted fusion.

[0044] First, a random forest model is used to determine the weights of each feature. Specifically, a random forest classifier is trained using historical fault data, with the equipment fault type as the label and each feature as the input. The importance score (such as the reduction in the Gini coefficient) of each feature in the random forest is calculated, and the importance score is normalized and used as the weight.

[0045] Secondly, calculate the initial health index: HI initial =Σ(w i ×x i ) Where w i Let x be the weight of the i-th feature. i Its normalized value.

[0046] To improve the accuracy of the health index, this embodiment introduces a digital twin simulation correction coefficient. Specifically, a three-dimensional geometric model mapped to the physical equipment is constructed, and the boundary conditions and material parameters of the simulation model are continuously calibrated using real-time acquired multimodal feature sets. Accelerated degradation simulation is performed in a virtual environment to obtain the degradation trend curve of the equipment under ideal operating conditions. The ratio of the simulated health level at the current moment to the theoretical initial health level is used as the correction coefficient δ to calculate the corrected health index. HI corrected =HI initial ×δ.

[0047] Furthermore, this embodiment also considers equipment association risk. A graph neural network is used to calculate the conditional risk probability of adjacent equipment failures on the target equipment: an association risk graph is constructed, where nodes represent various electromechanical devices, and edges represent physical distance (e.g., spatial proximity), electrical coupling (e.g., shared power lines), or functional dependencies between devices (e.g., fan failure causing temperature sensor malfunction). The multimodal features of each node are used as initial embeddings, and through the multi-layer message passing mechanism of the graph neural network, neighbor node information is aggregated to output the conditional risk probability p of the target equipment being affected by adjacent equipment failures. The final health index is calculated as follows: HI final =HI corrected ×(1-p).

[0048] Finally, based on HI final Equipment health levels are classified by value: Health (HI) final ≥0.8), sub-health (0.6≤HI) fina <0.8), fault warning (0.4≤HI) fina <0.6), critical fault (HI) fina <0.4).

[0049] S3: Failure Prediction and Remaining Useful Life Estimation: When the health level assessed by S2 reaches sub-health or below (i.e., HI) fina When <0.8), the fault prediction and remaining useful life estimation steps are initiated.

[0050] The real-time and historical multimodal feature sequences are input into a pre-trained Long Short-Term Memory (LSTM) network model. In this embodiment, the LSTM model adopts a two-layer structure, with each layer containing 64 neurons. The output layer is a fully connected layer, and the activation function is ReLU. The model input consists of daily feature data from the past 30 days (such as daily average vibration energy, daily average temperature, daily average current, etc.), and the output is the predicted remaining lifespan of the equipment (in days).

[0051] When training the model, the time difference between the actual fault time in the historical fault data and the current time is used as the RUL label, the mean squared error is used as the loss function, and the Adam optimizer is used to update the parameters.

[0052] S4: Maintenance Decision Optimization: Maintenance work orders are automatically generated based on the health index calculated by S2 and the remaining service life predicted by S3.

[0053] Specifically, a work order is triggered when the following combined conditions are met: the health index is below 0.6 and the predicted remaining useful life is less than 30 days. If only one condition is met, only an alert is issued without generating a work order to avoid over-maintenance.

[0054] After a work order is generated, maintenance resources are dynamically scheduled based on the location of maintenance personnel, spare parts inventory, and traffic flow information. Specifically, a hierarchical scheduling strategy is adopted: The first layer determines maintenance priority based on equipment type and health level (e.g., tunnel ventilation fan failures have a higher priority than information boards); the second layer uses Dijkstra's algorithm to plan the shortest path based on the distance between the maintenance personnel's GPS location and the spare parts warehouse; the third layer adjusts the work time window to avoid peak hours (such as 7:00-9:00 AM and 5:00-7:00 PM) based on historical traffic flow data.

[0055] S5: Closed-loop feedback and model iteration: After maintenance is completed, technicians record information such as the actual cause of the fault, the model of the replaced parts, the repair time, and the post-repair equipment status in the system, and upload it to the historical database.

[0056] The Long Short-Term Memory (LSTM) network model was retrained using the updated data, and the weights of the health index were optimized. Specifically, a sliding window incremental learning strategy was adopted, using only newly collected data from the most recent maintenance cycle (e.g., 30 days) to fine-tune the LSTM model, preserving historical degradation patterns from the old model and avoiding catastrophic forgetting. For the health index weights, a Bayesian update method was used, employing the actual fault causes and fault times recorded after each maintenance as observations to dynamically adjust the weight allocation of each feature, thus continuously optimizing the health index model.

[0057] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0058] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for intelligent maintenance of highway electromechanical equipment, characterized in that, Includes the following steps: S1: Collect equipment operation data, and perform data cleaning and feature extraction on the equipment operation data to obtain a standardized multimodal feature set; S2: Based on the multimodal feature set, calculate the current health index of the device through weighted fusion, and classify the device health level according to the health index; S3: When the health level reaches the preset trigger condition, the real-time and historical multimodal feature sequences are input into the pre-trained long short-term memory network model, and the remaining service life prediction value of the device is output. S4: Based on the health index and the predicted remaining service life, automatically generate maintenance work orders and dynamically schedule maintenance resources by combining the location of maintenance personnel, spare parts inventory and traffic flow information.

2. The intelligent maintenance method for highway electromechanical equipment according to claim 1, characterized in that, Also includes: S5: Record the maintenance execution results and update the historical database. Use the updated data to retrain the Long Short-Term Memory network model and optimize the weights of the health index.

3. The intelligent maintenance method for highway electromechanical equipment according to claim 1, characterized in that, The data cleaning process includes using the isolated forest algorithm to identify and remove outlier values ​​from sensors.

4. The intelligent maintenance method for highway electromechanical equipment according to claim 3, characterized in that, The isolated forest algorithm constructs multiple isolated trees to score data points for anomalies, and marks points with scores exceeding a preset threshold as outliers and removes them.

5. The intelligent maintenance method for highway electromechanical equipment according to claim 1, characterized in that, The health index is calculated using a weighted fusion method, with the weights determined by a random forest model based on the importance scores of each feature to the equipment failure type.

6. The intelligent maintenance method for highway electromechanical equipment according to claim 4, characterized in that, The calculation of the health index also includes a correction coefficient based on digital twin simulation; the digital twin simulation constructs a three-dimensional geometric model that maps to the physical device and continuously calibrates the simulation parameters using the multimodal feature set, performs accelerated degradation simulation in a virtual environment, and superimposes the degradation trend obtained from the simulation as a correction coefficient into the weighted fusion calculation result.

7. The intelligent maintenance method for highway electromechanical equipment according to claim 5, characterized in that, The health index is further corrected based on the equipment association risk, which is calculated by a graph neural network. The graph neural network uses each electromechanical device as a node and the physical distance and functional dependency between devices as edges. It uses the multimodal features of each node as the initial embedding, and outputs the conditional risk probability of adjacent device failures on the target device after message transmission. This probability is then added to the health index as a subtraction term.

8. The intelligent maintenance method for highway electromechanical equipment according to claim 1, characterized in that, The Long Short-Term Memory (LSTM) network model has a two-layer structure, with each layer containing 64 neurons. The output layer is a fully connected layer, and the activation function is ReLU. The mean squared error is used as the loss function and the Adam optimizer is used when training the model.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements a smart maintenance method for highway electromechanical equipment as described in any one of claims 1 to 8.

10. A computer storage medium, characterized in that, The computer storage medium stores a computer program, which, when executed by a processor, implements the steps of a smart maintenance method for highway electromechanical equipment as described in any one of claims 1 to 8.

Images