Cableway monitoring management method and system based on artificial intelligence and edge calculation

Abstract

The invention relates to the technical field of cableway fault monitoring, solves the technical problems that in the prior art, the difference between a fault prediction result and the reality is large, nonlinear data is difficult to predict accurately, and the prediction result is inaccurate, and particularly relates to a cableway monitoring management method and system based on artificial intelligence and edge calculation. The method comprises the following steps: S1, obtaining original data of a cableway monitoring system, preprocessing the original data to obtain alignment data, capturing dynamic features through time sequence dependence, providing feature importance interpretation, organically combining time sequence prediction and feature contribution by a weighted fusion mechanism to generate health indexes, and obtaining health indexes of the cableway monitoring system; and the continuous health state can be mapped into a quantifiable fault probability, and finally real-time early warning is realized through threshold triggering, so that the interpretability of a model output result is improved, independent optimization and fault traceability are supported, and the prediction precision and the reliability of operation and maintenance decision are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of cableway fault monitoring technology, and in particular to a cableway monitoring and management method and system based on artificial intelligence and edge computing. Background Technology

[0002] Cableway monitoring and management involves using technical means and standardized processes to comprehensively supervise and control all aspects of cableway design, manufacturing, installation, operation, and maintenance to ensure the safe and efficient operation of the cableway. As a special transportation device, cableways are exposed to complex natural environments for extended periods, enduring alternating loads, corrosion, and wear. Key components are prone to fatigue damage, structural deformation, and electrical faults, necessitating fault prediction. However, existing technologies for cableway fault prediction often result in significant discrepancies between predicted and actual faults, neglecting the importance of characteristic data in cableway processing. Furthermore, the nonlinear data in the raw data of the cableway monitoring system is difficult to predict accurately, leading to inaccurate prediction results. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a cableway monitoring and management method and system based on artificial intelligence and edge computing. This solves the technical problems of existing technologies, such as the large discrepancy between fault prediction results and actual results, and the difficulty in accurately predicting nonlinear data, which leads to inaccurate prediction results.

[0004] To address the aforementioned technical problems, this invention provides the following technical solution: a cableway monitoring and management method based on artificial intelligence and edge computing, comprising the following steps: S1. Obtain the raw data from the cableway monitoring system and preprocess the raw data to obtain aligned data; S2. Construct a multi-type, multi-protocol information transmission network based on the alignment data and obtain the transmission data. Then, perform aggregation processing on the transmission data to obtain aggregated data. S3. Based on aggregated data, a fused feature set is obtained through data fusion and feature extraction; S4. Based on the fusion feature set, predict cableway failure conditions to obtain failure prediction results and cableway safety index; S5. Based on the fault prediction results and the cableway safety index, construct a digital twin synchronization model, and perform simulation and anomaly detection on the digital twin synchronization model data to obtain simulation data and optimization strategies. S6. Generate early warning data based on simulation data and optimization strategies, and send the early warning data to the control center.

[0005] Preferably, in S1, the specific implementation steps are as follows: S11. Based on the raw data of the cableway monitoring system, anomaly detection and removal are performed to obtain cleaned data; S12. Based on the cleaned data, the mean interpolation method is used to process the data to obtain complete data, and the complete data is then normalized to obtain normalized data. S13. Obtain aligned data by time series alignment processing based on normalized data.

[0006] Preferably, in S2, the specific implementation steps are as follows: S21. Obtain encapsulated data by data encapsulation processing based on the alignment data; S22. Construct a multi-type, multi-protocol information transmission network based on encapsulated data and output the transmitted data; S23. Construct a time window of size A and sliding step b based on the transmitted data, and aggregate the transmitted data within the time window to obtain aggregated data.

[0007] Preferably, in S3, the specific implementation steps are as follows: S31. Obtain combined data by concatenating aggregated data; S32. Obtain feature data by feature extraction based on the combined data; S33. Obtain coded features by feature encoding processing based on feature data; S34. Construct a feature decoder based on feature encoding processing, and reconstruct the reconstructed features using the feature decoder based on the encoded features; S35. Calculate the error value based on the reconstructed features and feature data, determine the error threshold, and filter the feature data according to the error threshold to obtain the fused feature set; If the error value is greater than or equal to the error threshold, the feature data corresponding to the reconstructed feature is removed; otherwise, it is retained. All feature data are filtered and then concatenated to obtain the fused feature set.

[0008] Preferably, in S4, the specific implementation steps are as follows: S41. Obtain the time-series prediction value by processing the fused feature set through time-series prediction. S42. Construct a random forest model based on the fused feature set, and obtain the feature importance score through the random forest model; S43. The cableway safety index is obtained by weighted summation based on the time-series predicted values ​​and feature importance scores; S44. Calculate the failure probability using logistic regression based on the cableway safety index; S45. Generate fault prediction results and cableway safety index based on fault probability.

[0009] Preferably, in S41, the specific implementation steps are as follows: S411. Construct an LSTM processing layer and input the elements of the fused feature set one by one into the LSTM processing layer to obtain the hidden state sequence. S412. Calculate the attention score based on the hidden state sequence, and normalize the attention score to obtain the attention weight. S413. Obtain the context vector by weighted summation based on the attention weights and the hidden state sequence; S414. Initial weight vector and initial bias vector are obtained based on normal distribution initialization. Initial prediction output layer is constructed based on weight vector and bias vector. S415. The initial weight vector and initial bias vector are used as individuals to construct a particle swarm optimization layer. The optimized weight vector and optimized bias vector are obtained through iterative optimization, and an optimized prediction layer is constructed. S416. Input the context vector into the optimized prediction layer to obtain the time series prediction value.

[0010] Preferably, in S5, the specific implementation steps are as follows: S51. Based on the fault prediction results and the cableway safety index, a digital twin model is obtained through physical modeling; S52. Collect the real-time fusion feature set, and assimilate the data by Kalman filtering based on the digital twin model and the real-time fusion feature set to obtain the synchronization model. S53. Obtain the synchronization data in the synchronization model, and obtain the simulation data through simulation processing based on the synchronization data; S54. Construct a GAN discriminator and obtain the anomaly score of the simulation data through the GAN discriminator; S55. Set an anomaly threshold and generate an optimization strategy based on the anomaly score and the anomaly threshold. If the abnormal score is greater than or equal to the abnormal threshold, an optimization strategy is generated; otherwise, it indicates normal operation and proceeds to step S6.

[0011] Preferably, in S6, the specific implementation steps are as follows: S61. Obtain a structured list by standardizing the format based on simulation data and optimization strategies; S62. Obtain control instructions through instruction mapping based on the structured list; S63. Combine control commands, optimization strategies, and simulation data into early warning data, and transmit the early warning data to the control center.

[0012] The technical solution also provides a system for the aforementioned cableway monitoring and management method based on artificial intelligence and edge computing, the system comprising: The preprocessing module is used to acquire the raw data from the cableway monitoring system and preprocess the raw data to obtain aligned data. The aggregation module is used to construct a multi-type, multi-protocol information transmission network based on the alignment data and obtain the transmission data, and to perform aggregation processing on the transmission data to obtain aggregated data. The fusion module is used to obtain a fused feature set based on aggregated data through data fusion and feature extraction; The prediction module is used to predict cableway fault conditions based on the fused feature set, and obtain the fault prediction results and cableway safety index. The digital twin module is used to construct a digital twin synchronization model based on fault prediction results and cableway safety index, and to simulate and detect anomalies in the digital twin synchronization model data to obtain optimization strategies. The control module is used to generate early warning data based on simulation data and optimization strategies, and then send the early warning data to the control center.

[0013] By employing the above technical solutions, the present invention provides a cableway monitoring and management method and system based on artificial intelligence and edge computing, which has at least the following beneficial effects: 1. This invention improves the accuracy and robustness of feature extraction through PCA dimensionality reduction and autoencoder feature enhancement, effectively reducing data dimensionality while retaining key information. It uses a concatenation method to aggregate the original information of multi-source data, preserving the integrity of heterogeneous data and effectively reducing data dimensionality and removing redundant noise. It further mines higher-order feature representations using nonlinear coding to enhance the discriminative power of features. Based on the dynamic threshold selection of reconstruction error, it adaptively retains the most representative feature components, which can significantly improve the generalization ability and interpretability of subsequent steps.

[0014] 2. This invention captures dynamic features through time-series dependence and can provide an explanation of feature importance. The weighted fusion mechanism organically combines time-series prediction and feature contribution to generate a health index. It can also map continuous health status into quantifiable failure probabilities. Finally, it achieves real-time early warning through threshold triggering, which improves the interpretability of model output results and supports independent optimization and fault tracing, significantly improving prediction accuracy and the reliability of operation and maintenance decisions.

[0015] 3. This invention establishes a digital twin model through fault prediction and safety index, providing a precise foundation for subsequent analysis. The synchronous model is obtained by Kalman filtering and data assimilation, which can reflect the cableway status in real time. Combined with the GAN discriminator, anomaly scores are obtained, which can accurately detect anomalies. Anomaly threshold is set to generate optimization strategies, which can respond to abnormal situations in a timely manner and ensure the safe and stable operation of the cableway. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart of the cableway monitoring and management method based on artificial intelligence and edge computing according to the present invention. Figure 2 This is a structural block diagram of the cableway monitoring and management system based on artificial intelligence and edge computing of the present invention. Detailed Implementation

[0017] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. This will allow for a full understanding and implementation of how the present application uses technical means to solve technical problems and achieve technical effects.

[0018] Example 1 Due to the inherent limitations of existing technologies, such as significant discrepancies between predicted and actual faults, and the difficulty in accurately predicting nonlinear data, leading to inaccurate prediction results, please refer to [the relevant documentation / reference]. Figure 1 This embodiment provides a cableway monitoring and management method based on artificial intelligence and edge computing, which can significantly improve the accuracy of fault prediction results and accurately predict nonlinear data. The method includes the following steps: S1. Obtain the raw data from the cableway monitoring system and preprocess the raw data to obtain aligned data. Existing technologies typically rely on manual calibration and simple filtering during preprocessing, which cannot adaptively process multi-source data and is inefficient and prone to errors. To address these issues, the specific implementation steps are as follows: S11. Clean data is obtained by anomaly detection and removal based on the raw data of the cableway monitoring system. The raw data of the cableway monitoring system includes raw data collected by sensors such as vibration, temperature, tilt angle, pressure, wind speed, and lightning strike. The mean and standard deviation are calculated based on the raw data using the mean and standard deviation formulas. The Z value is obtained by subtracting the mean from the raw data and dividing by the standard deviation. If the absolute value of the Z value is greater than 3, the raw data is removed; otherwise, the raw data is retained. Clean data is obtained through this anomaly detection and removal method. The mean and standard deviation formulas are commonly used statistical calculation formulas and will not be elaborated here.

[0019] S12. Based on the cleaned data, the mean interpolation method is used to obtain complete data. Based on the complete data, normalization processing is performed to obtain normalized data. Based on the mean of the two adjacent data in the cleaned data, the mean is inserted into the position of the missing data to make the cleaned data more complete. Based on the complete data, the MinMax normalization method is used to obtain normalized data. The MinMax normalization method is to subtract the minimum value of the complete data from the complete data and then divide by the difference between the maximum and minimum values ​​of the complete data to obtain normalized data.

[0020] S13. Aligned data is obtained by time series alignment processing based on normalized data. The aligned data is obtained by dividing the difference between the normalized data at times t-1 and t+1 by the difference between times t-1 and t+1, then multiplying by the current t value minus the difference at t-1, and adding the product to the current normalized data. This invention can be executed on edge computing nodes to clean, normalize, and align the original data to eliminate noise and anomalies, improve data quality, and provide more accurate data for subsequent steps.

[0021] S2. Construct a multi-type, multi-protocol information transmission network based on the alignment data and obtain the transmission data. Perform aggregation processing on the transmission data to obtain aggregated data. Existing technologies generally use a single network with limited coverage, high power consumption, or high latency. Furthermore, aggregation through fixed time has significant latency and is susceptible to outlier interference, leading to deviations in data transmission and processing. To address these issues, the specific steps are as follows: S21. Based on the alignment data, encapsulated data is obtained through data encapsulation processing; the alignment data is converted into JSON format, which includes multiple formats such as timestamps, sensor IDs, and data values, and forms a complete JSON data encapsulation package that is easy to parse and transmit, i.e., encapsulated data. JSON format is a lightweight data exchange format. It is widely used in data transmission and storage scenarios because of its simplicity, readability, and parsing. It will not be elaborated on here.

[0022] S22. Construct a multi-type, multi-protocol information transmission network based on encapsulated data and output the transmitted data; the multi-type network includes various network types such as 5G, Ethernet, and LoRaWAN. Among them, 5G network is suitable for high-speed, low-latency, and large-capacity information transmission, Ethernet is suitable for broadcast and shared media, and LoRaWAN is suitable for low-power, long-distance information transmission. It can adapt to data transmission in various situations, improve the efficiency and adaptability of data transmission, and the multi-protocol network includes network protocols such as LoRaWAN, TCP, and IP. LoRaWAN is also a low-power wide area network protocol, TCP is a transport layer protocol, and IP is a network layer protocol. The multi-protocol network information transmission can ensure the reliability of data transmission.

[0023] S23. Construct a time window of size A and sliding step b based on the transmitted data, and aggregate the transmitted data within the time window to obtain aggregated data; obtain aggregated data by summing the transmitted data within the time window and dividing by the number of transmitted data; this invention unifies the data format based on JSON encapsulation, and the hybrid network transmission can dynamically select the optimal path according to the scenario. Time window aggregation realizes real-time smooth processing of data, making data processing and transmission efficient, flexible and low cost.

[0024] S3. Based on aggregated data, a fused feature set is obtained through data fusion and feature extraction. Existing technologies cannot handle nonlinear relationships well, feature extraction is easily affected by noise, and feature data processing is inefficient and lacks objectivity. To solve the above problems, the specific implementation steps are as follows: S31. Combined data is obtained by splicing aggregated data. Aggregated data has undergone multiple processing steps on the basis of the original data, but still retains the dimensions of the original data. Here, data splicing is performed directly to obtain a data matrix, i.e., combined data. The number of spliced ​​aggregated data can be determined by grouping according to time, so that multiple data at the same time are grouped into one combined data. It can also be classified according to the sensors of the cableway monitoring system, which is beneficial to the implementation of feature extraction and feature enhancement in the later stage. The data splicing method is a commonly used method to fuse multiple data into a matrix, which will not be elaborated here.

[0025] S32. Feature data is obtained through feature extraction based on the combined data; the eigenvector matrix is ​​calculated based on the combined data using the covariance matrix formula, that is, the eigenvector matrix is ​​obtained by multiplying the transpose matrix of the combined data by the combined data and then dividing by the number of sample rows of the combined data. The eigenvector matrix is ​​then multiplied by the combined data to obtain the dimensionality-reduced feature data. This method is principal component analysis, an unsupervised linear dimensionality reduction and feature extraction method. It projects the original high-dimensional data into a low-dimensional space through orthogonal transformation, retaining the direction of the largest variance in the data, thereby reducing the data dimensionality while preserving as much original information as possible. This will not be elaborated further here.

[0026] S33. Obtain encoded features through feature encoding processing based on feature data; set initial encoding weight matrix and encoding bias matrix, propagate the feature data forward through the initial encoding weight matrix and encoding bias matrix, determine the activation function as ReLU function to output encoded data, use mean squared error as loss function to calculate the difference between feature data and encoded data and calculate gradient through backpropagation, update the initial encoding weight matrix and encoding bias matrix through gradient descent method to obtain updated encoding weight matrix and updated encoding bias matrix, then multiply the feature data by the updated encoding weight matrix and add the updated encoding bias matrix to obtain the sum value, input the sum value into the ReLU function to obtain encoded features.

[0027] S34. Construct a feature decoder based on feature encoding processing. Reconstruct the features by using the feature decoder based on the encoded features. The feature decoder is based on feature encoding processing, which transforms the encoding process into a decoding process. The steps are similar to those of feature encoding processing, namely, obtaining the decoding weight matrix and decoding bias matrix by combining gradient descent and backpropagation. Then, the decoded sum is obtained by multiplying the decoded weight matrix by the encoded features and adding the decoded bias matrix. The decoded sum is then input into the Sigmoid function to obtain the reconstructed features.

[0028] S35. Calculate the error value based on the reconstructed features and feature data, determine the error threshold, and filter the feature data according to the error threshold to obtain the fused feature set; If the error value is greater than or equal to the error threshold, the feature data corresponding to the reconstructed feature is removed; otherwise, it is retained. All feature data are filtered and then concatenated to obtain a fused feature set. The error value is obtained by the square of the difference between the reconstructed feature and the feature data, and the error threshold is obtained by multiplying the mean of the squares of the differences between the reconstructed feature and the feature data by 0.9. The data concatenation method is the same as that in S31, i.e., the data is directly concatenated. This invention improves the accuracy and robustness of feature extraction through PCA dimensionality reduction and autoencoder feature enhancement, effectively reducing data dimensionality while retaining key information. The concatenation method realizes the aggregation of original information from multiple sources, retains the integrity of heterogeneous data, and effectively reduces data dimensionality and removes redundant noise. The nonlinear coding capability is used to further mine higher-order feature representations and enhance the discriminative power of features. Based on the dynamic threshold selection of reconstruction error, the most representative feature components are adaptively retained, which can significantly improve the generalization ability and interpretability of subsequent steps.

[0029] S4. Based on the fused feature set, predict cableway fault conditions to obtain fault prediction results and cableway safety index. Existing technologies often result in significant discrepancies between predicted and actual fault conditions during cableway fault prediction, neglecting the importance of feature data in cableway processing. Furthermore, the nonlinear data in the raw data of the cableway monitoring system is difficult to predict accurately, leading to inaccurate prediction results. To address these issues, the specific implementation steps are as follows: S41. Obtain the time-series predicted value through time-series prediction processing based on the fused feature set; the specific implementation method is as follows: S411. Construct an LSTM processing layer. Input the elements of the fused feature set one by one into the LSTM processing layer to obtain the hidden state sequence. The LSTM processing layer is a special recurrent neural network with memory units and gating mechanisms. The gating mechanism includes an input gate, a forget gate, and an output gate. The processing flow is as follows: first, the forget gate determines which old information in the memory unit to discard; then, the input gate selectively adds new information to the memory unit; then, based on the forgotten old information and the new input information, the state of the memory unit is updated; finally, the output gate determines which information in the memory unit needs to be passed to the hidden state of the next time step, i.e., the hidden state sequence. LSTM is a commonly used neural network model, which will not be elaborated here.

[0030] S412. Calculate the attention score based on the hidden state sequence, and normalize the attention score to obtain the attention weight. Multiply the hidden state sequence at time t by parameter a, and add the hidden state sequence at the last time step by parameter b. Process the sum using the tanh function to obtain the T function value. Initialize a parameter vector using a uniform distribution. Multiply the parameter vector by the T function value to obtain the attention score. Exp process the attention score to obtain the e value. Divide the e value by the sum of all e values ​​to obtain the attention weight. Parameters a and b can be obtained through a random generator. Parameters a, b, and the parameter vector can be optimized by constructing an optimization layer using a particle swarm optimization algorithm. That is, parameters a, b, and the parameter vector are treated as individuals in a particle swarm for optimization iteration, and finally, the optimal individual is output as the optimized parameters a, b, and the parameter vector.

[0031] S413. Obtain the context vector by weighted summation based on the attention weights and the hidden state sequence; S414. Initial weight vector and initial bias vector are obtained based on normal distribution initialization. Initial prediction output layer is constructed based on weight vector and bias vector. Normal distribution initialization is a commonly used method to obtain initialized weight vector and bias vector, which will not be elaborated here. The initial prediction output layer is the prediction model obtained by multiplying the context vector by the initial weight vector and then adding the initial bias vector.

[0032] S415. Construct a particle swarm optimization layer using the initial weight vector and initial bias vector as individuals. Obtain the optimized weight vector and optimized bias vector through iterative optimization, and construct an optimization prediction layer. The particle swarm optimization layer is based on the particle swarm optimization algorithm, and iteratively optimizes individuals. The position and velocity of each generation in the population are updated and iterated through inertia weights and random numbers, and finally finds a globally optimal individual, namely the optimized weight vector and optimized bias vector. The particle swarm optimization algorithm is a commonly used optimization method, which will not be elaborated here. The content of the optimization prediction layer is the product of the context vector and the optimized weight vector plus the optimized bias vector.

[0033] S416. Input the context vector into the optimized prediction layer to obtain the time series prediction value; S42. Construct a random forest model based on the fused feature set, and obtain the feature importance score through the random forest model. First, input each element in the fused feature set into the random forest model. The model performs ensemble learning by constructing a large number of decision trees. During training, the model quantifies the importance of each element by calculating the Gini index of each element in the node splits of all trees. Finally, take the average of the feature importance scores of all trees to obtain the feature importance score of each element. The random forest model is a commonly used model for obtaining feature importance, and will not be elaborated on here.

[0034] S43. The cableway safety index is obtained by weighted summation based on the time-series predicted values ​​and feature importance scores; S44. Calculate the failure probability based on the cableway safety index using logistic regression; obtain the logistic regression parameters through maximum likelihood estimation. The logistic regression parameters include the intercept term and the characteristic coefficient. The index value is obtained by multiplying the cableway safety index by the characteristic coefficient and adding the sum of the intercept term, taking the negative number, and raising the negative number to the natural constant e. The failure probability is obtained by dividing 1 by the sum of the index value and 1.

[0035] S45. Generate fault prediction results and cableway safety index based on fault probability; the fault prediction results include cableway non-fault results and cableway fault results. Obtain historical data of the cableway monitoring system, calculate the average fault probability based on the historical data, multiply by 0.8 to obtain the fault threshold. If the fault probability is less than or equal to the fault threshold, generate a cableway non-fault result; otherwise, generate a cableway fault result. This invention utilizes time-series dependency to capture dynamic features and can provide interpretation of feature importance. The weighted fusion mechanism organically combines time-series prediction and feature contribution to generate a health index. It can also map continuous health status to quantifiable fault probabilities. Finally, real-time early warning is achieved through threshold triggering, improving the interpretability of model output results and supporting independent optimization and fault tracing, significantly improving prediction accuracy and the reliability of operation and maintenance decisions.

[0036] S5. Based on the fault prediction results and the cableway safety index, construct a digital twin synchronization model. Simulate and detect anomalies in the digital twin synchronization model data to obtain simulation data and optimization strategies. Existing twin models rely on purely physical models, which are computationally complex and lack real-time synchronization adjustments. To address these issues, the specific implementation steps are as follows: S51. Based on the fault prediction results and the cableway safety index, a digital twin model is obtained through physical modeling. Physical modeling can be achieved through RecurDyn. The General Constraint function is used to define the rope length and distance constraints between connected objects, replacing the traditional finite element method. RecurDyn is a multibody dynamics simulation software that is often used for the construction of digital twin models, and will not be elaborated on here.

[0037] S52. Collect real-time fusion feature sets, and assimilate the data using Kalman filtering based on the digital twin model and the real-time fusion feature sets to obtain a synchronization model. Kalman filtering is a data synchronization process based on the predicted data in the digital twin model using the Kalman filtering formula. The Kalman filtering formula consists of three formulas: state prediction, covariance prediction, and state update. It is a recursive linear optimal estimation method. By combining the digital twin model with the real-time fusion feature sets, the state of the digital twin model can be estimated and predicted in real time under the presence of noise interference. This will not be elaborated on here.

[0038] S53. Obtain the synchronization data in the synchronization model, and obtain the simulation data through simulation processing based on the synchronization data. The synchronization data in the synchronization model is the real-time data updated in real time in the digital twin model. However, the real-time data is processed by Kalman filtering, so it is synchronization data. The simulation processing can be implemented by the Runge-Kutta method. The Runge-Kutta method is a type of numerical integration method used to solve the initial value problem of ordinary differential equations ODE. It approximates the true solution by performing multiple function evaluations, such as slope calculation, in each time step, combined with weighted averaging. It will not be elaborated here.

[0039] S54. Construct a GAN discriminator to obtain the anomaly score of the simulation data. The GAN discriminator is obtained by multiplying the simulation data by the discriminator coefficient and adding the discriminator bias term to obtain the processed data value. The processed data value is then used to calculate the true probability through the sigmoid function. The anomaly score is obtained by subtracting the true probability from 1.

[0040] S55. Set an anomaly threshold and generate an optimization strategy based on the anomaly score and the anomaly threshold. If the anomaly score is greater than or equal to the anomaly threshold, an optimization strategy is generated; otherwise, it indicates normal operation, and the process proceeds to step S6. The anomaly threshold is obtained by summing the anomaly scores and dividing by the number of anomaly scores. Alternatively, the anomaly threshold can be set directly based on historical data. In practice, the anomaly threshold is generally set to 0.7. If the anomaly score is greater than or equal to 0.7, an optimization strategy is generated, such as adjusting the speed of the cableway drive wheel. Otherwise, it indicates normal operation, no optimization strategy is given, and the process proceeds to the next step. This invention establishes a digital twin model through fault prediction and safety index, providing a precise foundation for subsequent analysis. A synchronous model is obtained through Kalman filtering data assimilation, which can reflect the cableway status in real time. Simulation processing acquires data, and combined with a GAN discriminator, anomaly scores are obtained, enabling accurate anomaly detection. Setting an anomaly threshold generates an optimization strategy, which can respond to abnormal situations in a timely manner and ensure the safe and stable operation of the cableway.

[0041] S6. Generate early warning data based on simulation data and optimization strategies, and send the early warning data to the control center. Existing technologies often have a disconnect between analysis results and control systems in practice, requiring staff to interpret and operate them, which can easily lead to system response delays and execution deviations. To address these issues, the specific solution is as follows: S61. Based on the simulation data and optimization strategy, a structured list is obtained through format standardization. The simulation data includes the load of the cableway or the physical data of the drive wheel, and the optimization strategy includes the adjustment of the physical data. These are all unstructured or semi-structured texts. This needs to be converted into a program-readable structured list through format standardization, such as target components, current state, target action, specific parameters, etc. This can be done through text processing using Python, which is a commonly used method for structured text conversion, and will not be elaborated here.

[0042] S62. Obtain control commands through instruction mapping based on the structured list; generate an instruction mapping table based on historical data. For example, if the current state of the target component and the target action are in the same range, the corresponding control command will be generated. That is, if the current state of the drive wheel is faulty and the rotation speed exceeds the abnormal threshold, an optimization strategy to reduce the speed will be generated, and then the control command will generate a control command to reduce the speed.

[0043] S63. The control commands, optimization strategies, and simulation data are merged into early warning data, and the early warning data is transmitted to the control center. This invention receives the simulation conclusions and optimization suggestions generated by the digital twin model, formats them into standard commands, generates early warning information, and distributes it to the corresponding control system. Its advantage lies in directly and efficiently implementing intelligent decision-making and achieving seamless connection from analysis to control execution.

[0044] Example 2 Due to the inherent technical problems of inaccurate predictions caused by significant discrepancies between existing fault prediction methods and actual results, and the difficulty in accurately predicting nonlinear data, please refer to [link to relevant documentation]. Figure 2 The diagram shown is a structural block diagram of the cableway monitoring and management system based on artificial intelligence and edge computing provided in this embodiment. The system includes: a preprocessing module, an aggregation module, a fusion module, a prediction module, a digital twin module, and a control module.

[0045] The system comprises the following modules: a preprocessing module to acquire raw data from the cableway monitoring system and preprocess it to obtain aligned data; an aggregation module to construct a multi-type, multi-protocol information transmission network based on the aligned data and obtain transmitted data, and then aggregate the transmitted data to obtain aggregated data; a fusion module to obtain a fusion feature set based on the aggregated data through data fusion and feature extraction; a prediction module to predict cableway fault conditions based on the fusion feature set and obtain fault prediction results and a cableway safety index; a digital twin module to construct a digital twin synchronization model based on the fault prediction results and the cableway safety index, and then perform simulation and anomaly detection on the digital twin synchronization model data to obtain optimization strategies; and a control module to generate early warning data based on simulation data and optimization strategies, and then send the early warning data to the control center.

[0046] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0047] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A cableway monitoring and management method based on artificial intelligence and edge computing, characterized in that, The method includes the following steps: S1. Obtain the raw data from the cableway monitoring system and preprocess the raw data to obtain aligned data; S2. Construct a multi-type, multi-protocol information transmission network based on the alignment data and obtain the transmission data. Then, perform aggregation processing on the transmission data to obtain aggregated data. S3. Based on aggregated data, a fused feature set is obtained through data fusion and feature extraction; S4. Based on the fusion feature set, predict cableway failure conditions to obtain failure prediction results and cableway safety index; S5. Based on the fault prediction results and the cableway safety index, construct a digital twin synchronization model, and perform simulation and anomaly detection on the data of the digital twin synchronization model to obtain simulation data and optimization strategies. S6. Generate early warning data based on simulation data and optimization strategies, and send the early warning data to the control center.

2. The cableway monitoring and management method based on artificial intelligence and edge computing according to claim 1, characterized in that, In S1, the specific implementation steps are as follows: S11. Based on the raw data of the cableway monitoring system, anomaly detection and removal are performed to obtain cleaned data; S12. Based on the cleaned data, the mean interpolation method is used to process the data to obtain complete data, and the complete data is then normalized to obtain normalized data. S13. Obtain aligned data by time series alignment processing based on normalized data.

3. The cableway monitoring and management method based on artificial intelligence and edge computing according to claim 1, characterized in that, In S2, the specific implementation steps are as follows: S21. Obtain encapsulated data by data encapsulation processing based on the alignment data; S22. Construct a multi-type, multi-protocol information transmission network based on encapsulated data and output the transmitted data; S23. Construct a time window of size A and sliding step b based on the transmitted data, and aggregate the transmitted data within the time window to obtain aggregated data.

4. The cableway monitoring and management method based on artificial intelligence and edge computing according to claim 1, characterized in that, In S3, the specific implementation steps are as follows: S31. Obtain combined data by concatenating aggregated data; S32. Obtain feature data by feature extraction based on the combined data; S33. Obtain coded features by feature encoding processing based on feature data; S34. Construct a feature decoder based on feature encoding processing, and reconstruct the reconstructed features using the feature decoder based on the encoded features; S35. Calculate the error value based on the reconstructed features and feature data, determine the error threshold, and filter the feature data according to the error threshold to obtain the fused feature set; If the error value is greater than or equal to the error threshold, the feature data corresponding to the reconstructed feature is removed; otherwise, it is retained. All feature data are filtered and then concatenated to obtain the fused feature set.

5. The cableway monitoring and management method based on artificial intelligence and edge computing according to claim 1, characterized in that, In S4, the specific implementation steps are as follows: S41. Obtain the time-series prediction value by processing the fused feature set through time-series prediction. S42. Construct a random forest model based on the fused feature set, and obtain the feature importance score through the random forest model; S43. The cableway safety index is obtained by weighted summation based on the time-series predicted values ​​and feature importance scores; S44. Calculate the failure probability using logistic regression based on the cableway safety index; S45. Generate fault prediction results and cableway safety index based on fault probability.

6. The cableway monitoring and management method based on artificial intelligence and edge computing according to claim 5, characterized in that, In S41, the specific implementation steps are as follows: S411. Construct an LSTM processing layer and input the elements of the fused feature set one by one into the LSTM processing layer to obtain the hidden state sequence. S412. Calculate the attention score based on the hidden state sequence, and normalize the attention score to obtain the attention weight. S413. Obtain the context vector by weighted summation based on the attention weights and the hidden state sequence; S414. Initial weight vector and initial bias vector are obtained based on normal distribution initialization. Initial prediction output layer is constructed based on weight vector and bias vector. S415. The initial weight vector and initial bias vector are used as individuals to construct a particle swarm optimization layer. The optimized weight vector and optimized bias vector are obtained through iterative optimization, and an optimized prediction layer is constructed. S416. Input the context vector into the optimized prediction layer to obtain the time series prediction value.

7. The cableway monitoring and management method based on artificial intelligence and edge computing according to claim 1, characterized in that, In S5, the specific implementation steps are as follows: S51. Based on the fault prediction results and the cableway safety index, a digital twin model is obtained through physical modeling; S52. Collect the real-time fusion feature set, and assimilate the data by Kalman filtering based on the digital twin model and the real-time fusion feature set to obtain the synchronization model. S53. Obtain the synchronization data in the synchronization model, and obtain the simulation data through simulation processing based on the synchronization data; S54. Construct a GAN discriminator and obtain the anomaly score of the simulation data through the GAN discriminator; S55. Set an anomaly threshold and generate an optimization strategy based on the anomaly score and the anomaly threshold. If the anomaly score is greater than or equal to the anomaly threshold, an optimization strategy is generated. Conversely, if it is normal, proceed to S6.

8. The cableway monitoring and management method based on artificial intelligence and edge computing according to claim 1, characterized in that, In S6, the specific implementation steps are as follows: S61. Obtain a structured list by standardizing the format based on simulation data and optimization strategies; S62. Obtain control instructions through instruction mapping based on the structured list; S63. Combine control commands, optimization strategies, and simulation data into early warning data, and transmit the early warning data to the control center.

9. A system applied to the cableway monitoring and management method based on artificial intelligence and edge computing as described in any one of claims 1-8, characterized in that, The system includes: The preprocessing module is used to acquire the raw data from the cableway monitoring system and preprocess the raw data to obtain aligned data. The aggregation module is used to construct a multi-type, multi-protocol information transmission network based on the alignment data and obtain the transmission data, and to perform aggregation processing on the transmission data to obtain aggregated data. The fusion module is used to obtain a fused feature set based on aggregated data through data fusion and feature extraction; The prediction module is used to predict cableway fault conditions based on the fused feature set, and obtain the fault prediction results and cableway safety index. The digital twin module is used to construct a digital twin synchronization model based on fault prediction results and cableway safety index, and to simulate and detect anomalies in the digital twin synchronization model data to obtain optimization strategies. The control module is used to generate early warning data based on simulation data and optimization strategies, and then send the early warning data to the control center.

Images