Escalator safety performance detection method and device, medium and equipment

Abstract

The invention discloses an escalator safety performance detection method and device, a medium and equipment. The method comprises the steps of obtaining operation monitoring data of an escalator; preprocessing the operation monitoring data; constructing an escalator safety performance detection model, and training the model; and inputting the preprocessed operation monitoring data into a trained escalator safety performance detection model, and outputting a safety index of the escalator. The safety index is obtained through calculation, and the safety performance detection precision of the escalator can be improved.

Description

Technical Field

[0001] The present application belongs to the field of elevator engineering technology, and specifically relates to a method, device, medium and equipment for detecting the safety performance of an escalator. Background Art

[0002] With the acceleration of urbanization, escalators, as an important part of modern public transportation facilities, are widely used in public places such as shopping malls and subway stations. However, safety issues of escalators also arise, such as accidental stops and step misalignment, which bring safety hazards to passengers. Therefore, it is very important to establish an efficient and reliable escalator safety performance testing method to prevent escalator accidents. Summary of the invention

[0003] In view of the deficiencies in the prior art, the purpose of the present application is to provide a method, device, medium and equipment for detecting the safety performance of an escalator. The present application aims to accurately and efficiently detect the safety performance of an escalator.

[0004] To achieve the above objectives, this application provides the following technical solutions:

[0005] An escalator safety performance detection method, the method comprising: obtaining operation monitoring data of an escalator; preprocessing the operation monitoring data; constructing an escalator safety performance detection model and training the model; the escalator safety performance detection model performs multi-scale time window sampling on the operation monitoring data and uses adaptive weight allocation to capture short-term transient events and long-term trend changes in the operation of the escalator; the preprocessed operation monitoring data is input into the trained escalator safety performance detection model, and the safety index of the escalator is output.

[0006] Optionally, the operation monitoring data is preprocessed, including: performing data cleaning on the operation monitoring data; performing data transformation on the operation monitoring data after data cleaning; and performing data reduction on the operation monitoring data after data transformation.

[0007] Optionally, the escalator safety performance detection model includes: an input layer, a feature extraction layer, a feature analysis layer and an output layer connected in sequence, wherein the input layer is used to receive the preprocessed operation monitoring data; the feature extraction layer is used to extract features from the preprocessed operation monitoring data; the feature analysis layer is used to analyze the extracted features to evaluate the immediate risk of the escalator and predict the long-term operation trend, and calculate the safety index of the escalator based on the risk assessment results and the long-term operation trend prediction; the output layer is used to output the safety index.

[0008] Optionally, the feature extraction layer includes: a multi-scale time window sampling unit, an adaptive weight allocation unit, a local feature generation unit, a global context fusion unit, a feature selection and enhancement unit, a feature mapping unit and an output interface connected in sequence.

[0009] Optionally, the feature analysis layer includes: an immediate risk assessment module, a long-term trend prediction module and a safety evaluation module connected in sequence, wherein the immediate risk assessment module is used to quickly identify possible risk factors of the escalator; the long-term trend prediction module is used to predict the potential risk change trend of the escalator in the future; the safety evaluation module is used to comprehensively analyze immediate risks, long-term trends and environmental factors, and calculate the safety index.

[0010] Optionally, the escalator safety performance detection model is trained through the following steps: obtaining historical data of the escalator and preprocessing it, and dividing the preprocessed historical data into a training set and a test set; setting training parameters, and using the training set to train the escalator safety performance detection model until the training meets the maximum number of iterations; using the test set to verify the trained model, during the verification process, if the mean absolute error MAE, mean square error MSE, and root mean square error RMSE, which are model performance evaluation indicators, are all less than a threshold, the model verification passes; otherwise, the training parameters are adjusted or the training set samples are expanded to retrain the model until the verification passes.

[0011] The present application also provides an escalator safety performance detection device, which includes: an acquisition module for acquiring operation monitoring data of the escalator; a preprocessing module for preprocessing the operation monitoring data; a model construction and training module for constructing an escalator safety performance detection model and training the model; an output module for inputting the preprocessed operation monitoring data into the trained escalator safety performance detection model and outputting the safety index of the escalator.

[0012] Optionally, the preprocessing module includes: a cleaning submodule for cleaning the operation monitoring data; a transformation submodule for transforming the operation monitoring data after data cleaning; and a reduction submodule for reducing the operation monitoring data after data transformation.

[0013] The present application also provides a storage medium, which includes instructions, and when the instructions are executed on a computer, the computer executes the method as described in any of the preceding items.

[0014] The present application also provides an electronic device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements any of the above methods when executing the program.

[0015] Compared with the prior art, the beneficial effects brought about by the present application are as follows: by constructing an escalator safety performance detection model, the present application can accurately reflect the safety status of the escalator by calculating the safety index of the activity, thereby significantly improving the accuracy and reliability of the escalator safety performance detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flowchart of an escalator safety performance detection method provided by an embodiment of the present application;

[0017] Figure 2 is a structural schematic diagram of an escalator safety performance detection model provided by another embodiment of the present application;

[0018] Figure 3 yes Figure 1 Schematic diagram of the structure of the feature extraction layer in the shown model;

[0019] Figure 4 is a structural schematic diagram of an escalator safety performance detection device provided by another embodiment of the present application;

[0020] Figure 5 is a schematic diagram of the structure of a storage medium provided by another embodiment of the present application;

[0021] Figure 6 It is a schematic diagram of the structure of an electronic device provided in another embodiment of the present application. DETAILED DESCRIPTION

[0022] The specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. Although the specific embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0023] It should be noted that certain words are used in the specification and claims to refer to specific components. Those skilled in the art should understand that technicians may use different nouns to refer to the same component. This specification and claims do not use the difference in nouns as a way to distinguish components, but use the difference in the functions of the components as the criterion for distinction. As mentioned throughout the specification and claims, "including" or "comprising" is an open term, so it should be interpreted as "including but not limited to". The subsequent description of the specification is a preferred embodiment of the implementation of the present application, but the description is based on the general principles of the specification and is not used to limit the scope of the present application. The scope of protection of the present application shall be determined by the attached claims.

[0024] To facilitate understanding of the embodiments of the present application, further explanation will be given below using specific embodiments as examples in conjunction with the accompanying drawings, and the various drawings do not constitute a limitation on the embodiments of the present application.

[0025] Figure 1 FIG. 1 is a flow chart of a method for detecting the safety performance of an escalator provided by an embodiment of the present application, such as Figure 1 As shown, the method comprises the following steps:

[0026] S100: Acquire operation monitoring data of the escalator, wherein the operation monitoring data includes mechanical performance data (for example, the operating speed of the escalator, the speed change rate when starting and stopping, the inclination angle, the step gap, and the step chain tension), dynamic response data (for example, the vibration of the escalator during operation, the noise level, and the motor current) and environmental data (for example, the ambient temperature, the operating temperature of the motor and the reducer, and the ambient humidity).

[0027] S200: Preprocessing the operation monitoring data;

[0028] S300: Build an escalator safety performance detection model and train the model;

[0029] S400: Input the pre-processed operation monitoring data into a trained escalator safety performance detection model, and output the escalator safety index.

[0030] By constructing an escalator safety performance detection model, this application can not only provide higher safety performance detection accuracy, but also identify potential failures and risk trends in advance through in-depth analysis of operation data, thereby achieving more accurate and timely safety assessments and preventive maintenance.

[0031] In another exemplary embodiment, in step S200, preprocessing the operation monitoring data includes the following steps:

[0032] S201: Cleaning the operation monitoring data, specifically including:

[0033] First, you need to identify and remove or correct obviously unreasonable data points. For example, if a sensor continues to report impossible values ​​(such as temperatures below absolute zero) for a period of time, then these records should be marked as invalid and removed from the data set. Secondly, for missing data points, you can choose a filling strategy, such as using the average or median of the previous and next time points, or using interpolation to estimate missing values; you can also choose to delete data rows or columns with missing values. Finally, you need to use statistical methods to find data points that deviate from the normal range and choose to retain, correct or delete them.

[0034] S202: performing data transformation on the operation monitoring data after data cleaning;

[0035] In this step, the numerical data needs to be adjusted to the range of [0, 1] to ensure that data of different scales will not have a disproportionate impact on the model; for non-numerical data, it needs to be converted into a form that the model can handle, such as one-hot encoding.

[0036] S203: performing data reduction on the operation monitoring data after data transformation.

[0037] In this step, methods such as principal component analysis and linear discriminant analysis can be used to reduce the dimension of the feature space, thereby reducing the computational complexity of the model and preventing overfitting.

[0038] In another exemplary embodiment, in step S300, Figure 2 As shown, the escalator safety performance detection model includes an input layer, a feature extraction layer, a feature analysis layer and an output layer connected in sequence, wherein the input layer is used to receive the preprocessed operation monitoring data; the feature extraction layer is used to extract features from the preprocessed operation monitoring data; the feature analysis layer is used to analyze the extracted features to evaluate the immediate risk of the escalator and predict the long-term operation trend, and calculate the safety index ESI of the escalator based on the risk assessment results and the long-term operation trend prediction; the output layer is used to output the safety index ESI.

[0039] In this embodiment, Figure 3 As shown, the feature extraction layer includes a multi-scale time window sampling unit, an adaptive weight allocation unit, a local feature generation unit, a global context fusion unit, a feature selection and enhancement unit, a feature mapping unit and an output interface connected in sequence. The multi-scale time window sampling unit is used to segment the pre-processed operation monitoring data according to time windows of different lengths, such as short time windows (such as 5 seconds), medium time windows (such as 30 seconds) and long time windows (such as 5 minutes), and each time window corresponds to a group of sensor readings or status records.

[0040] The multi-scale time window sampling unit is expressed as follows:

[0041]

[0042] Among them, T * i represents the optimal time window size, T i represents one of the candidate time window sizes, T s represents the set of all possible time window sizes, s represents the data stream or environmental conditions observed so far, Q(s,Ti ; θ) represents the Q value function, and the prediction takes the time window T i The expected return depends on the state s and the parameters θ of the Q network.

[0043] The safety performance of escalators may be affected by factors on multiple time scales, such as short-term transient events (such as sudden stops or accelerations) and long-term trend changes (such as gradual performance degradation caused by equipment aging). By using time windows of different lengths, these short-term and long-term change patterns can be captured simultaneously. In addition, different time windows can help the model learn richer feature representations, thereby improving its generalization ability. Data from shorter time windows helps to identify rapidly changing local features, while longer time windows can better reflect overall trends, allowing the model to make accurate predictions when faced with unseen data.

[0044] Furthermore, the data of each time window is fed into the adaptive weight allocation unit, which dynamically adjusts the proportion of its contribution to the overall feature by calculating the importance of each time point. The adaptive weight allocation unit is expressed as follows:

[0045]

[0046] Among them, w(t) represents the importance weight of time point t, softmax(·) represents the activation function, and q t represents the query vector, T represents the transpose of the query vector, k represents the key vector, d k Indicates the dimension of the key vector.

[0047] In addition, within each time window, the present application uses a local feature generation unit to replace the traditional convolution operation, and the local feature generation unit is expressed as follows:

[0048]

[0049] Among them, F(xT i ) represents the time window T obtained by combining multiple transformation methods i The internal features, f j (·) represents the jth transformation method, such as wavelet transform, Hilbert transform, etc., a j represents the weight of the jth transformation method, xT i Represents the time window T i The sensor reading sequence within.

[0050] Traditional convolution operations usually use fixed filters (or kernels), which are general feature extraction tools trained through a large number of images or other types of data. Although traditional convolution operations perform well in areas such as processing natural images, they cannot capture all relevant features most effectively for non-image data in specific application scenarios (such as time series data). The local feature generation unit can focus on extracting immediately useful features, which can respond more quickly to changes in the operating status of the escalator, thereby improving the accuracy of fault feature identification. In addition, the local feature generation unit can also help detect signs of slowly developing faults, such as increased wear or aging of mechanical parts, which helps to plan maintenance work in advance and avoid service interruptions caused by sudden failures.

[0051] The global context fusion unit is expressed as follows:

[0052]

[0053] Among them, S i (k+1) represents the updated state of node i after the k+1th iteration, σ(·) represents the activation function, such as ReLU, W (k) represents the weight matrix at the kth iteration, S i (k) represents the state of node i at the kth iteration, S j (k) represents the state of node j at the kth iteration, A ij Represents the adjacency matrix element, indicating the degree of association between nodes i and j.

[0054] The global context fusion unit can cross the local time window and combine the context information in the long time span with the features of the current time point. Therefore, the unit not only focuses on the local changes in the short time window, but also captures the long-term trends and development patterns in the entire time series. In this way, the model can understand the operating status of the escalator more comprehensively, not only identifying immediate risk factors, but also predicting potential problems in the future, so it can help to find those slowly developing fault signs in advance. In addition, the global context fusion unit can achieve effective cross-modal fusion of various types of data sources (such as mechanical properties, dynamic responses, environmental conditions, etc.) at the feature level, allowing data from different sources to complement and support each other, thereby forming a more complete and consistent view, so that the model can examine the safety performance of the escalator from multiple perspectives and improve the reliability and comprehensiveness of the overall evaluation. For example, by associating the working status of mechanical components with environmental factors (such as humidity and temperature), the aging degree or corrosion risk of the equipment can be more accurately assessed.

[0055] The feature selection and enhancement unit is used to automatically select the most valuable feature combination and perform enhancement processing on the selected features. The feature selection and enhancement unit is expressed as follows:

[0056] F e '=F'+εG(z),z~p z (z)

[0057] Among them, F e ' represents the feature subset enhanced by adversarial training, F' represents the optimal feature subset selected by Bayesian optimization, ε represents the perturbation intensity coefficient, which is used to control the degree of added noise, G(·) represents the generator, which is responsible for generating adversarial samples from the latent variable z, and p z (z) represents the prior distribution of the latent variable z, such as a Gaussian distribution.

[0058] The feature selection and enhancement unit first applies the Bayesian optimization algorithm to search and evaluate the model performance under different feature combinations among multiple candidate features including running speed, vibration, motor current, operating temperature, and ladder chain tension. By constructing a proxy model (such as Gaussian process regression), Bayesian optimization can efficiently explore the feature space and find those feature combinations that can best improve prediction accuracy and stability to form the optimal feature subset; in order to further enhance the expressiveness of these key features, the feature selection and enhancement unit introduces the generator G(·) in the generative adversarial network. The generator G(·) is derived from the prior distribution p of the latent variable z. z (z) to generate adversarial samples X adv =G(z), these adversarial samples are the result of adding controlled noise to the original features, aiming to simulate various disturbances that may be encountered in the real world. The feature selection and enhancement unit selects the optimal feature subset through Bayesian optimization and the feature X enhanced by adversarial training. adv The enhanced feature set not only captures the key indicators of escalator safety performance, but also enhances the model's ability to identify and robustness to abnormal situations.

[0059] The feature mapping unit is used to project the high-dimensional feature space into the low-dimensional space. The feature mapping unit includes a multi-scale feature extraction layer, a manifold learning and autoencoder combination layer, an attention mechanism layer and a final feature mapping layer connected in sequence. The multi-scale feature extraction layer is used to capture feature information at different scales, which is specifically expressed as follows:

[0060] h 1 =f multi -scale(X)

[0061] Among them, h 1represents the intermediate representation after multi-scale feature extraction, X represents the original high-dimensional feature vector, and f multi -scale(·) represents a multi-scale feature extraction function.

[0062] The manifold learning and autoencoder combination layer compresses features through the autoencoder and combines the manifold learning method to ensure that the low-dimensional representation retains the intrinsic geometric structure of the data, wherein the encoder is used to convert the high-dimensional feature h 1 Compress to a low-dimensional representation h 2 , the decoder is used to reconstruct the original input The manifold learning and autoencoder combination layer is specifically expressed as follows:

[0063] h 2 =f AE (h 1 θ enc )

[0064]

[0065] Among them, h 2 represents the low-dimensional representation after encoder compression, f AE (·;θ enc ) represents the encoder part of the autoencoder, θ enc represents the parameter geometry of the encoder, including weight matrices and bias vectors, etc. represents the approximate original input reconstructed by the decoder, g AE (·;θ dec ) represents the decoder part of the autoencoder, θ dec Represents the set of decoder parameters, including weight matrix and bias vector.

[0066] In order to further enhance the model's focus on key features, 2 The self-attention mechanism is applied above, which is specifically expressed as follows:

[0067] h 3 =f attention (h 2 θ att )

[0068] Among them, f attention (·;θ att ) represents the self-attention mechanism function, θ att Represents the set of parameters of the attention mechanism.

[0069] Finally, the feature h after the above processing 3 As the final low-dimensional feature z output, it is specifically expressed as follows:

[0070] z=f final(h 3 θ final )

[0071] Among them, z represents the final low-dimensional representation, f final (·;θ final ) represents the function of final adjustment of feature distribution and scale, θ final A collection of parameters representing the final mapping layer.

[0072] The feature mapping unit captures changes on different time scales through multi-scale feature extraction, combines manifold learning with autoencoders to retain the intrinsic geometric structure of the data and remove redundant information, uses the attention mechanism to dynamically adjust the feature weights to highlight key information, and optimizes the low-dimensional representation through the final feature mapping layer. This design not only improves the accuracy of fault prediction, enhances the robustness and generalization ability of the model, but also optimizes the utilization of computing resources, supports multimodal data analysis and continuous improvement, and thus provides efficient and reliable technical guarantees for intelligent monitoring and preventive maintenance.

[0073] The feature analysis layer includes an immediate risk assessment module, a long-term trend prediction module and a safety assessment module connected in sequence, wherein the immediate risk assessment module is used to quickly identify the possible risk factors of the escalator, specifically including a self-attention mechanism layer, a multi-head self-attention mechanism, a feedforward neural network (FFNN) and an output layer, wherein the self-attention mechanism layer is used to capture the long-term and short-term dependencies in the time series to help the model focus on key time points. The multi-head self-attention mechanism is used to enhance the model's ability to focus on different feature dimensions. The feedforward neural network is used for nonlinear transformation to increase the expression ability. The output layer is used to give an immediate risk score, indicating the risk level of the escalator at the current moment. The immediate risk assessment module also uses a residual connection, that is, the original input information of the input self-attention mechanism layer is directly passed to the output layer. Through the residual connection, the gradient disappearance problem can be avoided.

[0074] The instant risk assessment module is specifically represented as follows:

[0075] R(t)=FFNN(MultiHead(F(xT i ))+F(xT i ))

[0076] Among them, R(t) represents the output instant risk score, which represents the risk level at time t, and F(xT i ) represents the time window T i The feature vector generated by the data in, MultiHead(·) represents the multi-head self-attention mechanism, FFNN(·) represents the feedforward neural network, +F(xT i ) represents a residual connection.

[0077] The instant risk assessment module introduces a self-attention mechanism, which enables the model to capture long-term and short-term dependencies in the time series and helps it focus on key time points. The instant risk assessment module introduces a multi-head self-attention mechanism, which allows the model to focus on multiple different feature dimensions at the same time, enhancing the ability to understand complex patterns. This allows the model to not only identify a single type of anomaly, but also comprehensively consider the impact of multiple factors, thereby more comprehensively assessing risks. The introduction of a feedforward neural network helps capture the complex relationship between input features, allowing the model to better adapt to risk assessment needs in various situations.

[0078] The long-term trend prediction module is used to predict the potential risk change trend of the escalator in the future. The module includes a graph neural network, an LSTM unit and an attention pooling layer. The long-term trend prediction module is expressed as follows:

[0079] P(t+n)=LSTM(AttentionPool(GNN({F(xT 1 ),...,F(xT n )})))

[0080] Among them, P(t+n) represents the risk index predicted at the future time point t+n, {F(xT 1 ),...,F(xT n )} represents the feature set of multiple time windows, each F(xT i ) represents a feature vector within a time window, GNN(·) represents a graph neural network, AttentionPool(·) represents an attention pooling function, and LSTM(·) represents a long short-term memory network unit.

[0081] Graph neural networks are good at processing data with complex topological structures, such as the relationships between different components in escalators. Graph neural networks can transfer information between nodes (such as sensors and mechanical parts) and capture the dependencies between these components. LSTM can identify potential trends and development patterns of escalators during operation and help predict future performance changes. By combining graph neural networks and LSTM, the long-term trend prediction module can not only process the information flow in complex network structures, but also effectively capture time series, so that the model can more comprehensively understand the operating status of the escalator system, thereby improving the accuracy of long-term trend prediction. In addition, by introducing the attention pooling layer, the model's attention to key events can be improved, such as equipment maintenance records, the time points of major failures, etc., thereby providing a more accurate and reliable long-term trend prediction for the safety performance test of escalators.

[0082] The safety evaluation module is used to comprehensively analyze immediate risks, long-term trends and environmental factors, and calculate the safety index ESI. The safety evaluation module includes a multimodal fusion layer, an adaptive weighted network, an interactive decision unit, an explanatory output layer and a feedback loop mechanism connected in sequence. The multimodal fusion layer introduces a cross-attention mechanism to enable data from different sources (immediate risks, long-term trends, environmental factors) to interact with each other and enhance information exchange. The multimodal fusion layer is specifically represented as follows:

[0083] M=CrossAttn(R(t),P(t+n),E(t))

[0084] Among them, M represents the fused multimodal feature representation, which captures the relationship between immediate risk, long-term trend and environmental factors. CrossAttn(·) represents the cross-attention mechanism function, which is used to calculate the correlation between different input sources. R(t) represents the immediate risk score, which indicates the risk level at the current time t. P(t+n) represents the risk index at the future time point t+n output by the long-term trend prediction module. E(t) represents environmental data, including influencing factors such as temperature and humidity.

[0085] The multimodal fusion layer can ensure that the model's assessment of the safety status of the escalator is more comprehensive and accurate by fusing multiple types of data, thereby reducing the impact of abnormalities or omissions of a single data source on the overall assessment results and improving the stability and reliability of the system.

[0086] The adaptive weighted network introduces a small neural network, which includes a first fully connected layer and a second fully connected layer, and the first fully connected layer and the second fully connected layer are both provided with an activation function ReLU and a Dropout layer to dynamically adjust the importance weights of immediate risks, long-term trends and environmental factors, so that the model can respond flexibly according to actual conditions.

[0087] The adaptive weighted network is expressed as follows:

[0088] W=AWN(M)

[0089] Where W denotes an adaptive weight vector, which represents the importance of immediate risk, long-term trend, and environmental factors, and AWN(·) denotes a small neural network that learns how to dynamically adjust the weights based on the input M.

[0090] The adaptive weighted network can ensure that key information is given sufficient attention, thereby improving the accuracy and credibility of the evaluation results.

[0091] The interactive decision unit introduces a graph convolutional network to regard immediate risks and long-term trends as nodes and environmental factors as edge attributes to capture the complex relationship between them. The interactive decision unit is expressed as follows:

[0092] D=GCN(M,W)

[0093] Where D represents the decision feature after being processed by the graph convolutional network, reflecting the complex relationship between immediate risks, long-term trends and environmental factors. GCN(·) represents the graph convolutional network function, which is used to process the relationship between nodes and update the node status by aggregating neighbor information.

[0094] In addition, the graph convolutional network function is expressed as follows:

[0095]

[0096] Among them, H (l) represents the node feature matrix of the lth layer, A represents the adjacency matrix of the graph, which represents the connection relationship between nodes, and D represents the degree matrix. represents the adjacency matrix after adding the self-loop, I N represents the identity matrix, represents the updated degree matrix, W (l) represents the weight matrix of the lth layer, and σ(·) represents the activation function, such as ReLU.

[0097] The explanatory output layer is used to calculate the safety index ESI and simultaneously generate the contribution of each input feature to the final decision. The explanatory output layer is expressed as follows:

[0098] (ESI,C)=EOL(D)

[0099] Among them, ESI represents the safety index, C represents the contribution vector, that is, the contribution of each input feature to the final decision, and EOL(·) represents the explanatory output layer function.

[0100] Specifically, ESI is obtained by the following expression:

[0101] ESI=σ eol (W esi T D+b esi )

[0102] Among them, W esi represents the weight matrix from decision features to safety index, b esi represents the bias term, D represents the feature after being processed by the interactive decision unit, σ eol (·) represents an activation function, such as Sigmoid.

[0103] The contribution vector C is obtained by the following expression:

[0104]

[0105] Among them, C i Indicates the contribution of the i-th feature, indicating the influence of this feature on the safety index ESI, D i represents the i-th feature of the decision feature matrix D, Representation feature D i The partial derivative of ESI represents the characteristic D i The local sensitivity to ESI, α represents a scalar parameter in the path integral, ranging from 0 to 1, Represents the path integral, which calculates the total contribution by accumulating the gradients at different interpolation points.

[0106] The feedback loop mechanism is used to collect actual operating data after evaluation and feed it back to the model to update model parameters, thereby continuously improving the model prediction accuracy.

[0107] The feedback loop mechanism is expressed as follows:

[0108] θ new =FLM(θ old ,Maintenance Logs)

[0109] Among them, θ new represents the updated model parameters, θ old represents the old model parameters, which are used for the state before the feedback loop, and FLM(·) represents the feedback loop mechanism function.

[0110] In another exemplary embodiment, in step S300, training the escalator safety performance detection model includes the following steps:

[0111] S201: Obtain and preprocess the historical operation monitoring data of the escalator, and divide the preprocessed historical data into a training set and a validation set in a ratio of 7:3;

[0112] In this step, the preprocessing of the historical operation monitoring data is based on the above-mentioned preprocessing method, which will not be repeated here.

[0113] S202: Setting training parameters, for example, booster selects the tree-based model gbtree by default, learning_rate is set to 0.01, max_depth is set to 3, and the maximum number of iterations is set to 300. The escalator safety performance detection model is trained using the training set. During the model training process, when the number of model iterations meets the preset value, the model training is completed.

[0114] S203: Use the validation set to validate the trained model. During the validation process, if the mean absolute error (MAE), mean square error (MSE) and root mean square error (RMSE) as model performance evaluation indicators are all less than the threshold (the threshold of MAE is set to 0.05, and the threshold of MSE is set to 0.001), the model validation is passed; otherwise, adjust the training parameters or expand the training set samples to retrain the model, for example, adjust the learning_rate to 0.005, adjust the max_depth to 5, and then retrain the model until the validation is passed.

[0115] Below, the present application combines specific data to exemplarily illustrate how the method described in the present application performs safety performance testing on an escalator.

[0116] The safety performance of the escalator in a subway station is tested. First, the operation monitoring data of the escalator is collected over a period of time (for example, 3 months). The collection interval is 15 minutes. The data at each time point includes:

[0117] Mechanical performance data:

[0118] Running speed: average value is 0.5m / s, standard deviation is 0.02m / s.

[0119] Speed ​​change rate when starting and stopping: average acceleration is 0.8m / s 2 , the deceleration rate is -0.9m / s 2 .

[0120] Tilt angle: Constant at 30°.

[0121] Step gap: average value is 5mm, standard deviation is 0.5mm.

[0122] Step chain tension: average value is 2000N, standard deviation is 100N.

[0123] Dynamic response data:

[0124] Vibration conditions: average amplitude is 0.05g, maximum value does not exceed 0.1g.

[0125] Noise level: average value is 65dB(A), peak value can reach 70dB(A).

[0126] Motor current: average value is 20A, fluctuation range is ±2A.

[0127] Environmental data:

[0128] Ambient temperature: average value is 22℃, fluctuation range is ±5℃.

[0129] The operating temperature of the motor and reducer: the average value is 70℃, the fluctuation range is ±10℃.

[0130] Ambient humidity: The average relative humidity is 60%, with a fluctuation range of ±10%.

[0131] After preprocessing the above data, it was input into the trained escalator safety performance detection model, and the output safety index ESI was 0.92 (the full score was set to 1 point).

[0132] The above scores indicate that the current safety condition of the escalator is good, but maintenance is still needed.

[0133] It should be noted that in the process of calculating the safety index ESI, this application uses advanced feature extraction techniques, such as multi-scale time window sampling, local feature generation, etc., as well as feature selection and enhancement units to select the most valuable feature combinations, which helps the model focus on those factors that have the greatest impact on safety performance, thereby improving the accuracy of the prediction. In addition, ESI not only takes into account the risk level at the current moment (immediate risk assessment), but also combines the trend forecast of potential risk changes in the future (long-term trend forecast). This dual assessment method enables the model to better capture factors that may cause failures, thereby effectively improving the accuracy of safety performance detection.

[0134] In addition, the explanatory output layer also gives the contribution of each input feature to the final decision, for example:

[0135] Running speed: Contribution 15%

[0136] Speed ​​change rate when starting and stopping: Contribution 10%

[0137] Tilt angle: Contribution 5%

[0138] Step gap: Contribution 12%

[0139] Step chain tension: Contribution 10%

[0140] Vibration: Contribution 8%

[0141] Noise amount: Contribution 8%

[0142] Motor current: Contribution 12%

[0143] Ambient temperature: Contribution 5%

[0144] Operating temperature of motor and reducer: contribution 7%

[0145] Ambient humidity: Contribution 8%

[0146] The contribution output can help technicians understand which factors have a greater impact on the safety status of the escalator, so that they can take targeted preventive measures or arrange necessary maintenance work.

[0147] In another exemplary embodiment, the present application also provides an escalator safety performance detection device, such as Figure 4 As shown, the device includes: an acquisition module 100, used to acquire the operation monitoring data of the escalator; a preprocessing module 200, used to preprocess the operation monitoring data; a model building and training module 300, used to build an escalator safety performance detection model and train the model; an output module 400, used to input the preprocessed operation monitoring data into the trained escalator safety performance detection model, and output the safety index of the escalator.

[0148] Optionally, the preprocessing module 200 includes: a cleaning submodule for cleaning the operation monitoring data; a transformation submodule for transforming the operation monitoring data after data cleaning; and a reduction submodule for reducing the operation monitoring data after data transformation.

[0149] Based on the above embodiments, Figure 5 , for an explanation of the computer-readable storage medium of the exemplary embodiment of the present application, please refer to Figure 5 , the computer-readable storage medium shown is a CD 40, on which a computer program (i.e., a program product) is stored. When the computer program is executed by the processor, each step recorded in the above method implementation will be implemented, for example, obtaining the operation monitoring data of the escalator; preprocessing the operation monitoring data; constructing an escalator safety performance detection model and training the model; inputting the preprocessed operation monitoring data into the trained escalator safety performance detection model, and outputting the safety index of the escalator. The specific implementation method of each step will not be repeated here.

[0150] It should be noted that the computer-readable storage medium includes but is not limited to phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical or magnetic storage media, which are not listed here one by one.

[0151] Based on the above embodiments, the present application also provides an electronic device, as shown below: Figure 6 An electronic device for downloading a file according to an exemplary embodiment of the present application is described.

[0152] Figure 6 A block diagram of an exemplary electronic device 50 suitable for implementing the embodiments of the present application is shown, and the electronic device 50 may be a computer system or a cloud server. Figure 6 The electronic device 50 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0153] like Figure 6 As shown, the electronic device 50 includes but is not limited to: one or more processors or processing units 501, a system memory 502, and a bus 503 connecting different system components (including the system memory 502 and the processing unit 501).

[0154] The electronic device 50 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the electronic device 50, including volatile and non-volatile media, removable and non-removable media.

[0155] The system memory 502 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 5021 and / or cache memory 5022. The electronic device 50 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the ROM 5023 may be used to read and write non-removable, non-volatile magnetic media ( Figure 6 is not shown in the Figure 6 As shown in FIG. 5 , a disk drive for reading and writing a removable non-volatile disk (e.g., a "floppy disk") and an optical disk drive for reading and writing a removable non-volatile optical disk (e.g., a CD-ROM, a DVD-ROM, or other optical media) can be provided. In these cases, each drive can be connected to bus 503 via one or more data medium interfaces. System memory 502 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of each embodiment of the present application.

[0156] A program / utility 5025 having a set (at least one) of program modules 5024 may be stored, for example, in system memory 502, and such program modules 5024 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. Program modules 5024 generally perform the functions and / or methods of the embodiments described herein.

[0157] The electronic device 50 may also communicate with one or more external devices 504 (e.g., a keyboard, a pointing device, a display, etc.). Such communication may be performed via an input / output (I / O) interface 505. Furthermore, the electronic device 50 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 506. Figure 6 As shown, the network adapter 506 communicates with other modules (such as the processing unit 501, etc.) of the electronic device 50 via the bus 503. It should be understood that although Figure 6 Not shown, other hardware and / or software modules may be used in conjunction with the electronic device 50 .

[0158] The processing unit 501 executes various functional applications and data processing by running the program stored in the system memory 502, for example, obtaining the operation monitoring data of the escalator; preprocessing the operation monitoring data; constructing an escalator safety performance detection model and training the model; inputting the preprocessed operation monitoring data into the trained escalator safety performance detection model, and outputting the safety index of the escalator. The specific implementation method of each step is not repeated here. It should be noted that although several units / modules or sub-units / sub-modules of the file concurrent downloading device are mentioned in the above detailed description, this division is only exemplary and not mandatory. In fact, according to the implementation mode of the present application, the features and functions of two or more units / modules described above can be concretized in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided into multiple units / modules to be concretized.

[0159] In the description of the present application, it should be noted that the terms "first", "second" and "third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0160] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0161] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.

[0162] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0163] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0164] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that can be executed by a processor. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a cloud server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0165] The above embodiments are only for illustrating the technical concept and features of the present application, and their purpose is to enable people familiar with the technology to understand the content of the present application and implement it accordingly, and they cannot be used to limit the protection scope of the present application. Any equivalent changes or modifications made according to the spirit of the present application should be included in the protection scope of the present application.

Claims

1. A method for detecting the safety performance of an escalator, characterized in that: The method comprises: Obtain escalator operation monitoring data; Preprocessing the operation monitoring data; An escalator safety performance detection model is constructed and trained; the escalator safety performance detection model performs multi-scale time window sampling on the operation monitoring data and allocates adaptive weights to capture short-term transient events and long-term trend changes in the operation of the escalator; The pre-processed operation monitoring data is input into a trained escalator safety performance detection model, and the safety index of the escalator is output.

2. The escalator safety performance detection method according to claim 1, characterized in that: Preprocessing the operation monitoring data includes: Performing data cleaning on the operation monitoring data; Performing data transformation on the operation monitoring data after data cleaning; The operation monitoring data after data transformation is subjected to data reduction.

3. The escalator safety performance detection method according to claim 1, characterized in that: The escalator safety performance detection model includes: The input layer, feature extraction layer, feature analysis layer and output layer are connected in sequence, where The input layer is used to receive the pre-processed operation monitoring data; The feature extraction layer is used to extract features from the preprocessed operation monitoring data; The feature analysis layer is used to analyze the extracted features to evaluate the immediate risk of the escalator and predict the long-term operation trend, and calculate the safety index of the escalator based on the risk assessment result and the long-term operation trend prediction; The output layer is used to output the security index.

4. The escalator safety performance detection method according to claim 3 is characterized in that: The feature extraction layer comprises: A multi-scale time window sampling unit, an adaptive weight allocation unit, a local feature generation unit, a global context fusion unit, a feature selection and enhancement unit, a feature mapping unit and an output interface are connected in sequence.

5. The escalator safety performance detection method according to claim 3, characterized in that: The feature analysis layer includes: The immediate risk assessment module, long-term trend prediction module and safety evaluation module are connected in sequence, among which, The instant risk assessment module is used to quickly identify possible risk factors of the escalator; The long-term trend prediction module is used to predict the potential risk change trend of the escalator in the future; The safety evaluation module is used to comprehensively analyze immediate risks, long-term trends and environmental factors, and calculate and obtain a safety index.

6. The escalator safety performance detection method according to claim 1, characterized in that: The escalator safety performance detection model is trained through the following steps: Obtain and preprocess the historical data of the escalator, and divide the preprocessed historical data into a training set and a test set; Setting training parameters, and using the training set to train the escalator safety performance detection model until the training meets the maximum number of iterations; The trained model is verified using the test set. During the verification process, if the mean absolute error (MAE), mean square error (MSE), and root mean square error (RMSE), which are the model performance evaluation indicators, are all less than the threshold, the model verification is successful; otherwise, the training parameters are adjusted or the training set samples are expanded to retrain the model until the verification is successful.

7. An escalator safety performance detection device, characterized in that: The device comprises: An acquisition module, used for acquiring the operation monitoring data of the escalator; A preprocessing module, used for preprocessing the operation monitoring data; Model building and training module, used to build escalator safety performance detection model and train the model; The output module is used to input the pre-processed operation monitoring data into the trained escalator safety performance detection model and output the safety index of the escalator.

8. An escalator safety performance detection device according to claim 7, characterized in that: The preprocessing module comprises: A cleaning submodule, used for cleaning the operation monitoring data; A transformation submodule, used for performing data transformation on the operation monitoring data after data cleaning; The reduction submodule is used to reduce the operation monitoring data after the data transformation.

9. A storage medium, characterized in that: The method comprises instructions, which, when executed on a computer, enable the computer to execute the method according to any one of claims 1 to 6.

10. An electronic device, characterized in that: The electronic device comprises: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method according to any one of claims 1 to 6 is implemented.

Images