A method and system for intelligent detection of water conservancy gate status

Abstract

The present invention discloses a method and system for intelligent detection of water conservancy gate state, the method includes data collection, data preprocessing, gate opening and closing detection, water level and water pressure detection, mechanical wear detection and water conservancy gate state detection. The present invention relates to the field of gate state detection technology, specifically to a method and system for intelligent detection of water conservancy gate state, the scheme adopts the analysis of three states of gate opening and closing state, water level and water pressure state and mechanical wear state, analyzes short-term decision and early warning through gate opening and closing state and water level and water pressure state, and analyzes long-term maintenance and use through mechanical wear state, so as to perform multi-faceted and multi-angle intelligent gate state detection; adopts a bidirectional temporal convolutional network method combined with convolutional feature extraction to perform water level and water pressure state detection; adopts an integrated model combining gated cyclic unit and recursive neural network to perform mechanical wear state detection.

Description

Technical Field

[0001] The present invention relates to the technical field of gate state detection, and in particular to an intelligent detection method and system for a water conservancy gate state. Background Art

[0002] The intelligent detection method and system for water conservancy gate status is a monitoring solution based on sensor technology and data analysis, which aims to evaluate the operating status of water conservancy gates in real time. The system collects various data such as displacement, vibration, pressure, etc. of the gate, and uses machine learning and deep learning algorithms for analysis to accurately identify gate failures and abnormal conditions. Its main function is to improve the safety and reliability of water conservancy facilities, timely warn of potential risks, reduce manual inspection costs, and provide data support for the scientific management of water resources.

[0003] However, in the existing intelligent detection of water conservancy gate status, there is a problem that the existing methods mostly use physical detection and manual analysis to deal with the detection and monitoring of water conservancy gate status, and the monitoring of water conservancy gates involves too many tasks and the detection process is relatively cumbersome. The efficiency of manual processing is low, and the detection through intelligent methods must consider the complex factors affecting the gate status. The existing intelligent system is difficult to effectively provide short-term and long-term effective support for the intelligent detection of gate status. In the existing gate opening and closing status detection method, there is a traditional method for solving the gate opening and closing status. The solution is generally to use a logical model based on fixed rules to identify the signal in the sensor data, but the result of this method is too far away. The data are scattered and cannot be effectively quantified, which leads to technical problems such as unintuitive results and low reliability. In the existing water level and water pressure state detection methods, the traditional methods do not make targeted adjustments to short-term decisions and high-frequency data for water level and water pressure state detection, which leads to the technical problem that it is difficult to cooperate with the results of gate opening and closing state detection to conduct effective water conservancy environment analysis. In the existing mechanical wear state detection methods, the traditional methods mostly use image data-based analysis or complex sensor data-based integration for mechanical wear state detection. In fact, mechanical wear faced by gate state detection is a long-term problem, and overly complex sensor design or image analysis may waste computing resources. Summary of the invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a method and system for intelligent detection of water conservancy gate status. In the existing intelligent detection of water conservancy gate status, the existing methods mostly use physical detection and manual analysis to deal with the detection and monitoring of water conservancy gate status, and the monitoring of water conservancy gates involves too many tasks and the detection process is relatively cumbersome. The efficiency of manual processing is low, and the detection by intelligent methods must consider the complex factors affecting the gate status. The existing intelligent system is difficult to effectively provide short-term and long-term effective support for the intelligent detection of gate status. The technical problem, this scheme creatively adopts the analysis of the three states of gate opening and closing state, water level and water pressure state and mechanical wear state, analyzes short-term decision-making and early warning through the gate opening and closing state and water level and water pressure state, and analyzes long-term maintenance and use through mechanical wear state, and conducts multi-faceted and multi-angle intelligent gate status detection, which provides strong practical support and an explainable integrated method for the intelligentization of water conservancy gate control; in the existing gate opening and closing state detection method, there is a traditional method for solving the gate opening and closing state, which generally adopts a logical model based on a fixed Rules are used to identify signals in sensor data, but the results of this method are too discrete and cannot be effectively quantified, which leads to technical problems such as the results are not intuitive enough and the reliability is low. This solution creatively adopts a fusion decision support vector machine model combined with extreme gradient boosting to analyze position sensor signals. Through the integration and analysis of two machine learning models, it can not only realize the intelligent classification of the opening and closing states, but also realize the quantitative representation of the opening and closing states, effectively improving the result availability and process interpretability of the gate opening and closing states, and at the same time improving the computational efficiency of the overall method; In the existing water level and water pressure state detection methods, there is a traditional method for water level and water pressure state detection that does not make targeted adjustments to short-term decisions and high-frequency data, which leads to the technical problem that it is difficult to cooperate with the results of gate opening and closing state detection to conduct effective water conservancy environment analysis. This solution creatively adopts a bidirectional temporal convolutional network method combined with convolutional feature extraction to detect water level and water pressure states. By introducing a bidirectional temporal convolutional network based on the features extracted by standard convolution, it further classifies the features, thereby improving the sensitivity of water level and water pressure state detection to time, and at the same time improving the real-time and accuracy of water level and water pressure state detection;In the existing mechanical wear state detection methods, traditional methods for mechanical wear state detection mostly use image data analysis or complex sensor data integration. In fact, mechanical wear faced by gate state detection is a long-term problem. Overly complex sensor design or image analysis may waste computing resources. This solution creatively uses an integrated model combining gated recurrent units and recursive neural networks to analyze mechanical wear state from the perspective of historical data recording and vibration displacement sensing, effectively reducing the computing resources consumed by the long-term maintenance task of mechanical wear state detection, and also optimizes the overall process of intelligent detection of water conservancy gate state. ;

[0005] The technical solution adopted by the present invention is as follows: The present invention provides a method for intelligent detection of water conservancy gate status, the method comprising the following steps:

[0006] Step S1: data collection;

[0007] Step S2: data preprocessing;

[0008] Step S3: gate opening and closing detection;

[0009] Step S4: water level and water pressure detection;

[0010] Step S5: Mechanical wear detection;

[0011] Step S6: Detection of water gate status.

[0012] Further, in step S1, the data collection is used to collect the original data set required for the intelligent detection of the water conservancy gate state, specifically, to obtain the original data for the water conservancy gate state detection through wireless sensor network data collection;

[0013] The original data of water conservancy gate state detection specifically includes sensor data and environmental data;

[0014] The sensor data specifically includes water level sensor data, water pressure sensor data, vibration sensor data and displacement sensor data;

[0015] The environmental data specifically includes meteorological data, water flow condition data and gate operation history data.

[0016] Further, in step S2, the data preprocessing is used to optimize the original data, specifically to optimize the original data of the water conservancy gate state detection to obtain the optimized data of the water conservancy gate state detection, which specifically includes the following steps:

[0017] Step S21: data filtering, specifically processing the missing data, abnormal data and noise in the original data of the water conservancy gate state detection, specifically using the mean interpolation method to process the missing values ​​and abnormal values, and using the median filtering method to remove the noise, to obtain a filtered optimized data set;

[0018] Step S22: data standardization, specifically, performing Z-score standardization on the data in the filtered optimized data set to obtain a standard optimized data set;

[0019] Step S23: extracting signal features, specifically extracting features from the signal data in the filtered optimization data set to obtain a signal feature data set;

[0020] Step S24: extracting environmental features, specifically extracting features from the environmental data in the filtered optimization data set to obtain an environmental condition feature data set;

[0021] Step S25: data comprehensive processing, specifically, data optimization and feature integration are performed through the data filtering, data standardization, signal feature extraction and environmental feature extraction to obtain water conservancy gate state detection optimization data;

[0022] The water conservancy gate state detection optimization data specifically includes optimized sensor signal data and integrated feature data sets.

[0023] Further, in step S3, the gate opening and closing detection is used to monitor the opening and closing degree of the gate in real time and record data. Specifically, based on the water conservancy gate state detection optimization data, a fusion decision support vector machine model combined with extreme gradient boosting is used to perform position sensor signal analysis to obtain gate opening and closing state detection data, which specifically includes the following steps:

[0024] Step S31: constructing a support vector machine model, specifically, training the support vector machine model by constructing an RBF kernel function and setting an optimization objective function;

[0025] The optimization objective function is used to classify the opening and closing states of the gate opening and closing detection;

[0026] Step S32: constructing an extreme gradient boosting model, specifically training a standard extreme boosting model, and training the extreme gradient boosting model by constructing a regularized optimized cross entropy loss function;

[0027] Step S33: fusion decision design, specifically, integrating the outputs of the support vector machine model and the extreme gradient boosting model by weighted average to obtain a decision fusion output;

[0028] Step S34: gate opening and closing detection model training, specifically, through the construction of the support vector machine model, the construction of the extreme gradient boosting model and the fusion decision design, the gate opening and closing detection model training is performed to obtain the gate opening and closing detection model Model O ;

[0029] Step S35: gate opening and closing detection, specifically using the gate opening and closing detection model Model O , performing gate opening and closing detection according to the water conservancy gate state detection optimization data to obtain gate opening and closing state detection data;

[0030] The gate opening and closing state detection data specifically includes the gate opening and closing state classification and the gate opening quantitative prediction value.

[0031] Further, in step S4, the water level and water pressure detection is used to perform intelligent detection of water level and water pressure changes by intelligent prediction and comparison, specifically, based on the water conservancy gate state detection optimization data, a bidirectional temporal convolutional network method combined with convolutional feature extraction is used to perform water level and water pressure detection to obtain water level and water pressure detection data, specifically including the following steps:

[0032] Step S41: constructing a convolutional extraction layer, specifically constructing a convolutional feature extraction subnet including an input layer, a convolutional layer and a pooling layer, and activating it with a nonlinear activation function to extract water level and water pressure features in the water conservancy gate state detection optimization data;

[0033] Step S42: constructing a bidirectional temporal convolutional layer, specifically, after the convolutional extraction layer, constructing a bidirectional temporal convolutional layer to replace the fully connected layer of the convolutional extraction layer, and optimizing the temporal information processing by constructing the bidirectional temporal convolutional layer, and enhancing the short-term high-frequency feature processing capability of the model by introducing causal convolution and residual link, and constructing the bidirectional temporal convolutional layer;

[0034] Step S43: constructing a support vector machine classifier, specifically, according to the operation described in step S31, constructing a standard support vector machine model as a model classifier, performing model classification refinement and optimization, and obtaining a water level and water pressure classification output;

[0035] Step S44: water level and water pressure detection model training, specifically, by constructing the convolution extraction layer, constructing the bidirectional temporal convolution layer and constructing the support vector machine classifier, the water level and water pressure detection model training is performed to obtain the water level and water pressure detection model Model P ;

[0036] Step S45: water level and water pressure detection, specifically using the water level and water pressure detection model Model P, performing water level and water pressure detection according to the water conservancy gate state detection optimization data to obtain water level and water pressure detection data;

[0037] The water level and water pressure detection data specifically include water level anomaly detection information and water pressure anomaly detection information; the anomaly detection information specifically includes quantitative values ​​and classification identifiers.

[0038] Further, in step S5, the mechanical wear detection is used to perform wear detection of the mechanical system by predicting long-term mechanical wear conditions, specifically, based on the hydraulic gate state detection optimization data, an integrated model combining a gated recurrent unit and a recursive neural network is used to perform mechanical wear detection to obtain mechanical wear detection data, specifically including the following steps:

[0039] Step S51: data reconstruction, specifically, data fusion and time-series alignment of the vibration sensor data, displacement sensor data and gate operation history data in the water conservancy gate state detection optimization data, and obtaining mechanical wear detection support data through manual data annotation;

[0040] The manual data labeling includes labels such as normal machinery, slight mechanical wear, and severe mechanical wear;

[0041] Step S52: constructing a gated loop unit, specifically constructing a standard gated loop unit to capture the long-term dependency relationship in the mechanical wear detection support data, and obtaining a gated loop hidden state output;

[0042] Step S53: constructing a recursive neural network layer, specifically constructing a recursive neural network to update the hidden state output by the gated loop hidden state, so as to optimize the prediction of the long-term trend time series relationship and obtain an updated hidden state;

[0043] Step S54: constructing a classifier, specifically constructing a softmax classifier activated by an S-type activation function as a model classifier, and performing mechanical wear detection classification according to the updated hidden state to obtain a mechanical wear classification output;

[0044] Step S55: Mechanical wear detection model training, specifically, through the data reconstruction, the construction of the gated recurrent unit, the construction of the recursive neural network layer and the construction of the classifier, the mechanical wear detection model training is performed to obtain the mechanical wear detection model Model Q ;

[0045] Step S56: Mechanical wear detection, specifically using the mechanical wear detection model Model Q , performing mechanical wear detection according to the hydraulic gate state detection optimization data to obtain mechanical wear detection data;

[0046] The mechanical wear detection data specifically includes a mechanical wear state assessment level and a predicted mechanical wear fault type.

[0047] Further, in step S6, the hydraulic gate state detection is used to perform comprehensive intelligent gate state detection in combination with real-time gate state, water level and water pressure detection, and long-term mechanical wear detection. Specifically, by obtaining the gate opening value from the gate opening and closing state detection data, and obtaining the water level state and water pressure state data from the water level and water pressure detection data, a comprehensive evaluation of the hydraulic gate opening and closing state is performed to obtain hydraulic gate opening and closing control decision reference data, and by combining the mechanical wear detection data for long-term maintenance analysis to obtain long-term gate switch state decision reference data, and by combining the hydraulic gate opening and closing control decision reference data and the long-term gate switch state decision reference data for comprehensive analysis to obtain hydraulic gate state change decision reference data.

[0048] The present invention provides a water conservancy gate state intelligent detection system, comprising a data collection module, a data preprocessing module, an intelligent detection module and a control feedback module;

[0049] The data collection module is used for data collection, and obtains the original data of the water conservancy gate state detection through data collection, and sends the original data of the water conservancy gate state detection to the data preprocessing module;

[0050] The data preprocessing module is used for data preprocessing, and obtains the optimized data for water conservancy gate state detection through data preprocessing, and sends the optimized data for water conservancy gate state detection to the intelligent detection module;

[0051] The intelligent detection module is used for valve opening and closing detection, water level and water pressure detection and mechanical wear detection. Through the gate opening and closing detection, water level and water pressure detection and mechanical wear detection, the gate opening and closing state detection data, water level and water pressure detection data and mechanical wear detection data are obtained, and the gate opening and closing state detection data, water level and water pressure detection data and mechanical wear detection data are sent to the control feedback module;

[0052] The control feedback module is used for data analysis and long-term maintenance. Through control feedback, reference data for decision-making on the state change of the water conservancy gate is obtained, and the water conservancy gate is controlled and maintained based on the reference data for decision-making on the state change of the water conservancy gate.

[0053] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0054] (1) In the existing intelligent detection of water conservancy gate status, the existing methods mostly use physical detection and manual analysis to handle the detection and monitoring of water conservancy gate status. However, the monitoring of water conservancy gates involves too many tasks and the detection process is relatively cumbersome. The efficiency of manual processing is low, and the detection through intelligent methods must consider the complex factors that affect the gate status. The existing intelligent system is difficult to effectively provide short-term and long-term effective support for the intelligent detection of gate status. This solution creatively adopts the analysis of three states: gate opening and closing state, water level and water pressure state, and mechanical wear state. Through the gate opening and closing state and water level and water pressure state analysis, short-term decision-making and early warning are carried out, and through the mechanical wear state analysis, long-term maintenance and use are carried out. Multi-faceted and multi-angle intelligent gate status detection is carried out, which provides strong practical support and an explainable integrated method for the intelligentization of water conservancy gate control;

[0055] (2) In the existing gate opening and closing state detection methods, the traditional method for solving the gate opening and closing state generally adopts a logical model based on fixed rules to identify the signal in the sensor data. However, the results of this method are too discrete and cannot be effectively quantified, which leads to technical problems that the results are not intuitive and have low reliability. This solution creatively adopts a fusion decision support vector machine model combined with extreme gradient boosting to analyze the position sensor signal. Through the integration and analysis of the two machine learning models, it can realize the intelligent classification of the opening and closing state and the quantitative representation of the opening and closing state, effectively improving the result availability and process interpretability of the gate opening and closing state, and at the same time improving the computational efficiency of the overall method.

[0056] (3) In the existing water level and water pressure state detection methods, there is a technical problem that the traditional methods for water level and water pressure state detection do not make targeted adjustments to short-term decisions and high-frequency data, which leads to the difficulty in conducting effective water conservancy environment analysis in conjunction with the results of gate opening and closing state detection. This solution creatively adopts a bidirectional temporal convolutional network method combined with convolutional feature extraction to perform water level and water pressure state detection. By introducing a bidirectional temporal convolutional network to further classify the features extracted based on standard convolution, the sensitivity of water level and water pressure state detection to time is improved, while the real-time and accuracy of water level and water pressure state detection are improved.

[0057] (4) In the existing mechanical wear status detection methods, traditional methods for mechanical wear status detection mostly use image data-based analysis or complex sensor data-based integration. In fact, mechanical wear faced by gate status detection is a long-term problem. Overly complex sensor design or image analysis may waste computing resources. This solution creatively uses an integrated model combining gated recurrent units and recursive neural networks to analyze mechanical wear status from the perspective of historical data recording and vibration displacement sensing, effectively reducing the computing resources consumed by mechanical wear status detection, a long-term maintenance task, and also optimizing the overall process of intelligent detection of water conservancy gate status. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 A schematic diagram of a flow chart of a method for intelligent detection of water conservancy gate status provided by the present invention;

[0059] Figure 2 A schematic diagram of an intelligent detection system for water conservancy gate status provided by the present invention;

[0060] Figure 3 This is a schematic diagram of the process of data preprocessing in step S2;

[0061] Figure 4 This is a flow chart of gate opening and closing detection in step S3;

[0062] Figure 5 This is a schematic diagram of the process of water level and water pressure detection in step S4;

[0063] Figure 6 This is a schematic diagram of the process of mechanical wear detection in step S5;.

[0064] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0065] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0066] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0067] Example 1, see Figure 1 The present invention provides a method for intelligently detecting the state of a water conservancy gate, the method comprising the following steps:

[0068] Step S1: data collection;

[0069] Step S2: data preprocessing;

[0070] Step S3: gate opening and closing detection;

[0071] Step S4: water level and water pressure detection;

[0072] Step S5: Mechanical wear detection;

[0073] Step S6: Detection of water gate status.

[0074] By performing the above operations, in the existing intelligent detection of water conservancy gate status, there is a problem that the existing methods mostly use physical detection and manual analysis to deal with the detection and monitoring of water conservancy gate status, while the monitoring of water conservancy gates involves too many tasks and the detection process is relatively cumbersome, and the efficiency of manual processing is low. Detection through intelligent methods must consider complex factors that affect the gate status. The existing intelligent system is difficult to effectively provide short-term and long-term effective support for the intelligent detection of gate status. This solution creatively adopts the analysis of the three states of gate opening and closing state, water level and water pressure state, and mechanical wear state. Short-term decision-making and early warning are analyzed through the gate opening and closing state and water level and water pressure state, and long-term maintenance and use are analyzed through the mechanical wear state, so as to conduct multi-faceted and multi-angle intelligent gate status detection, which provides strong practical support and an explainable integrated method for the intelligentization of water conservancy gate control.

[0075] Example 2, see Figure 1 and Figure 2 In step S1, the data collection is used to collect the original data set required for the intelligent detection of the water conservancy gate state, specifically, to obtain the original data for the water conservancy gate state detection through wireless sensor network data collection;

[0076] The original data of water conservancy gate state detection specifically includes sensor data and environmental data;

[0077] The sensor data specifically includes water level sensor data, water pressure sensor data, vibration sensor data and displacement sensor data;

[0078] The environmental data specifically includes meteorological data, water flow condition data and gate operation history data.

[0079] Example 3, see Figure 1 , Figure 2 and Figure 3 This embodiment is based on the above embodiment. In step S2, the data preprocessing is used to optimize the original data, specifically to optimize the original data of the water conservancy gate state detection to obtain the optimized data of the water conservancy gate state detection, which specifically includes the following steps:

[0080] Step S21: data filtering, specifically processing the missing data, abnormal data and noise in the original data of the water conservancy gate state detection, specifically using the mean interpolation method to process the missing values ​​and abnormal values, and using the median filtering method to remove the noise, to obtain a filtered optimized data set;

[0081] Step S22: data standardization, specifically, performing Z-score standardization on the data in the filtered optimized data set to obtain a standard optimized data set;

[0082] Step S23: extracting signal features, specifically extracting features from the signal data in the filtered optimization data set to obtain a signal feature data set;

[0083] Step S24: extracting environmental features, specifically extracting features from the environmental data in the filtered optimization data set to obtain an environmental condition feature data set;

[0084] Step S25: data comprehensive processing, specifically, data optimization and feature integration are performed through the data filtering, data standardization, signal feature extraction and environmental feature extraction to obtain water conservancy gate state detection optimization data;

[0085] The water conservancy gate state detection optimization data specifically includes optimized sensor signal data and integrated feature data sets.

[0086] Example 4, see Figure 1 , Figure 2 and Figure 4 This embodiment is based on the above embodiment. In step S3, the gate opening and closing detection is used to monitor the gate opening and closing degree in real time and record data. Specifically, based on the water conservancy gate state detection optimization data, a fusion decision support vector machine model combined with extreme gradient boosting is used to perform position sensor signal analysis to obtain gate opening and closing state detection data. Specifically, the following steps are included:

[0087] Step S31: constructing a support vector machine model, specifically, training the support vector machine model by constructing an RBF kernel function and setting an optimization objective function;

[0088] The optimization objective function is used to classify the opening and closing states of the gate opening and closing detection;

[0089] The calculation formula of the RBF kernel function is:

[0090] K(x i , x j )=exp(-γ||x i -x j || 2 );

[0091] Where K(·) is the RBF kernel function, x i is the first input data, x j is the second input data, exp(·) is the natural base function, and γ is the width parameter;

[0092] The calculation formula of the optimization objective function is:

[0093]

[0094] In the formula, F tr is the optimization objective function value, min is the minimization function, w is the model parameter to be optimized, which is used to represent the model parameter of the support vector machine model, C is the regularization penalty factor, N is the total number of input data, n is the input data index, μ n It is the slack variable corresponding to the nth input data, which is used to indicate the degree of error of the classification corresponding to the nth input data;

[0095] Step S32: constructing an extreme gradient boosting model, specifically training a standard extreme boosting model, and training the extreme gradient boosting model by constructing a regularized optimized cross entropy loss function;

[0096] The calculation formula of the regularized optimization cross entropy loss function is:

[0097]

[0098] Where L(·) is the regularized optimization cross entropy loss function, f m is the model parameter of the extreme gradient boosting model, N is the total number of input data, n is the input data index, yn is the true value of the gate opening and closing state corresponding to the nth input data, is the predicted value of the gate opening and closing state corresponding to the nth input data, M is the total number of parameters of the limited gradient boosting model, m is the parameter index of the limited gradient boosting model, and Ω(·) is the regularization term function, which is used to represent the model parameter f of the limited gradient boosting model.m complexity;

[0099] Step S33: fusion decision design, specifically, integrating the outputs of the support vector machine model and the extreme gradient boosting model by weighted average to obtain a decision fusion output;

[0100] The calculation formula of the weighted average integration is:

[0101] Output fu =αO SVM +(1-α)O XGB ;

[0102] In the formula, Output fu is the decision fusion output, α is the weighted fusion weight, the specific value is 0.65, O SVM is the prediction output of the support vector machine model, O XGB is the prediction output of the extreme gradient boosting model;

[0103] Step S34: gate opening and closing detection model training, specifically, through the construction of the support vector machine model, the construction of the extreme gradient boosting model and the fusion decision design, the gate opening and closing detection model training is performed to obtain the gate opening and closing detection model Model O ;

[0104] Step S35: gate opening and closing detection, specifically using the gate opening and closing detection model Model O , performing gate opening and closing detection according to the water conservancy gate state detection optimization data to obtain gate opening and closing state detection data;

[0105] The gate opening and closing state detection data specifically includes the gate opening and closing state classification and the gate opening quantitative prediction value.

[0106] By performing the above operations, in the existing gate opening and closing state detection method, there is a traditional method for solving the gate opening and closing state. Generally, the solution is to use a logical model based on fixed rules to identify the signal in the sensor data, but the result of this method is too discrete and cannot be effectively quantified, which leads to the technical problem that the result is not intuitive enough and the reliability is low. This solution creatively uses a fusion decision support vector machine model combined with extreme gradient boosting to analyze the position sensor signal. Through the integration and analysis of the two machine learning models, it can realize the intelligent classification of the opening and closing state and the quantitative representation of the opening and closing state, effectively improving the result availability and process interpretability of the gate opening and closing state, and at the same time improving the computational efficiency of the overall method.

[0107] Example 5, see Figure 1 and Figure 2This embodiment is based on the above embodiment. In step S4, the water level and water pressure detection is used to perform intelligent detection of water level and water pressure changes by intelligent prediction and comparison. Specifically, based on the water conservancy gate state detection optimization data, a bidirectional temporal convolutional network method combined with convolutional feature extraction is used to perform water level and water pressure detection to obtain water level and water pressure detection data. Specifically, the following steps are included:

[0108] Step S41: constructing a convolutional extraction layer, specifically constructing a convolutional feature extraction subnet including an input layer, a convolutional layer and a pooling layer, and activating it with a nonlinear activation function to extract water level and water pressure features in the water conservancy gate state detection optimization data;

[0109] Step S42: constructing a bidirectional temporal convolutional layer, specifically, after the convolutional extraction layer, constructing a bidirectional temporal convolutional layer to replace the fully connected layer of the convolutional extraction layer, and optimizing the temporal information processing by constructing the bidirectional temporal convolutional layer, and enhancing the short-term high-frequency feature processing capability of the model by introducing causal convolution and residual link, and constructing the bidirectional temporal convolutional layer;

[0110] The calculation formula for constructing the bidirectional temporal convolutional layer is:

[0111]

[0112] In the formula, Z t It is the output of the bidirectional temporal convolution layer, which is used to indicate that the addition of residual links will activate the causal convolution output A. t and the original data input X t Output feature to be concatenated, Y t is the causal convolution output, t is the time index, X is the data input, which is used to represent the water level and water pressure characteristics of the convolution extraction layer output, W is the causal convolution weight, K is the total size of the causal convolution weight sequence, k is the causal convolution weight sequence index, which specifically represents the weighted summation of the past k time steps of the data input X at the current time t, W k is the kth causal convolution weight, X t-k is the data input of the past k time steps at the current time t; A t is the causal convolution activation output, ReLU (·) nonlinear activation function, and b is the activation bias term;

[0113] Step S43: constructing a support vector machine classifier, specifically, according to the operation described in step S31, constructing a standard support vector machine model as a model classifier, performing model classification refinement and optimization, and obtaining a water level and water pressure classification output;

[0114] Step S44: water level and water pressure detection model training, specifically, by constructing the convolution extraction layer, constructing the bidirectional temporal convolution layer and constructing the support vector machine classifier, the water level and water pressure detection model training is performed to obtain the water level and water pressure detection model Model P ;

[0115] Step S45: water level and water pressure detection, specifically using the water level and water pressure detection model Model P , performing water level and water pressure detection according to the water conservancy gate state detection optimization data to obtain water level and water pressure detection data;

[0116] The water level and water pressure detection data specifically include water level anomaly detection information and water pressure anomaly detection information; the anomaly detection information specifically includes quantitative values ​​and classification identifiers.

[0117] By performing the above operations, in the existing water level and water pressure status detection methods, there is a technical problem that the traditional methods for water level and water pressure status detection do not make targeted adjustments to short-term decisions and high-frequency data, which leads to difficulty in conducting effective water conservancy environment analysis in conjunction with the results of gate opening and closing status detection. This scheme creatively adopts a bidirectional temporal convolutional network method combined with convolutional feature extraction to perform water level and water pressure status detection. By introducing a bidirectional temporal convolutional network to further classify the features extracted based on standard convolution, the sensitivity of water level and water pressure status detection to time is improved, while the real-time and accuracy of water level and water pressure status detection are improved.

[0118] Example 6, see Figure 1 , Figure 2 and Figure 6 This embodiment is based on the above embodiment. In step S5, the mechanical wear detection is used to detect the wear of the mechanical system by predicting the long-term mechanical wear. Specifically, based on the hydraulic gate state detection optimization data, an integrated model combining a gated recurrent unit and a recurrent neural network is used to perform mechanical wear detection to obtain mechanical wear detection data. Specifically, the following steps are included:

[0119] Step S51: data reconstruction, specifically, data fusion and time-series alignment of the vibration sensor data, displacement sensor data and gate operation history data in the water conservancy gate state detection optimization data, and obtaining mechanical wear detection support data through manual data annotation;

[0120] The manual data labeling includes labels such as normal machinery, slight mechanical wear, and severe mechanical wear;

[0121] Step S52: constructing a gated loop unit, specifically constructing a standard gated loop unit to capture the long-term dependency relationship in the mechanical wear detection support data, and obtaining a gated loop hidden state output;

[0122] Step S53: constructing a recursive neural network layer, specifically constructing a recursive neural network to update the hidden state output by the gated loop hidden state, so as to optimize the prediction of the long-term trend time series relationship and obtain an updated hidden state;

[0123] Step S54: constructing a classifier, specifically constructing a softmax classifier activated by an S-type activation function as a model classifier, and performing mechanical wear detection classification according to the updated hidden state to obtain a mechanical wear classification output;

[0124] Step S55: Mechanical wear detection model training, specifically, through the data reconstruction, the construction of the gated recurrent unit, the construction of the recursive neural network layer and the construction of the classifier, the mechanical wear detection model training is performed to obtain the mechanical wear detection model Model Q ;

[0125] Step S56: Mechanical wear detection, specifically using the mechanical wear detection model Model Q , performing mechanical wear detection according to the hydraulic gate state detection optimization data to obtain mechanical wear detection data;

[0126] The mechanical wear detection data specifically includes a mechanical wear state assessment level and a predicted mechanical wear fault type.

[0127] By performing the above operations, in the existing mechanical wear status detection methods, the traditional methods for mechanical wear status detection mostly use image data-based analysis or complex sensor data-based integration. In fact, the mechanical wear faced by gate status detection is a long-term problem. Overly complex sensor design or image analysis may waste computing resources. This solution creatively uses an integrated model combining gated recurrent units and recursive neural networks to analyze the mechanical wear status from the perspective of historical data recording and vibration displacement sensing, effectively reducing the computing resources consumed by the long-term maintenance task of mechanical wear status detection, and also optimizes the overall process of intelligent detection of water conservancy gate status.

[0128] Embodiment 7, see Figure 1 and Figure 2This embodiment is based on the above embodiment. In step S6, the water conservancy gate state detection is used to perform comprehensive intelligent gate state detection in combination with real-time gate state, water level and water pressure detection and long-term mechanical wear detection. Specifically, by obtaining the gate opening value from the gate opening and closing state detection data, and obtaining the water level state and water pressure state data from the water level and water pressure detection data, a comprehensive evaluation of the water conservancy gate opening and closing state is performed to obtain water conservancy gate opening and closing control decision reference data, and by combining the mechanical wear detection data for long-term maintenance analysis to obtain long-term gate switch state decision reference data, and by combining the water conservancy gate opening and closing control decision reference data and the long-term gate switch state decision reference data for comprehensive analysis to obtain water conservancy gate state change decision reference data.

[0129] Embodiment 8, see Figure 1 and Figure 2 , this embodiment is based on the above embodiment, and the present invention provides a water conservancy gate state intelligent detection system, including a data collection module, a data preprocessing module, an intelligent detection module and a control feedback module;

[0130] The data collection module is used for data collection, and obtains the original data of the water conservancy gate state detection through data collection, and sends the original data of the water conservancy gate state detection to the data preprocessing module;

[0131] The data preprocessing module is used for data preprocessing, and obtains the optimized data for water conservancy gate state detection through data preprocessing, and sends the optimized data for water conservancy gate state detection to the intelligent detection module;

[0132] The intelligent detection module is used for valve opening and closing detection, water level and water pressure detection and mechanical wear detection. Through the gate opening and closing detection, water level and water pressure detection and mechanical wear detection, the gate opening and closing state detection data, water level and water pressure detection data and mechanical wear detection data are obtained, and the gate opening and closing state detection data, water level and water pressure detection data and mechanical wear detection data are sent to the control feedback module;

[0133] The control feedback module is used for data analysis and long-term maintenance. Through control feedback, reference data for decision-making on the state change of the water conservancy gate is obtained, and the water conservancy gate is controlled and maintained based on the reference data for decision-making on the state change of the water conservancy gate.

[0134] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0135] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0136] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A method for intelligent detection of water conservancy gate status, characterized in that: The method comprises the following steps: Step S1: data collection, obtaining the original data of water conservancy gate status detection; Step S2: data preprocessing to obtain optimized data for water conservancy gate status detection; Step S3: gate opening and closing detection, using a fusion decision support vector machine model combined with extreme gradient boosting to analyze position sensor signals and obtain gate opening and closing state detection data; Step S4: water level and water pressure detection, using a bidirectional temporal convolutional network method combined with convolutional feature extraction to perform water level and water pressure detection to obtain water level and water pressure detection data; Step S5: Mechanical wear detection, using an integrated model combining a gated recurrent unit and a recurrent neural network to perform mechanical wear detection to obtain mechanical wear detection data, including the following steps: Step S51: data reconstruction; Step S52: constructing a gated recurrent unit; Step S53: constructing a recursive neural network layer; Step S54: constructing a classifier; Step S55: mechanical wear detection model training; Step S56: mechanical wear detection; Step S6: hydraulic gate state detection, which is used to perform comprehensive intelligent gate state detection in combination with real-time gate state, water level and water pressure detection, and long-term mechanical wear detection. By obtaining the gate opening value from the gate opening and closing state detection data, and obtaining the water level state and water pressure state data from the water level and water pressure detection data, a comprehensive evaluation of the hydraulic gate opening and closing state is performed to obtain hydraulic gate opening and closing control decision reference data, and by combining the mechanical wear detection data for long-term maintenance analysis, long-term gate switch state decision reference data is obtained, and by combining the hydraulic gate opening and closing control decision reference data and the long-term gate switch state decision reference data for comprehensive analysis, the hydraulic gate state change decision reference data is obtained.

2. The method for intelligent detection of water conservancy gate status according to claim 1 is characterized in that: In step S1, the water conservancy gate state detection raw data includes sensor data and environmental data; The sensor data includes water level sensor data, water pressure sensor data, vibration sensor data and displacement sensor data; The environmental data includes meteorological data, water flow condition data and gate operation history data.

3. A method for intelligent detection of water conservancy gate status according to claim 2, characterized in that: In step S2, the data preprocessing is used to optimize the original data, optimize the original data of the water conservancy gate state detection, and obtain the optimized data of the water conservancy gate state detection, including the following steps: Step S21: data filtering, processing missing data, abnormal data and noise in the original data of the water conservancy gate state detection, using mean interpolation method to process missing values ​​and abnormal values, and using median filtering method to remove noise, to obtain a filtered optimized data set; Step S22: data standardization, performing Z-score standardization on the data in the filtered optimized data set to obtain a standard optimized data set; Step S23: extracting signal features, extracting features from the signal data in the filtered optimization data set to obtain a signal feature data set; Step S24: extracting environmental features, extracting features from the environmental data in the filtered optimization data set to obtain an environmental condition feature data set; Step S25: data comprehensive processing, through the data filtering, the data standardization, the signal feature extraction and the environmental feature extraction, data optimization and feature integration are performed to obtain water conservancy gate state detection optimization data; The water conservancy gate state detection optimization data includes optimized sensor signal data and integrated feature data sets.

4. The method for intelligent detection of water conservancy gate status according to claim 3 is characterized in that: In step S3, the gate opening and closing detection is used to monitor the gate opening and closing degree in real time and record data. According to the water conservancy gate state detection optimization data, a fusion decision support vector machine model combined with extreme gradient boosting is used to analyze the position sensor signal to obtain the gate opening and closing state detection data, including the following steps: Step S31: constructing a support vector machine model, training the support vector machine model by constructing an RBF kernel function and setting an optimization objective function; The optimization objective function is used to classify the opening and closing states of the gate opening and closing detection; Step S32: construct an extreme gradient boosting model, train a standard extreme boosting model, and train the extreme gradient boosting model by constructing a regularized optimized cross entropy loss function; Step S33: Fusion decision design, using weighted average to integrate the outputs of the support vector machine model and the extreme gradient boosting model to obtain a decision fusion output; Step S34: gate opening and closing detection model training, through the construction of the support vector machine model, the construction of the extreme gradient boosting model and the fusion decision design, the gate opening and closing detection model training is performed to obtain the gate opening and closing detection model Model O ; Step S35: Gate opening and closing detection, using the gate opening and closing detection model Model O , performing gate opening and closing detection according to the water conservancy gate state detection optimization data to obtain gate opening and closing state detection data; The gate opening and closing state detection data includes gate opening and closing state classification and gate opening quantitative prediction value.

5. A method for intelligent detection of water conservancy gate status according to claim 4, characterized in that: In step S4, the water level and water pressure detection is used to perform intelligent detection of water level and water pressure changes by intelligent prediction and comparison. According to the water conservancy gate state detection optimization data, a bidirectional temporal convolutional network method combined with convolutional feature extraction is used to perform water level and water pressure detection to obtain water level and water pressure detection data, including the following steps: Step S41: constructing a convolution extraction layer, constructing a convolution feature extraction subnet including an input layer, a convolution layer and a pooling layer, and activating it with a nonlinear activation function to extract water level and water pressure features in the water conservancy gate state detection optimization data; Step S42: constructing a bidirectional temporal convolution layer. After the convolution extraction layer, a bidirectional temporal convolution layer is constructed to replace the fully connected layer of the convolution extraction layer. By constructing the bidirectional temporal convolution layer, the temporal information processing is optimized. By introducing causal convolution and residual link, the short-term high-frequency feature processing capability of the model is enhanced. The bidirectional temporal convolution layer is constructed. Step S43: constructing a support vector machine classifier, according to the operation of step S31, constructing a standard support vector machine model as a model classifier, performing model classification refinement and optimization, and obtaining a water level and water pressure classification output; Step S44: water level and water pressure detection model training, through the construction of the convolution extraction layer, the construction of the bidirectional temporal convolution layer and the construction of the support vector machine classifier, the water level and water pressure detection model training is performed to obtain the water level and water pressure detection model Model P ; Step S45: water level and water pressure detection, using the water level and water pressure detection model Model P , performing water level and water pressure detection according to the water conservancy gate state detection optimization data to obtain water level and water pressure detection data; The water level and water pressure detection data include water level anomaly detection information and water pressure anomaly detection information; the anomaly detection information includes a quantitative value and a classification mark.

6. A method for intelligent detection of water conservancy gate status according to claim 5, characterized in that: In step S5, the mechanical wear detection is used to detect the wear of the mechanical system by predicting the long-term mechanical wear. According to the hydraulic gate state detection optimization data, an integrated model combining a gated recurrent unit and a recursive neural network is used to perform mechanical wear detection to obtain mechanical wear detection data, including the following steps: Step S51: data reconstruction, fusing and time-series aligning the vibration sensor data, displacement sensor data and gate operation history data in the water conservancy gate state detection optimization data, and obtaining mechanical wear detection support data through manual data annotation; The manual data labeling includes the following labels: normal machinery, slight mechanical wear, and severe mechanical wear; Step S52: constructing a gated loop unit, constructing a standard gated loop unit to capture the long-term dependency relationship in the mechanical wear detection support data, and obtaining a gated loop hidden state output; Step S53: constructing a recursive neural network layer, constructing a recursive neural network to update the hidden state output by the gated loop hidden state, so as to optimize the prediction of the long-term trend time series relationship and obtain an updated hidden state; Step S54: constructing a classifier, constructing a softmax classifier activated by an S-type activation function as a model classifier, and performing mechanical wear detection classification according to the updated hidden state to obtain a mechanical wear classification output; Step S55: Mechanical wear detection model training, through the data reconstruction, the construction of the gated recurrent unit, the construction of the recursive neural network layer and the construction of the classifier, the mechanical wear detection model training is performed to obtain the mechanical wear detection model Model Q ; Step S56: Mechanical wear detection, using the mechanical wear detection model Model Q , performing mechanical wear detection according to the hydraulic gate state detection optimization data to obtain mechanical wear detection data; The mechanical wear detection data includes a mechanical wear state assessment grade and a predicted mechanical wear fault type.

7. A water conservancy gate state intelligent detection system, used to implement a water conservancy gate state intelligent detection method as described in any one of claims 1 to 6, characterized in that: It includes data collection module, data preprocessing module, intelligent detection module and control feedback module; The data collection module is used for data collection, and obtains the original data of the water conservancy gate state detection through data collection, and sends the original data of the water conservancy gate state detection to the data preprocessing module; The data preprocessing module is used for data preprocessing, and obtains the optimized data for water conservancy gate state detection through data preprocessing, and sends the optimized data for water conservancy gate state detection to the intelligent detection module; The intelligent detection module is used for valve opening and closing detection, water level and water pressure detection and mechanical wear detection. Through gate opening and closing detection, a fusion decision support vector machine model combined with extreme gradient boosting is used to perform position sensing signal analysis to obtain gate opening and closing state detection data. Through water level and water pressure detection, a bidirectional temporal convolutional network method combined with convolution feature extraction is used to perform water level and water pressure detection to obtain water level and water pressure detection data. Through mechanical wear detection, an integrated model combined with a gated recurrent unit and a recursive neural network is used to perform mechanical wear detection to obtain mechanical wear detection data. The gate opening and closing state detection data, water level and water pressure detection data and mechanical wear detection data are sent to the control feedback module; The control feedback module is used for data analysis and long-term maintenance. Through control feedback, reference data for decision-making on the state change of the water conservancy gate is obtained, and the water conservancy gate is controlled and maintained based on the reference data for decision-making on the state change of the water conservancy gate.

Images