Hydraulic hoist dynamic fault diagnosis method based on random forest and fault tree

Abstract

This invention discloses a dynamic fault diagnosis method for hydraulic gate hoists based on random forest and fault tree, comprising the following steps: formulating a fault simulation scheme and collecting fault sample data; preprocessing the fault data; constructing and optimizing a random forest model; analyzing fault modes; and determining specific fault causes. This invention utilizes selected features and a fault diagnosis dataset to train the random forest diagnostic model. During training, cross-validation is used to evaluate the model's performance, ensuring good generalization ability. Finally, a random forest diagnostic model capable of accurately predicting fault modes in hydraulic gate hoists is obtained. This invention clarifies the logical relationship between fault modes and fault causes by constructing a fault tree model, forming a closed-loop diagnostic process of "fault mode identification → fault cause tracing," balancing data-driven and knowledge-driven approaches, and overcoming the limitations of a single model. This invention improves the efficiency and accuracy of fault diagnosis.

Description

Technical Field

[0001] This invention belongs to the field of hydraulic equipment fault diagnosis technology, specifically relating to a dynamic fault diagnosis method for hydraulic gate hoists based on random forest and fault tree. Background Technology

[0002] Hydraulic gate hoists are important equipment used in water conservancy and hydropower projects to open and close hydraulic gates, and their normal operation is crucial to the safety and integrity of the project.

[0003] However, due to the complex structure of hydraulic gate hoists and their long-term standby state, fault diagnosis is quite difficult. Traditional fault diagnosis methods mostly rely on manual periodic maintenance and inspection of the hydraulic system's operating status, which cannot achieve accurate, online fault diagnosis based on key state parameters such as hydraulic oil acidity, viscosity, and water content. Although existing technologies can judge the state by monitoring single or a few characteristic parameters of hydraulic oil, it is still difficult to comprehensively and dynamically identify complex fault modes caused by the coupling of multiple factors, and the diagnostic process often lacks the ability to systematically trace the root cause of the fault, resulting in inaccurate fault location and delayed maintenance response. Summary of the Invention

[0004] To address the aforementioned problems in existing technologies, this invention designs a dynamic fault diagnosis method for hydraulic gate hoists based on random forests and fault trees. This method overcomes the limitations of scarce and incomplete historical fault data in actual engineering projects, providing a sufficient and reliable training foundation for subsequent intelligent diagnostic models and significantly improving the accuracy of diagnosis.

[0005] To achieve the above objectives, the basic idea of ​​this invention is as follows: First, fault data is obtained through digital simulation modeling of the hydraulic gate hoist, and a training dataset is constructed; then, the fault data is trained using the random forest algorithm to construct a fault diagnosis model; next, fault tree analysis is combined to conduct in-depth analysis of the fault modes and determine the causes of the faults; finally, dynamic fault diagnosis of the hydraulic gate hoist is realized, improving the accuracy and efficiency of fault diagnosis.

[0006] The technical solution of the present invention is as follows: A dynamic fault diagnosis method for hydraulic gate hoists based on random forest and fault tree, comprising the following steps:

[0007] A. Develop a fault simulation plan and collect fault sample data; A digital simulation model of the hydraulic gate hoist was built using professional simulation software. Different sensors were set up in key components, a fault simulation scheme was developed, and fault sample data was collected.

[0008] B. Preprocess the fault data.

[0009] C. Construct and optimize a random forest model; We organized and selected feature and fault diagnosis datasets, used multiple decision trees for model training, optimized the hyperparameters of the random forest through grid search, and used cross-validation to assess model performance.

[0010] D. Analyze failure modes; Based on the structure and working principle of the hydraulic gate hoist, a fault tree model is constructed, and the fault tree analysis method is used to analyze the fault modes.

[0011] E. Determine the specific cause of the malfunction. The sensor collects data in real time and inputs it into the random forest diagnostic model. Based on the fault mode output by the random forest diagnostic model and the fault tree analysis results, the specific cause of the fault is determined.

[0012] Furthermore, the method for formulating a fault simulation scheme and collecting fault sample data described in step A includes the following steps: A1. Building a simulation model A digital simulation model of a hydraulic gate hoist is built using professional simulation software. The digital simulation model accurately simulates the hydraulic gate hoist and sets physical parameters that conform to reality. The hydraulic gate hoist includes a hydraulic pump, hydraulic cylinder, control valve, oil pipes, and seals. The physical parameters include the viscosity and density of the hydraulic oil, the length, diameter, and roughness of the pipes, as well as the elastic modulus and damping coefficient of each component. A2. Based on the working principle and operation process of the hydraulic gate hoist, set the corresponding control logic and boundary conditions in the digital simulation model. The control logic includes the start and stop control of the hydraulic pump, the extension and retraction control of the hydraulic cylinder, and the opening and closing logic of the valve. The boundary conditions include the initial pressure, load conditions, simulation time, and external constraints.

[0013] A3. Filter fault modes; A31. Identify the failure modes of hydraulic gate hoists through literature review, expert consultation and analysis of actual cases; The failure modes include leakage, pressure loss, system blockage, hydraulic pump cavitation, and control valve failure.

[0014] A32. Provide a detailed description of each failure mode, including the failure phenomenon, failure cause, and impact on performance, to provide a clear objective for subsequent fault injection and diagnosis.

[0015] A33. Set up a fault injection scheme: First, design a corresponding fault injection scheme for each of the selected fault modes.

[0016] The fault injection scheme includes the conditions for the fault to occur, the range of changes in fault parameters, and the duration of the fault.

[0017] A34. Implement fault injection functionality in the digital simulation model. Modify the parameter settings of the digital simulation model or introduce additional control logic to trigger faults according to a predetermined scheme during the simulation. Then, perform multiple simulations for each fault mode, adjusting the fault parameter values ​​in each simulation to obtain monitoring data under different fault severity levels. This data will serve as the basis for subsequent training and testing of the random forest diagnostic model. Finally, record the monitoring data, the time point of fault occurrence, and the duration of each simulation to provide complete information for subsequent data analysis.

[0018] Furthermore, the method for preprocessing fault data described in step B includes the following steps: B1. Preprocess the collected fault dataset, including data cleaning and normalization. The normalization method is to scale the data to the range [0,1] using the min-max scaling method, as shown in the following formula:

[0019] in For the reduced data, This is the original data. The minimum value of the original data. This represents the maximum value of the original data.

[0020] B12. Feature selection is performed using the Lasso regression algorithm.

[0021] The Lasso regression algorithm is used to find a set of feature coefficients that minimizes the error between the predicted values ​​of the random forest diagnostic model and the actual fault labels. While minimizing the error, a penalty is applied to the sum of the absolute values ​​of all feature coefficients. The Lasso regression algorithm tends to compress the coefficients of features that contribute weakly or unstablely to the prediction target to exactly zero. A coefficient of zero means that the feature is excluded by the Lasso regression algorithm. The cost function of the Lasso regression algorithm is as follows:

[0022] in For the first i The actual value of each observed sample. For the first j The regression coefficients corresponding to each feature For the first i The first observation sample j 1 eigenvalue, This is the regularization strength parameter.

[0023] Furthermore, the method for constructing and optimizing the random forest diagnostic model described in step C includes the following steps: C1. Data preparation: Integrate the preprocessed hydraulic gate hoist fault dataset, clearly divide it into training and test sets, and ensure that each sample has an accurate and consistent fault mode label.

[0024] C2. Building Decision Trees: Decision trees are the basic unit of random forests, and each decision tree is trained independently. When building a decision tree, the next split node will randomly select m features from M candidate features. The feature with the highest information gain will be selected to split the decision tree. Maximizing the information gain of the split node... It is expressed as follows:

[0025] in, Indicates the impurity of the parent node. The impurity of the child nodes. It is a child node k The number of samples, N It represents the total number of samples at the current node.

[0026] Impurity is measured by the Gini index. The metric is represented as follows:

[0027] in, It is the first i The proportion of samples in each category, where C is the set of all categories.

[0028] In the construction of the random forest diagnostic model, each decision tree is independently trained based on randomly sampled data and feature subsets, forming a set of differentially distributed base classifiers. The final predicted category is then determined through a voting method. The formula is as follows:

[0029] in, It represents the total number of decision trees; It is the first j Decision trees for samples x The prediction results; It is an index function; it is 1 if the condition is met, and 0 otherwise. c This represents a candidate category label.

[0030] In random forests, feature importance is evaluated based on the reduction in node impurity. Feature importance is calculated by the cumulative reduction in impurity across all nodes split using that feature, as shown in the following formula:

[0031] in, The impurity accumulation for all nodes, It is a feature f In decision tree t The amount of impurity reduction during the split.

[0032] C3. Constructing the Random Forest Diagnostic Model: Initialize the basic parameters of the random forest diagnostic model, setting the number of decision trees and the maximum number of features per decision tree. When constructing each decision tree, firstly, random sampling is performed, drawing samples with replacement from the training set to form a training subset; then, when splitting a node, features are randomly selected as candidates; next, the best feature is selected for splitting based on the principle of maximizing information gain, and the entire decision tree is recursively constructed until the stopping condition is met; finally, all decision trees are integrated to form the random forest diagnostic model, and a voting mechanism is used to predict new samples.

[0033] C4. Optimize the random forest diagnostic model by using grid search or random search to find the optimal hyperparameter combination within a predefined parameter space. This hyperparameter combination includes the number of decision trees and the maximum number of features. Simultaneously, cross-validation is used to evaluate the performance of the random forest diagnostic model under different hyperparameter combinations, ensuring that the random forest diagnostic model has good generalization ability.

[0034] C5. Evaluate the optimized random forest diagnostic model, using accuracy, precision, recall, and F1 score to measure its performance on the test set. Visually demonstrate the model's prediction performance under different failure modes using a confusion matrix, analyze the causes of misdiagnosis and missed diagnosis, and provide a basis for improving the random forest diagnostic model.

[0035] Furthermore, the method for analyzing failure modes described in step D includes the following steps: D1. Define the basis for fault tree construction; The foundation for constructing the fault tree includes the structural composition, working principle, and interaction relationships between the components of the hydraulic gate hoist. It also clarifies the fault modes that need to be analyzed, including the fault location of the hydraulic gate hoist, the fault phenomenon occurring in each location, and the corresponding fault cause for each fault phenomenon.

[0036] D2. Construct a fault tree; First, define the top event, which is the overall failure of the hydraulic hoist. Then, identify the direct cause of the top event as an intermediate event. Next, further analyze the causes of the intermediate events to find the smallest fault cause that cannot be further decomposed as the basic event. Finally, use logic gates to connect the top event, intermediate events, and basic events to form a complete fault tree structure.

[0037] D3. Analyze the fault tree; First, the minimum cut set algorithm is used to solve for the minimum cut sets of the fault tree. These minimum cut sets represent the smallest combination of basic events that lead to the occurrence of the top event. Then, a qualitative analysis is performed based on the minimum cut sets to determine which basic events or combinations of basic events have the greatest impact on the occurrence of the top event. With sufficient data, a quantitative analysis is further performed to calculate the probability of occurrence of each basic event and intermediate event, and to assess the probability and risk level of the top event.

[0038] D4. Apply the results of fault tree analysis; Based on the analysis results, a detailed troubleshooting process is developed to guide technicians in quickly locating the cause of the fault.

[0039] Furthermore, the method for formulating a fault simulation scheme and collecting fault sample data in step E includes the following steps: E. Determine the specific cause of the malfunction. E1. Real-time data acquisition: By installing sensors at key parts of the hydraulic hoist, monitoring data reflecting the operating status of the equipment is collected in real time. The monitoring data includes the outlet pressure of the hydraulic pump, the extension and retraction speed of the hydraulic cylinder, the opening and closing status of the control valve, the valve inlet and outlet pressure, flow rate, and the time and duration of the fault occurrence.

[0040] E2. Preprocess the collected real-time data: Preprocess the real-time data according to the method in step B.

[0041] E3. Input the preprocessed real-time data into the pre-trained random forest diagnostic model. Based on the characteristics of the input data, the random forest diagnostic model quickly predicts the current fault modes of the hydraulic gate hoist.

[0042] E4. Based on the fault tree analysis results constructed in step D, conduct an in-depth analysis of the fault modes predicted by the random forest diagnostic model.

[0043] By finding the minimum cut set in the fault tree, the cause or combination of causes leading to the fault mode is determined, and a diagnostic result is formed.

[0044] E5. The diagnostic results are displayed on the operating interface in real time, and a detailed fault report is generated. Based on this information, operators can quickly locate the fault and take appropriate maintenance measures.

[0045] Compared with the prior art, the present invention has achieved the following beneficial effects: 1. This invention utilizes selected features and fault diagnosis datasets to train a random forest diagnostic model. The hyperparameters of the random forest, including the number of decision trees and the maximum number of features, are optimized through grid search to find the optimal model configuration. During training, cross-validation is used to evaluate the model's performance, ensuring good generalization ability. Ultimately, a random forest diagnostic model capable of accurately predicting fault modes in hydraulic gate hoists is obtained.

[0046] 2. This invention clarifies the logical relationship between fault modes and causes by constructing a fault tree model. Based on the structure and working principle of the hydraulic gate hoist, a detailed fault tree model is constructed. The fault tree model includes top events, intermediate events, and basic events, which are connected by logic gates (such as AND gates and OR gates) to form a complete fault logic diagram. It innovatively integrates the data-driven diagnostic results of random forests with the knowledge-driven analysis of fault trees: the fault mode probabilities output by the random forest are used as input for quantitative analysis of the fault tree, driving the real-time calculation of failure probabilities using minimal cut sets. This provides an important theoretical basis for subsequent fault investigation and maintenance. This innovation dynamically combines the random forest classification model with fault tree analysis, forming a closed-loop diagnostic process of "fault mode identification → fault cause tracing," balancing data-driven and knowledge-driven approaches, and overcoming the limitations of a single model.

[0047] 3. This invention first collects real-time monitoring data of the hydraulic gate hoist using sensors and inputs this data into a trained random forest diagnostic model. The model predicts the current fault mode based on the input data and outputs the prediction results. Next, combining the fault tree analysis results, it conducts in-depth analysis of the predicted fault mode to determine the specific cause of the fault. This step finds the root cause of the fault by searching for the minimum cut set in the fault tree. Finally, the diagnostic results (including the fault mode and cause) are displayed on the operation interface to guide operators in troubleshooting and repair. Through this real-time fault diagnosis system, dynamic fault diagnosis of the hydraulic gate hoist is achieved, improving the efficiency and accuracy of fault diagnosis. Attached Figure Description

[0048] Figure 1 This is a flowchart of the present invention; Figure 2 This is a technical roadmap of the present invention; Figure 3 This is a schematic diagram of a digital simulation model of a hydraulic gate opener according to the present invention; Figure 4 This is a schematic diagram of piston displacement of a hydraulic gate opener according to the present invention; Figure 5 A schematic diagram of the force applied by an actuator according to the present invention; Figure 6This is a schematic diagram of the inlet pressure of a hydraulic cylinder according to the present invention; Figure 7 This is a schematic diagram of the outlet pressure of a hydraulic cylinder according to the present invention; Figure 8 This is a schematic diagram of the inlet and outlet pressures of an electromagnetic directional valve according to the present invention; Figure 9 This is a schematic diagram of the inlet and return flow of an electromagnetic reversing valve according to the present invention; Figure 10 This is a schematic diagram of the inlet pressure of an overflow valve according to the present invention; Figure 11 This is a schematic diagram of the outlet flow rate of a pressure reducing valve according to the present invention; Figure 12 A schematic diagram of regression coefficients for a key feature of this invention; Figure 13 This is a schematic diagram of a fault diagnosis decision tree X according to the present invention; Figure 14 This is a schematic diagram of the confusion matrix of the fault mode judgment result of a hydraulic gate hoist according to the present invention; Figure 15 This is a schematic diagram of the overall fault tree of a hydraulic gate hoist according to the present invention; Figure 16 This is a schematic diagram of a leakage fault tree for a hydraulic gate hoist system according to the present invention; Figure 17 This is a schematic diagram of a blockage fault tree for a hydraulic gate hoist system according to the present invention; Figure 18 This is a schematic diagram of a cavitation fault tree for a hydraulic pump in a hydraulic gate hoist according to the present invention. Figure 19 This is a schematic diagram of a fault tree for a sudden pressure drop in a hydraulic gate hoist system according to the present invention; Figure 20 This is a schematic diagram of a pressureless fault tree for a hydraulic gate hoist system according to the present invention. Figure 21 This is a schematic diagram of a fault tree for insufficient pressure in a hydraulic gate hoist system according to the present invention. Detailed Implementation

[0049] The invention will now be further described with reference to the accompanying drawings.

[0050] like Figure 1-2 As shown, a dynamic fault diagnosis method for hydraulic gate hoists based on random forest and fault tree includes the following steps: Step 101: Use AMESIM software to build a digital simulation model of the hydraulic gate hoist, and set different sensors in key components. Develop a fault simulation scheme and collect fault sample data.

[0051] Building the simulation model: First, a detailed digital simulation model of the hydraulic gate hoist is built using professional simulation software (Amesim), such as... Figure 3 As shown. This model needs to accurately simulate all components of a hydraulic gate hoist, including but not limited to hydraulic pumps, hydraulic cylinders, control valves, oil pipes, and seals, and set realistic physical parameters, such as the viscosity and density of the hydraulic oil, the length, diameter, and roughness of the pipes, as well as the elastic modulus and damping coefficient of each component. By setting these parameters in detail, it is ensured that the simulation model can accurately reflect the dynamic behavior of the actual hydraulic gate hoist. Secondly, according to the working principle and operation process of the hydraulic gate hoist, corresponding control logic and boundary conditions are set in the simulation model. The control logic includes the start and stop control of the hydraulic pump, the extension and retraction control of the hydraulic cylinder, and the valve opening and closing logic. The boundary conditions include: initial system pressure, load conditions, simulation time, and external constraints.

[0052] Screening failure modes: First, through literature review, expert consultation and analysis of actual cases, we identify the possible failure modes of hydraulic gate hoists.

[0053] Common failure modes include, but are not limited to, system leakage, pressure loss, system blockage, hydraulic pump cavitation, and control valve failure. Secondly, each failure mode is described in detail, including the symptoms, possible causes, and impact on system performance, providing a clear objective for subsequent fault injection and diagnosis.

[0054] Setting up a fault injection scheme: First, design a corresponding fault injection scheme for each selected fault mode. The fault injection scheme needs to clearly define the conditions under which the fault occurs, the range of changes in fault parameters, and the duration of the fault.

[0055] Fault injection functionality is implemented in the simulation model. By modifying the simulation model's parameter settings or introducing additional control logic, faults are triggered according to a predetermined scheme during the simulation. Next, multiple simulations are performed for each fault mode, adjusting the fault parameter values ​​during each simulation to obtain monitoring data under different fault severity levels. This data will serve as the basis for subsequent fault diagnosis model training and testing. Finally, monitoring data from each simulation is recorded, including the extension and retraction speed of the hydraulic cylinder (e.g.,...). Figure 4 As shown), the force applied by the actuator (such as...) Figure 5 As shown), hydraulic cylinder inlet pressure (such as...) Figure 6 As shown), the outlet pressure of the hydraulic pump (e.g.) Figure 7 As shown), the inlet and outlet pressures of the solenoid directional valve (e.g.) Figure 8 As shown), the inlet and return flow of the solenoid directional valve (e.g.) Figure 9 As shown), the overflow valve inlet pressure (such as...) Figure 10 As shown), the outlet flow rate of the pressure reducing valve (e.g.) Figure 11(as shown in the image), as well as the time and duration of the fault, provide complete information for subsequent data analysis.

[0056] Step 102: Clean and normalize the fault data to eliminate the influence of different dimensions and units. Use the Lasso algorithm for feature selection.

[0057] First, a comprehensive preprocessing process was performed on the collected fault dataset, including data cleaning and normalization. The Min-Max scaling method was used to scale the data to the range [0,1] to eliminate the influence of different dimensions and units on the data analysis results, ensuring data comparability and consistency. The formula is as follows:

[0058] in For the reduced data, This is the original data. The minimum value of the original data. This represents the maximum value of the original data.

[0059] Secondly, Lasso regression is used for feature selection. Lasso regression introduces an L1 regularization term to promote the sparsity of feature coefficients, thereby eliminating features with regression coefficients approaching 0 and selecting the features most relevant to the fault mode. This step not only reduces the dimensionality of the data and improves the training efficiency of the random forest diagnostic model, but also enhances prediction accuracy, enabling subsequent models to more accurately capture fault features. The regression coefficients of the main features are as follows: Figure 12 As shown. The cost function of the Lasso regression algorithm is as follows:

[0060] in For the first i The actual value of each observed sample. For the first j The regression coefficients corresponding to each feature For the first i The first observation sample j 1 eigenvalue, This is the regularization strength parameter.

[0061] Step 103: Organize the selected features and fault diagnosis dataset, use multiple decision trees to train the model, optimize the hyperparameters of the random forest through grid search, and use cross-validation to assess the model's performance.

[0062] First, data preparation is required. This step involves integrating the preprocessed hydraulic gate hoist fault dataset (including data cleaning, normalization, and feature selection using the Lasso regression algorithm), clearly defining the training and test sets, and ensuring that each sample has an accurate and consistent fault mode label, such as system leakage or pressure loss, as the basis for subsequent model training and evaluation.

[0063] Build each decision tree model: such as Figure 13 As shown, a decision tree is the basic unit of a random forest, and each tree is trained independently. When constructing a decision tree, the next split node will randomly select m features from M candidate features. The feature with the highest information gain will be selected to split the decision tree. Maximizing the information gain of the split node...

[0064]

[0065] in, Indicates the impurity of the parent node. The impurity of the child nodes. It is a child node k The number of samples, N It represents the total number of samples at the current node.

[0066] The Gini index is commonly used to measure impurities. The metric is represented as follows:

[0067] in, It is the first i The proportion of samples in each category, where C is the set of all categories.

[0068] In the construction of the fault diagnosis model based on random forest, each decision tree is independently trained based on randomly sampled data and feature subsets, forming a set of differential base classifiers. The final predicted category y is determined by a voting method, as shown in the following formula:

[0069] in, It represents the total number of decision trees; It is the prediction result of the j-th tree for sample x; It is an indicator function, which is 1 if the condition is met, and 0 otherwise; C is the set of all categories; It is the category of the final prediction.

[0070] In random forests, the importance of a feature can be evaluated based on the reduction in node impurity. The importance of a feature is calculated by the cumulative reduction in impurity across all nodes split using that feature:

[0071] in, The impurity accumulation for all nodes, It is the amount by which the impurity of feature f is reduced when it is split on decision tree t.

[0072] Secondly, a random forest diagnostic model is constructed, and the basic parameters of the model are initialized, including the number of decision trees and the maximum number of features that each tree can consider. When constructing each decision tree, random sampling is first performed, drawing samples with replacement from the training set to form a training subset. Then, when splitting a node, features are randomly selected as candidates. Next, the best feature is selected for splitting based on the principle of maximizing information gain, and the entire tree is recursively constructed until the stopping condition is met. Finally, all decision trees are integrated to form a random forest diagnostic model, and a voting mechanism is used to predict new samples.

[0073] The next step is to optimize the random forest diagnostic model by using grid search or random search to find the optimal combination of hyperparameters, including the number of decision trees and the maximum number of features, within a predefined parameter space. Simultaneously, cross-validation is employed to evaluate the model's performance under different hyperparameter combinations, ensuring the model has good generalization ability.

[0074] Finally, the optimized model was evaluated using metrics such as accuracy, precision, recall, and F1 score to measure its performance on the test set. A confusion matrix was used to visually demonstrate the model's prediction performance across different failure modes (e.g., Figure 14 As shown in the figure, the reasons for misdiagnosis and missed diagnosis are analyzed to provide a basis for subsequent model improvement.

[0075] Step 104: Based on the structure and working principle of the hydraulic gate hoist, construct a fault tree model and use the fault tree analysis method to analyze the fault modes.

[0076] First, it is essential to define the foundation for fault tree construction. This includes a deep understanding of the structure, working principle, and interaction between components of the hydraulic gate hoist. It also involves identifying the fault modes that need to be analyzed, including the fault location of the hydraulic gate hoist, the possible fault phenomena of each location, and the corresponding causes of each fault phenomenon.

[0077] Secondly, a fault tree is constructed. First, the top event is defined, representing the overall failure of the hydraulic hoist. Then, the direct causes leading to the top event are identified as intermediate events, such as system leakage. Next, the causes of the intermediate events are further analyzed to find the smallest, indivisible fault cause as the basic event, such as wear of the hydraulic cylinder seals. Finally, logic gates are used to connect the top event, intermediate events, and basic events to form a complete fault tree structure, such as... Figure 15-21 As shown.

[0078] The next step is to analyze the fault tree. First, the minimum cut set algorithm is used to solve for the minimum cut sets of the fault tree. These minimum cut sets represent the minimum combination of basic events that leads to the occurrence of the top event. Then, a qualitative analysis is performed based on the minimum cut sets to determine which basic events or combinations of basic events have the greatest impact on the occurrence of the top event. With sufficient data, a quantitative analysis can also be performed to calculate the probability of occurrence of each basic event and intermediate event, and to assess the probability and risk level of the top event.

[0079] Finally, the results of fault tree analysis are applied to practice, and a detailed fault diagnosis process is developed based on the analysis results to guide technicians to quickly locate the cause of the fault.

[0080] Step 105: Input the real-time data collected by the sensor into the random forest diagnostic model. Based on the fault mode output by the model and the fault tree analysis results, determine the specific cause of the fault.

[0081] First, real-time data acquisition is carried out by installing sensors on key parts of the hydraulic hoist to collect monitoring data reflecting the equipment's operating status in real time, including but not limited to the outlet pressure of the hydraulic pump, the extension and retraction speed of the hydraulic cylinder, the on / off status of the control valve, and key parameters such as oil temperature and oil level, to ensure the comprehensiveness and accuracy of the data.

[0082] Secondly, the collected real-time data undergoes preprocessing. This step is similar to the data preprocessing in the second step, but with a greater emphasis on real-time performance and efficiency. This includes data cleaning to remove noise and outliers, data normalization to eliminate the influence of different dimensions and units, and necessary feature extraction and transformation to provide a suitable input format for the subsequent random forest fault diagnosis model.

[0083] The next step is to input the pre-processed real-time data into the pre-trained random forest fault diagnosis model. Based on the characteristics of the input data, the model will quickly predict the possible fault modes of the current hydraulic gate hoist. This process leverages the powerful classification capabilities of the random forest diagnostic model, enabling it to provide accurate fault prediction results in a short time.

[0084] Next, based on the fault tree analysis results constructed in step four, an in-depth analysis is conducted on the fault modes predicted by the random forest diagnostic model.

[0085] By finding the minimum cut set in the fault tree, the most likely cause or combination of causes leading to the fault mode is determined. This step leverages the advantages of fault tree analysis in fault cause localization, providing specific direction and basis for troubleshooting.

[0086] Finally, the diagnostic results (including predicted failure modes and determined failure causes) are displayed in real time on the operating interface, and a detailed failure report is generated. Operators can quickly locate the fault and take appropriate maintenance measures based on this information. Simultaneously, the system can automatically trigger early warning or shutdown protection mechanisms based on the fault diagnosis results to prevent further escalation of the fault and ensure the safe operation of the hydraulic gate hoist.

[0087] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0088] The above descriptions are merely embodiments of this application and are not intended to limit the scope of this application. For those skilled in the art, various modifications and variations can be made to the embodiments of this application. These modifications or substitutions do not cause the essence of the corresponding technical solutions to depart from the spirit and scope of the technical solutions in the embodiments of this application.

Claims

1. A dynamic fault diagnosis method for hydraulic gate hoists based on random forest and fault tree, characterized in that: Includes the following steps: A. Develop a fault simulation plan and collect fault sample data; A digital simulation model of the hydraulic gate hoist was built using professional simulation software. Different sensors were set up in key components, a fault simulation scheme was developed, and fault sample data was collected. B. Preprocess the fault data; C. Construct and optimize a random forest model; Organize and select feature and fault diagnosis datasets, use multiple decision trees to train the model, optimize the hyperparameters of random forest through grid search, and use cross-validation to assess model performance. D. Analyze failure modes; Based on the structure and working principle of the hydraulic gate hoist, a fault tree model is constructed, and the fault tree analysis method is used to analyze the fault modes. E. Determine the specific cause of the malfunction. The sensor collects data in real time and inputs it into the random forest diagnostic model. Based on the fault mode output by the random forest diagnostic model and the fault tree analysis results, the specific cause of the fault is determined.

2. The dynamic fault diagnosis method for hydraulic gate hoists based on random forest and fault tree as described in claim 1, characterized in that: The method for developing a fault simulation scheme and collecting fault sample data as described in step A includes the following steps: A1. Building a simulation model A digital simulation model of a hydraulic gate hoist is built using professional simulation software. The digital simulation model accurately simulates the hydraulic gate hoist and sets physical parameters that conform to reality. The hydraulic gate hoist includes a hydraulic pump, hydraulic cylinder, control valve, oil pipes, and seals. The physical parameters include the viscosity and density of the hydraulic oil, the length, diameter, and roughness of the pipes, as well as the elastic modulus and damping coefficient of each component. A2. Based on the working principle and operation process of the hydraulic gate hoist, set the corresponding control logic and boundary conditions in the digital simulation model. The control logic includes the start and stop control of the hydraulic pump, the extension and retraction control of the hydraulic cylinder, and the opening and closing logic of the valve. The boundary conditions include the initial pressure, load conditions, simulation time, and external constraints. A3. Filter fault modes; A31. Identify the failure modes of hydraulic gate hoists through literature review, expert consultation and analysis of actual cases; The failure modes include leakage, pressure loss, system blockage, hydraulic pump cavitation, and control valve failure. A32. Provide a detailed description of each failure mode, including the failure phenomenon, failure cause, and impact on performance, to provide a clear objective for subsequent fault injection and diagnosis; A33. Set up a fault injection scheme: First, design a corresponding fault injection scheme for each of the selected fault modes; The fault injection scheme includes the conditions for fault occurrence, the range of variation of fault parameters, and the duration of the fault. A34. Implement fault injection functionality in the digital simulation model; modify the parameter settings of the digital simulation model or introduce additional control logic to trigger faults according to a predetermined scheme during the simulation process; then, perform multiple simulations for each fault mode, adjusting the values ​​of fault parameters each time to obtain monitoring data under different fault levels; these data will serve as the basis for subsequent training and testing of the random forest diagnostic model; finally, record the monitoring data during each simulation process, as well as the time point and duration of the fault occurrence, to provide complete information for subsequent data analysis.

3. The dynamic fault diagnosis method for hydraulic gate hoists based on random forest and fault tree as described in claim 1, characterized in that: The method for preprocessing fault data described in step B includes the following steps: B1. Preprocess the collected fault dataset, including data cleaning and normalization. The normalization method is to scale the data to the range [0,1] using the min-max scaling method, as shown in the following formula: in For the reduced data, This is the original data. The minimum value of the original data. The maximum value of the original data; B12. Feature selection is performed using the Lasso regression algorithm; The Lasso regression algorithm is used to find a set of feature coefficients that minimizes the error between the predicted values ​​of the random forest diagnostic model and the actual fault labels. While minimizing the error, a penalty is applied to the sum of the absolute values ​​of all feature coefficients. The Lasso regression algorithm tends to compress the coefficients of features that contribute weakly or unstablely to the prediction target to exactly zero; a coefficient of zero means that the feature is excluded by the Lasso regression algorithm. The cost function of the Lasso regression algorithm is as follows: in For the first i The actual value of each observed sample. For the first j The regression coefficients corresponding to each feature For the first i The first observation sample j 1 eigenvalue, This is the regularization strength parameter.

4. The dynamic fault diagnosis method for hydraulic gate hoists based on random forest and fault tree as described in claim 1, characterized in that: The method for constructing and optimizing the random forest diagnostic model described in step C includes the following steps: C1. Data preparation: Integrate the preprocessed hydraulic gate hoist fault dataset, clearly divide the training set and test set, and ensure that each sample has an accurate and consistent fault mode label. C2. Building Decision Trees: Decision trees are the basic unit of random forests, and each decision tree is trained independently. When building a decision tree, the next node split will randomly select m features from M candidate features. The feature with the highest information gain will be selected to split the decision tree; maximizing the information gain of the splitting node... It is expressed as follows: in, Indicates the impurity of the parent node. The impurity of the child nodes. It is a child node k The number of samples, N This represents the total number of samples in the current node; Impurity is measured by the Gini index. The metric is represented as follows: in, It is the first i The proportion of samples in each category, where C is the set of all categories; In the construction of the random forest diagnostic model, each decision tree is independently trained based on randomly sampled data and feature subsets, forming a set of differentially distributed base classifiers. The final predicted category is then determined through a voting method. The formula is as follows: in, It represents the total number of decision trees; It is the first j Decision trees for samples x The prediction results; It is an index function; it is 1 if the condition is met, and 0 otherwise. c Represents a candidate category label; In random forests, the importance of a feature is evaluated based on the reduction in node impurity. The importance of a feature is calculated by the cumulative reduction in impurity across all nodes split using that feature, as shown in the following formula: in, The impurity accumulation for all nodes, It is a feature f In decision tree t The amount of impurity reduction during the splitting process; C3. Constructing a Random Forest Diagnostic Model: Initialize the basic parameters of the random forest diagnostic model, setting the number of decision trees and the maximum number of features per decision tree; when constructing each decision tree, firstly, perform random sampling, drawing samples with replacement from the training set to form a training subset; then, randomly select features as candidates when splitting nodes; next, select the best feature for splitting based on the principle of maximizing information gain, and recursively construct the entire decision tree until the stopping condition is met; finally, integrate all decision trees to form a random forest diagnostic model, and use a voting mechanism to predict new samples; C4. Optimize the random forest diagnostic model by using grid search or random search to find the optimal hyperparameter combination within a predefined parameter space. The hyperparameter combination includes the number of decision trees and the maximum number of features. At the same time, cross-validation is used to evaluate the performance of the random forest diagnostic model under different hyperparameter combinations to ensure that the random forest diagnostic model has good generalization ability. C5. Evaluate the optimized random forest diagnostic model, using accuracy, precision, recall, and F1 score to measure its performance on the test set. Visually demonstrate the model's prediction performance under different failure modes using a confusion matrix, analyze the causes of misdiagnosis and missed diagnosis, and provide a basis for improving the random forest diagnostic model.

5. The dynamic fault diagnosis method for hydraulic gate hoists based on random forest and fault tree as described in claim 1, characterized in that: The method for analyzing failure modes described in step D includes the following steps: D1. Define the basis for fault tree construction; The foundation for the fault tree construction includes the structural composition, working principle, and interaction relationship between the components of the hydraulic gate hoist. It also clarifies the fault modes that need to be analyzed, including the fault location of the hydraulic gate hoist, the fault phenomenon occurring in each location, and the corresponding fault cause for each fault phenomenon. D2. Construct a fault tree; First, define the top event, which is the overall failure of the hydraulic hoist. Then, identify the direct cause of the top event as an intermediate event. Next, further analyze the causes of the intermediate events to find the smallest fault cause that cannot be further decomposed as the basic event. Finally, use logic gates to connect the top event, intermediate events, and basic events to form a complete fault tree structure. D3. Analyze the fault tree; First, the minimum cut set algorithm is used to solve for the minimum cut sets of the fault tree. These minimum cut sets represent the minimum combination of basic events that lead to the occurrence of the top event. Then, a qualitative analysis is performed based on the minimum cut sets to determine which basic events or combinations of basic events have the greatest impact on the occurrence of the top event. With sufficient data, a quantitative analysis is further performed to calculate the probability of occurrence of each basic event and intermediate event, and to assess the probability and risk level of the top event. D4. Apply the results of fault tree analysis; Based on the analysis results, a detailed troubleshooting process is developed to guide technicians in quickly locating the cause of the fault.

6. The dynamic fault diagnosis method for hydraulic gate hoists based on random forest and fault tree as described in claim 1, characterized in that: Step E, the method for developing a fault simulation scheme and collecting fault sample data, includes the following steps: E. Determine the specific cause of the malfunction. E1. Real-time data acquisition: By installing sensors at key parts of the hydraulic hoist, real-time monitoring data reflecting the operating status of the equipment is collected. The monitoring data includes the outlet pressure of the hydraulic pump, the extension and retraction speed of the hydraulic cylinder, the opening and closing status of the control valve, the valve inlet and outlet pressure, flow rate, and the time and duration of the fault occurrence. E2. Preprocess the collected real-time data: Preprocess the real-time data according to the method in step B; E3. Input the preprocessed real-time data into the trained random forest diagnostic model; the random forest diagnostic model quickly predicts the fault modes of the current hydraulic gate hoist based on the characteristics of the input data. E4. Based on the fault tree analysis results constructed in step D, conduct an in-depth analysis of the fault modes predicted by the random forest diagnostic model. By finding the minimum cut set in the fault tree, the cause or combination of causes leading to the fault mode can be determined, and a diagnostic result can be formed. E5. The diagnostic results are displayed on the operation interface in real time, and a detailed fault report is generated; the operator can quickly locate the fault location based on this information and take corresponding maintenance measures.

Images