Unmanned mine car braking system health assessment method based on deep learning

Abstract

The invention relates to an unmanned mine car braking system health assessment method based on deep learning, belongs to the technical field of fault prediction and health assessment, and solves the problem that the stability and safety of a braking system are affected due to the fact that health state assessment of an unmanned mine car is not comprehensive enough in the prior art. The data extraction module obtains original data; s2, processing the original data to obtain preprocessed data; s3, the data processing module performs data cleaning to obtain cleaned data; s4, the data processing module performs processing to obtain data passing verification; s5, the model training module constructs and trains a braking system performance prediction model; s6, optimizing to obtain an optimized braking system performance prediction model; step S7, the health assessment module calculates a residual error between a model prediction value and a real value; step S8, performing trend analysis on the residual error; and step S9, performing health state evaluation based on the residual error and the trend analysis result, and sending out an early warning prompt.

Description

Technical Field

[0001] The present invention relates to the technical field of fault prediction and health assessment, and in particular to a health assessment method for an unmanned mining vehicle braking system based on deep learning. Background Art

[0002] The mining and utilization of mineral resources are key factors in promoting the process of industrialization and urbanization. The introduction of unmanned mining vehicles is a major innovation in the traditional mining production model. Through advanced automation and intelligent technologies, unmanned mining vehicles have achieved precise adaptation and efficient operation in complex mining environments, which not only improves production efficiency, but also greatly reduces the risk of accidents caused by human factors. In complex mining environments, unmanned mining vehicles need to frequently accelerate, decelerate, stop and other operations, all of which are inseparable from the effective support of the braking system. Once the braking system fails or performance declines, it will directly affect the braking effect and driving stability of the vehicle, thereby increasing the probability of accidents. Therefore, ensuring the healthy operation of the braking system is the premise and basis for the safe operation of unmanned mining vehicles. By real-time monitoring and evaluation of the health status of the braking system, potential faults and performance decline trends can be discovered in a timely manner, so as to take preventive maintenance measures to avoid the occurrence and expansion of faults. This can not only improve the safety and stability of the vehicle, but also extend the service life of the braking system and reduce maintenance costs. Health management can also improve the overall operating efficiency of unmanned mining vehicles.

[0003] Traditional health assessment methods focus more on the status analysis of a single device in a mine car and present the results in the form of graphs. They lack digital characterization of the structure, properties, operation process, etc. of the mine car and its components. The assessment content is not deep and rich enough, and it is difficult to fully reflect the overall performance of the mine car.

[0004] The health assessment methods currently available in the prior art lack health status assessment of non-critical components, resulting in an incomplete health status assessment of the entire system, which in turn affects the stability and safety of the braking system.

[0005] In addition, the current digital health assessment system requires the collaborative work of a large number of technical modules, and the system integration is highly complex. In addition, in order to ensure the stable operation and continuous optimization of the system, a high level of technical support and maintenance is also required, resulting in high costs for system construction and maintenance, which has become the main factor limiting its widespread application.

[0006] In summary, the health assessment methods in the prior art are lacking in the coverage of health assessment and have problems in the complexity and cost of system integration. Therefore, there is a need in the technical field for a health assessment method that can solve the above problems. Summary of the invention

[0007] In order to solve the above problems, the present invention provides a data-driven unmanned mine car brake system health assessment system. The system is designed to comprehensively and accurately evaluate the health status of the unmanned mine car brake system. In particular, the current distributed brake system health assessment method mainly focuses on key components (such as air charging solenoid valves and exhaust solenoid valves). The present invention is committed to achieving a comprehensive assessment of the overall health status of the brake system. This improvement includes timely detection and early warning of the overall performance of the brake system, thereby avoiding the potential impact of a single component failure on the performance of the entire system, thereby ensuring the comprehensiveness and accuracy of the assessment results. In view of the fact that the digital twin system can achieve high-precision evaluation but the technical integration is complex and costly, the present invention has taken the following optimization measures: optimize the data acquisition, processing and analysis process, adopt a lightweight and efficient data-driven model, reduce dependence on complex technical modules, so as to reduce construction and maintenance costs, and enable more mining companies to afford and effectively use the system.

[0008] According to one embodiment of the present invention, a method for health assessment of a brake system of an unmanned mining vehicle based on deep learning is provided, comprising the following steps:

[0009] Step S1, the data extraction module connects to the data source, and automatically captures data closely related to the brake system performance evaluation from the data source as raw data, where the data closely related to the brake system performance evaluation includes brake pedal opening, vehicle speed, acceleration, and driving slope;

[0010] Step S2, the data extraction module performs data screening and data verification on the original data to obtain pre-processed data;

[0011] Step S3, the data processing module receives the preprocessed data, and performs data cleaning on the preprocessed data to obtain cleaned data;

[0012] Step S4, the data processing module performs feature selection on the cleaned data to obtain highly relevant feature data, and further performs data conversion and normalization, as well as data verification, to obtain and store the verified data;

[0013] Step S5, the model training module uses the obtained highly correlated feature data to construct and train a brake system performance prediction model through an LSTM algorithm;

[0014] Step S6, the model training module optimizes the trained brake system performance prediction model to obtain an optimized brake system performance prediction model;

[0015] Step S7, the health assessment module obtains a model prediction value based on the optimized brake system performance prediction model, and calculates the residual between the model prediction value and the real value collected in real time by the mine car sensor;

[0016] Step S8, the health assessment module performs trend analysis on the residuals to obtain the residual variation trend over time as the trend analysis result, wherein the residual variation trend includes whether it is stable, the magnitude of the fluctuation, whether there is an obvious increasing trend and periodic fluctuation;

[0017] Step S9, the health assessment module performs a health status assessment based on the residual value and trend analysis results. If the residual changes stably within a preset range, the braking system is determined to be in a healthy state. Otherwise, the braking system is determined to be in an abnormal state and an early warning reminder is issued.

[0018] Optionally, the step S2 specifically includes the following steps:

[0019] Step S2.1, the data extraction module performs data screening, including using statistical analysis and threshold setting to identify abnormal fluctuation points in the original data;

[0020] Step S2.2, using interpolation and mean substitution to correct and remove outliers and noise related to abnormal fluctuation points in the original data to obtain filtered data;

[0021] Step S2.3, perform data verification on the screened data to obtain preprocessed data.

[0022] Optionally, the step S3 specifically includes the following steps:

[0023] Step S3.1, the data processing module receives the pre-processed key data and identifies key information therefrom;

[0024] Step S3.2, using a moving average filtering algorithm to smooth and filter the identified key information to obtain cleaned data.

[0025] Optionally, the step S4 specifically includes the following steps:

[0026] Step S4.1, the data processing module calculates the Pearson correlation coefficient between the features of the cleaned data, identifies the features that are highly correlated with the brake system performance evaluation, and retains them as highly correlated feature data;

[0027] Step S4.2, the data processing module performs data conversion and normalization processing on the highly correlated feature data to obtain converted data;

[0028] Step S4.3, the data processing module verifies the converted data to see whether it meets the expected format and quality standards. If so, the converted data is stored as the verified data. Otherwise, it returns to step S1 to re-extract the original data.

[0029] Optionally, the step S5 specifically includes the following steps:

[0030] Step S5.1, constructing an LSTM model through an LSTM algorithm;

[0031] Step S5.2, setting the initial parameters of the LSTM model, including setting the number of network layers of the LSTM model, the number of units in each network layer, and the loss function;

[0032] Step S5.3, using the obtained highly correlated feature data as a training set, the constructed LSTM model is trained to obtain a trained braking system performance prediction model.

[0033] Optionally, step S7 specifically includes the following steps:

[0034] Step S7.1, the health assessment module receives the verified data from the data processing module as real-time data, and inputs the real-time data into the optimized brake system performance prediction model to obtain the model prediction value;

[0035] Step S7.2, compare the model prediction value with the true value to obtain the residual.

[0036] Compared with the prior art, the unmanned mining vehicle braking system health assessment method based on deep learning provided by the embodiments of the present invention has at least the following advantages.

[0037] The deep learning-based health assessment method for the unmanned mining vehicle braking system provided in the embodiment of the present invention realizes the diagnosis of its health status by comprehensively evaluating the overall performance of the braking system, rather than being limited to the analysis of single or specific key components. Thus, it has the sensitivity to instantly capture abnormal conditions in the braking system, ensuring that once the system performance deviates from the normal standard, it can be quickly and accurately detected and identified, thereby providing users with timely and reliable maintenance suggestions to ensure driving safety.

[0038] The implementation method of the present invention provides a deep learning-based unmanned mining vehicle braking system health assessment method that utilizes the powerful capabilities of deep learning networks. While significantly reducing implementation costs, it can still achieve an accurate assessment of the health status of the braking system. And through the efficient data processing and analysis capabilities of the deep learning network, it can more directly and economically capture and understand the performance changes of the braking system, providing an economical, efficient and reliable maintenance solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. The features and advantages of the present invention can be more clearly understood by referring to the drawings. The drawings are schematic and should not be understood as limiting the present invention in any way. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0040] Figure 1 A system schematic diagram of a deep learning-based health assessment method for an unmanned mining vehicle braking system according to an embodiment of the present invention.

[0041] Figure 2 The present invention is a flowchart of a data extraction process of a deep learning-based unmanned mining vehicle braking system health assessment method provided according to an embodiment of the present invention.

[0042] Figure 3 The present invention is a flowchart of a data processing method for health assessment of an unmanned mining vehicle braking system based on deep learning according to an embodiment of the present invention.

[0043] Figure 4 The present invention is a flowchart of a model training process of a deep learning-based unmanned mining vehicle braking system health assessment method provided according to an embodiment of the present invention.

[0044] Figure 5 The present invention is a flowchart of a health assessment process of a deep learning-based health assessment method for an unmanned mining vehicle braking system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0046] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0047] The following describes in detail a method for health assessment of an unmanned mining vehicle braking system based on deep learning according to an embodiment of the present invention with reference to the accompanying drawings.

[0048] like Figure 1As shown, a data-driven unmanned mine car brake system health assessment system provided according to an embodiment of the present invention is intended to achieve accurate assessment of the health status of the unmanned mine car brake system through efficient data processing and intelligent analysis technology. The system as a whole is composed of multiple core modules, and each module works closely together to complete the comprehensive monitoring and assessment of the health status of the brake system. The core modules include: data extraction module, data processing module, model training module and health assessment module.

[0049] The data extraction module automatically extracts key information from massive data to provide data support for subsequent analysis. As the input end of the system, it is responsible for automatically capturing data closely related to the performance evaluation of the braking system from historical real vehicle logs, including key parameters such as brake pedal opening, vehicle speed, acceleration, and driving slope. These data are the basis for subsequent analysis.

[0050] The model training module uses advanced algorithms to build a prediction model to achieve accurate prediction of the performance of the brake system. It uses processed historical data to build a prediction model through machine learning or deep learning algorithms. This module is responsible for optimizing model parameters to improve the model's sensitivity to changes in brake system performance and prediction accuracy. To ensure the reliability and accuracy of the model in practical applications, after the model training is completed, it uses cross-validation and other methods to comprehensively evaluate the model's performance, including indicators such as accuracy and stability, to ensure that the model can achieve the expected evaluation effect in practical applications. Output the prediction results of the health status of the brake system. It is connected to the real-time data interface, and the real-time collected brake system-related data is input into the trained model. The model outputs prediction values, which reflect the current health status or potential problems of the brake system.

[0051] The health assessment module comprehensively judges the health status of the brake system based on the prediction results and provides decision support. Based on the output of the model prediction module and the preset health standards or thresholds, it makes a final assessment of the health status of the brake system and may generate alarms or maintenance recommendations to guide operators or maintenance teams to take appropriate measures.

[0052] In the data-driven unmanned mine car brake system health assessment system, the data processing module plays a vital role. Given the complexity of the open-pit mine environment, the directly collected information often contains a lot of noise and outliers. These bad data will seriously affect the accuracy of subsequent model training. Therefore, the data processing module uses a series of carefully designed preprocessing steps to deeply clean and optimize the original data to ensure the effectiveness of model training. For the original data collected in the complex environment of the open-pit mine, the data processing module first identifies and processes noise and outliers. Specifically, through statistical analysis and threshold setting, the abnormal fluctuation points in the data are identified, and appropriate methods (such as interpolation, mean substitution, etc.) are used to correct or eliminate them. This process effectively reduces the impact of noise and outliers on model training. In order to improve the accuracy of the model, the data processing module filters the key information in the original data. These key information include speed, slope, acceleration, pitch angle, roll angle, yaw angular velocity, and speed deviation changes, which are crucial for the evaluation of braking system performance. For these key information, the module uses algorithms such as moving average filtering for smoothing to reduce the interference of random noise and improve the stability and reliability of the data. After completing the noise and outlier processing and filtering, the data processing module further performs feature selection and optimization. By calculating the Pearson correlation coefficient between the features of the original data, the module can identify features that are highly correlated with the performance evaluation of the braking system. These highly correlated features will be retained as feature inputs for the next step of model training, while features with weaker correlation may be eliminated or downgraded. This process helps to reduce the complexity of the model and improve the training efficiency and prediction accuracy of the model.

[0053] The model training module is one of the core components of the entire health assessment system. It is responsible for using pre-processed high-quality data to build a prediction model that can accurately predict the performance of the brake system through advanced machine learning or deep learning algorithms. This module not only focuses on the model construction process, but also optimizes the model parameters to improve the model's sensitivity to changes in brake system performance and the accuracy of prediction. LSTM is a special type of recurrent neural network (RNN) that is particularly good at processing time series data. The performance indicators of the brake system, such as brake pedal opening, vehicle speed, acceleration, etc., all change dynamically over time and have significant time series characteristics. Therefore, the LSTM algorithm can make full use of the time dependencies in these time series data to make more accurate predictions. Traditional RNNs are prone to gradient vanishing or gradient exploding problems when processing long time series, resulting in the inability to effectively capture long-term dependencies. LSTM effectively solves this problem by introducing structures such as forget gates, input gates, and output gates. It can capture and retain long-term dependency information in time series, thereby improving the accuracy of predictions. The LSTM algorithm has been widely used and verified in the field of time series prediction, including financial prediction, weather prediction, natural language processing, and many other fields. In the field of brake system health assessment, LSTM can also demonstrate its powerful predictive ability.

[0054] As the core component of the unmanned mine car brake system health assessment system, the health assessment module undertakes the important task of converting the model prediction results into actual health status assessment. This module accurately judges the health status of the brake system by deeply analyzing the residuals between the model prediction values ​​and the true values, as well as the changing trends of these residuals over time. When the brake system is in a healthy state, the model can accurately capture its operating rules, so the residuals between the predicted values ​​and the true values ​​are small, and the changes in the residuals remain stable within the allowable range. In this case, it can be considered that the brake system is in good operating condition and no special attention is needed.

[0055] However, when the brake system is abnormal or about to fail, its actual operating data will deviate from the established benchmark model, resulting in a significant increase in the residual between the predicted value and the true value, and a significant trend change may occur. The trend of this residual change is the key basis for evaluating the health status of the brake system. First, the health assessment module calculates the residual between the model predicted value and the true value. The size of the residual directly reflects the accuracy of the model prediction and is also an important indicator for evaluating the health status of the brake system. Then, the module performs a trend analysis on the residual. By observing the trend of the residual over time, it can be judged whether the operating status of the brake system is stable. If the residual fluctuates within a relatively stable range and the fluctuation amplitude is small, it means that the brake system is in a healthy state; if the residual shows a significant increase trend or periodic fluctuation, it may indicate that there are potential problems in the brake system. Based on the size of the residual and the trend analysis results, the health assessment module determines the health status of the brake system. If the residual changes steadily within the allowable range, the brake system is judged to be in a healthy state; if the residual exceeds the allowable range and shows a significant trend change, the brake system is judged to be in an abnormal state and may be about to fail.

[0056] refer to Figures 1 to 5 According to an embodiment of the present invention, a method for health assessment of an unmanned mining vehicle braking system based on deep learning is provided, which includes the following steps.

[0057] Step S1, the data extraction module obtains driving data, connects to the data source, and automatically captures data closely related to the brake system performance evaluation from the data source as raw data. Optionally, the data source can be a historical real vehicle log database. The data closely related to the brake system performance evaluation may include key parameters such as brake pedal opening, vehicle speed, acceleration, and driving slope. Figure 2 As shown, in step S1, key information is automatically extracted from massive data through the data extraction module to provide data support for subsequent analysis. As the input end of the system, the data extraction module automatically captures data closely related to the performance evaluation of the braking system from the historical real vehicle log database, including key parameters such as brake pedal opening, vehicle speed, acceleration, and driving slope. These data are the basis for subsequent analysis.

[0058] Step S2: The data extraction module performs data screening and data verification on the original data to obtain pre-processed data. Figure 2 As shown, step S2 specifically includes the following steps.

[0059] Step S2.1, the data extraction module performs data screening, including using statistical analysis and threshold setting to identify abnormal fluctuation points in the original data.

[0060] Step S2.2, using interpolation, mean substitution and other methods to correct or eliminate the outliers and noise related to the abnormal fluctuation points in the original data to obtain the screened data.

[0061] In view of the complexity of the open-pit mining environment, the information directly collected often contains a large amount of noise and outliers. These bad data will seriously affect the accuracy of subsequent model training. In this step S2, the data processing module performs a series of preprocessing steps to deeply clean and optimize the original data to ensure the effectiveness of model training. For the original data collected in the complex environment of the open-pit mining area, the data processing module first identifies and processes noise and outliers. Specifically, through statistical analysis and threshold setting and other methods, abnormal fluctuation points in the data are identified, and appropriate methods (such as interpolation, mean substitution, etc.) are used to correct or eliminate them. This process effectively reduces the impact of noise and outliers on model training.

[0062] Step S2.3, perform data verification on the screened data to obtain pre-processed data. The data verification process can ensure the integrity and accuracy of the data.

[0063] Step S3: The data processing module receives the preprocessed data and performs data cleaning on the preprocessed data to obtain cleaned data. Figure 3 As shown, step S3 specifically includes the following steps.

[0064] Step S3.1, the data processing module receives the pre-processed key data and identifies key information therefrom. The key information may include speed, slope, acceleration, pitch angle, roll angle, yaw rate, and speed deviation change, which are crucial for evaluating the performance of the braking system.

[0065] Step S3.2, smoothing and filtering the identified key information using algorithms such as moving average filtering to obtain cleaned data. In order to improve the accuracy of the model, the identified key information is filtered by the data processing module. For these key information, the data processing module uses algorithms such as moving average filtering for smoothing to reduce the interference of random noise and improve the stability and reliability of the data.

[0066] Step S4: The data processing module performs feature selection on the cleaned data to obtain highly relevant feature data, and further performs data conversion and normalization, as well as data verification, to obtain and store the verified data. Figure 3 As shown, step S4 specifically includes the following steps.

[0067] Step S4.1, the data processing module calculates the Pearson correlation coefficient between the features of the cleaned data, identifies the features that are highly correlated with the brake system performance evaluation, removes or downgrades the features in the original data that are less correlated with the brake system performance evaluation, and retains the features that are highly correlated with the brake system performance evaluation as highly correlated feature data, which are used as feature input for model training. This process helps reduce the complexity of the model and improve the training efficiency and prediction accuracy of the model.

[0068] Step S4.2: the data processing module performs data conversion and normalization processing on the highly correlated feature data to obtain converted data.

[0069] Step S4.3, the data processing module verifies the converted data to see whether it meets the expected format and quality standards. If so, the converted data is stored as the verified data. Otherwise, it returns to step S1 to re-extract the original data.

[0070] Step S5: The model training module uses the obtained highly correlated feature data to construct and train a brake system performance prediction model through an LSTM (Long Short-Term Memory Neural Network) algorithm, and obtains the brake system performance prediction model, which outputs acceleration.

[0071] like Figure 4 As shown, in step S5, the model training module uses advanced algorithms to construct a brake system performance prediction model to achieve accurate prediction of brake system performance, and uses processed historical data, that is, the highly correlated feature data obtained, to construct a prediction model through machine learning or deep learning algorithms. The model training module is one of the core components of the entire health assessment system. It uses pre-processed high-quality data and advanced machine learning or deep learning algorithms to construct a prediction model that can accurately predict the performance of the brake system. LSTM is a special type of recurrent neural network (RNN) that is good at processing time series data. The performance indicators of the brake system, such as brake pedal opening, vehicle speed, acceleration, etc., all change dynamically over time and have significant time series characteristics. Therefore, the LSTM algorithm can make full use of the time dependency in these time series data to make more accurate predictions. Traditional RNNs are prone to gradient vanishing or gradient explosion problems when processing long time series, resulting in the inability to effectively capture long-term dependencies. LSTM effectively solves this problem by introducing structures such as forget gates, input gates, and output gates, and can capture and retain long-term dependency information in time series, thereby improving the accuracy of predictions.

[0072] like Figure 4 As shown, step S5 specifically includes the following steps.

[0073] Step S5.1, the model training module constructs an LSTM model through the LSTM algorithm.

[0074] Step S5.2, set the initial parameters of the LSTM model, including setting the number of network layers of the model, the number of units in each layer of the network, the loss function, etc.

[0075] Step S5.3, using the obtained highly correlated feature data as a training set, the constructed LSTM model is trained to obtain a trained braking system performance prediction model.

[0076] Step S6, optimizing the trained brake system performance prediction model to obtain an optimized brake system performance prediction model, and improving the model's sensitivity to brake system performance changes and prediction accuracy. The optimization includes using methods such as cross-validation to comprehensively evaluate the performance of the brake system performance prediction model, including indicators such as accuracy and stability, to ensure that the brake system performance prediction model can achieve the expected evaluation effect in practical applications.

[0077] In step S6, the parameters of the brake system performance prediction model are optimized through the model training module to improve the sensitivity and prediction accuracy of the brake system performance prediction model to changes in brake system performance, thereby ensuring the reliability and accuracy of the brake system performance prediction model in practical applications.

[0078] Step S7, the health assessment module obtains the model prediction value based on the optimized brake system performance prediction model, and calculates the residual between the model prediction value and the real value collected in real time by the mine car sensor. The size of the residual directly reflects the accuracy of the brake system performance prediction model prediction and is an important indicator for evaluating the health status of the brake system.

[0079] In step S7, the health assessment module is connected to the real-time data interface, and the verified data received from the data processing module is input as real-time data into the trained braking system performance prediction model through the real-time data interface, and the model prediction values ​​are output. These model prediction values ​​reflect the health status or potential problems of the current braking system.

[0080] The step S7 specifically includes the following steps.

[0081] Step S7.1, the health assessment module receives the verified data from the data processing module as real-time data, and inputs the real-time data into the optimized brake system performance prediction model to obtain the model prediction value.

[0082] Step S7.2, compare the model prediction value with the real value collected in real time by the mine car sensor to obtain the residual.

[0083] In this step S7, it can also include that the health assessment module comprehensively judges the health status of the brake system based on the model prediction value, outputs the prediction result of the health status of the brake system, and provides decision support. Among them, the comprehensive judgment of the health status of the brake system includes the model prediction value output by the model prediction module, combined with the preset health standard or threshold, to make a final evaluation of the health status of the brake system to obtain the prediction result, and can generate an alarm or maintenance suggestion to guide the operator or maintenance team to take corresponding measures. In step S8, the health assessment module performs trend analysis on the residual to obtain the residual change trend over time as the trend analysis result, and the residual change trend includes whether it is stable, the amplitude of fluctuation, and whether there is an obvious increase trend or periodic fluctuation.

[0084] Step S9, the health assessment module performs a health status assessment based on the size of the residual and the trend analysis results. If the residual changes stably within the allowable range and the fluctuation amplitude is small, the braking system is judged to be in a healthy state. If the residual exceeds the allowable range and shows an obvious trend change (such as an increasing trend or periodic fluctuation), the braking system is judged to be in an abnormal state and may be about to fail, and an early warning reminder is issued.

[0085] like Figure 5 As shown, in step S7 to step S9, the health assessment module, as the core component of the health assessment system of the unmanned mining vehicle braking system, converts the model prediction value into the actual health status assessment. This module accurately judges the health status of the braking system by deeply analyzing the residuals between the model prediction value and the true value, as well as the changing trend of these residuals over time. When the braking system is in a healthy state, the braking system performance prediction model can accurately capture its operating rules, so the residuals between the model prediction value and the true value are small, and the change of the residuals remains stable within the allowable range. In this case, it can be considered that the operating state of the braking system is good and no special attention is required. However, when the braking system is abnormal or about to fail, its actual operating data will deviate from the established benchmark model, resulting in a significant increase in the residuals between the model prediction value and the true value, and there may be obvious trend changes. This residual change trend is the key basis for evaluating the health status of the braking system.

[0086] In this embodiment, a maintenance suggestion generation module may also be included, which is used to automatically generate maintenance suggestions based on early warning reminders when it is determined that the braking system is in an abnormal state, and provide them to the maintenance team.

[0087] The present invention aims to diagnose the health status of the brake system by comprehensively evaluating its overall performance, rather than limiting the analysis to a single or specific key component. This innovative solution is sensitive enough to instantly capture abnormalities in the brake system, ensuring that once the system performance deviates from normal standards, it can be quickly and accurately detected and identified, thereby providing users with timely and reliable maintenance recommendations to ensure driving safety.

[0088] The system proposed in this invention uses the powerful capabilities of deep learning networks. Compared with building a complex digital twin system, this design can significantly reduce the implementation cost while still achieving an accurate assessment of the health status of the brake system. Through the efficient data processing and analysis capabilities of the deep learning network, the system can more directly and economically capture and understand the performance changes of the brake system, providing users with cost-effective and reliable maintenance solutions.

[0089] All the above optional technical solutions can be arbitrarily combined to form optional embodiments of the present application, which will not be described one by one here.

[0090] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0091] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for health assessment of unmanned mining vehicle braking system based on deep learning, characterized in that: The following steps are involved: Step S1, the data extraction module connects to the data source, and automatically captures data closely related to the brake system performance evaluation from the data source as raw data, where the data closely related to the brake system performance evaluation includes brake pedal opening, vehicle speed, acceleration, and driving slope; Step S2, the data extraction module performs data screening and data verification on the original data to obtain pre-processed data; Step S3, the data processing module receives the preprocessed data, and performs data cleaning on the preprocessed data to obtain cleaned data; Step S4, the data processing module performs feature selection on the cleaned data to obtain highly relevant feature data, and further performs data conversion and normalization, as well as data verification, to obtain and store the verified data; Step S5, the model training module uses the obtained highly correlated feature data to construct and train a brake system performance prediction model through an LSTM algorithm; Step S6, the model training module optimizes the trained brake system performance prediction model to obtain an optimized brake system performance prediction model; Step S7, the health assessment module obtains a model prediction value based on the optimized brake system performance prediction model, and calculates the residual between the model prediction value and the real value collected in real time by the mine car sensor; Step S8, the health assessment module performs trend analysis on the residuals to obtain the residual variation trend over time as the trend analysis result, wherein the residual variation trend includes whether it is stable, the magnitude of the fluctuation, whether there is an obvious increasing trend and periodic fluctuation; Step S9, the health assessment module performs a health status assessment based on the residual value and trend analysis results. If the residual changes stably within a preset range, the braking system is determined to be in a healthy state. Otherwise, the braking system is determined to be in an abnormal state and an early warning reminder is issued.

2. The method for health assessment of unmanned mining vehicle braking system based on deep learning according to claim 1 is characterized in that: The step S2 specifically includes the following steps: Step S2.1, the data extraction module performs data screening, including using statistical analysis and threshold setting to identify abnormal fluctuation points in the original data; Step S2.2, using interpolation and mean substitution to correct and remove outliers and noise related to abnormal fluctuation points in the original data to obtain filtered data; Step S2.3, perform data verification on the screened data to obtain preprocessed data.

3. The method for health assessment of unmanned mining vehicle braking system based on deep learning according to claim 2 is characterized in that: The step S3 specifically comprises the following steps: Step S3.1, the data processing module receives the pre-processed key data and identifies key information therefrom; Step S3.2, using a moving average filtering algorithm to smooth and filter the identified key information to obtain cleaned data.

4. The method for health assessment of unmanned mining vehicle braking system based on deep learning according to claim 3 is characterized in that: The step S4 specifically comprises the following steps: Step S4.1, the data processing module calculates the Pearson correlation coefficient between the features of the cleaned data, identifies the features that are highly correlated with the brake system performance evaluation, and retains them as highly correlated feature data; Step S4.2, the data processing module performs data conversion and normalization processing on the highly correlated feature data to obtain converted data; Step S4.3, the data processing module verifies the converted data to see whether it meets the expected format and quality standards. If so, the converted data is stored as the verified data. Otherwise, it returns to step S1 to re-extract the original data.

5. The method for health assessment of unmanned mining vehicle braking system based on deep learning according to claim 4 is characterized in that: The step S5 specifically comprises the following steps: Step S5.1, constructing an LSTM model through an LSTM algorithm; Step S5.2, setting the initial parameters of the LSTM model, including setting the number of network layers of the LSTM model, the number of units in each network layer, and the loss function; Step S5.3, using the obtained highly correlated feature data as a training set, the constructed LSTM model is trained to obtain a trained braking system performance prediction model.

6. The method for health assessment of unmanned mining vehicle braking system based on deep learning according to claim 5 is characterized in that: The step S7 specifically includes the following steps: Step S7.1, the health assessment module receives the verified data from the data processing module as real-time data, and inputs the real-time data into the optimized brake system performance prediction model to obtain the model prediction value; Step S7.2, compare the model prediction value with the true value to obtain the residual.

Images