Intelligent fault diagnosis and safety protection method for combine harvester

Abstract

The invention provides an intelligent fault diagnosis and safety protection method for a combine harvester, and the method comprises the steps: carrying out the quantitative analysis of the influence degree of a fault on the system safety and operation continuity through a risk assessment model, calculating a fault danger grade score through combining historical fault data and equipment aging state information, and carrying out the calculation of the fault danger grade score. A five-level danger classification result from slight danger to serious danger is obtained; acquiring system response data after isolation control is executed, monitoring running state changes and load redistribution conditions of residual normal parts, adjusting control parameters through a real-time feedback mechanism, and judging the effectiveness of isolation measures and the maintenance degree of system stability; a closed-loop feedback mechanism is adopted to continuously monitor the fault isolation effect and the system operation performance, key data in the processing process are collected to update fault diagnosis model parameters, the risk level evaluation accuracy and the protection range division rationality are optimized, and a continuously improved intelligent fault diagnosis and safety protection system is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of combine harvesters, and in particular to an intelligent fault diagnosis and safety protection method for a combine harvester. Background Art

[0002] As the degree of agricultural mechanization continues to increase, the structure of combine harvesters is becoming increasingly complex, and the coordination and cooperation requirements between subsystems are becoming increasingly high. Failure of any component may affect the performance of the entire machine and even cause serious losses.

[0003] Current fault diagnosis methods mainly rely on traditional periodic maintenance and simple alarm prompts, and lack the ability to accurately judge the severity of the fault and handle it in a graded manner.

[0004] These methods can often only respond passively after a fault occurs, and are unable to adopt differentiated response strategies based on the fault type and hazard level, resulting in minor faults being over-handled and serious faults not being controlled in a timely and effective manner.

[0005] The core challenge of combine harvester fault diagnosis lies in the coordination and unification of fault detection and safety protection mechanisms.

[0006] After detecting a fault, traditional diagnostic systems often adopt a unified shutdown protection strategy. This simple and crude approach ignores the differences in the impact of different faults on system safety.

[0007] When the system cannot accurately assess the danger level of the fault, it is difficult to formulate appropriate protection measures, which in turn leads to technical difficulties in determining the scope of protection.

[0008] Over-protection can lead to unnecessary shutdown of the entire machine, affecting operational efficiency, while insufficient protection can cause faults to spread and lead to more serious equipment damage.

[0009] Improper determination of the protection range further exposes the technical bottleneck of fault isolation control, namely how to achieve precise control of dangerous components while ensuring safety, while maintaining the continued operation of other normal components and avoiding system-wide shutdowns caused by local faults.

[0010] Therefore, how to establish an intelligent diagnostic method that can accurately assess the severity of the fault, reasonably determine the protection range and achieve precise fault isolation, while ensuring equipment safety and maximizing the system operation efficiency, has become a key issue in the development of combine harvester fault diagnosis technology. Summary of the Invention

[0011] The present invention provides an intelligent fault diagnosis and safety protection method for a combine harvester, comprising the following steps: S1. Obtain real-time operating parameter data: Obtain real-time operating parameter data for each subsystem of the combine harvester, collect engine speed, hydraulic pressure, vibration frequency, temperature change, and current fluctuation signals through a preset sensor network, establish a multi-dimensional equipment monitoring data matrix, and obtain the original fault feature vector containing time series characteristics; S2. Determine the current fault category and preliminary severity level; based on the fault feature vector, use a multi-level fault diagnosis algorithm to perform pattern recognition and classification processing on the abnormal signal. If the detected parameter deviates from the normal operating range, calculate the deviation metric value and match it with a pre-established fault type library to determine the current fault category and preliminary severity level; S3. Obtaining a hazard classification result; for each fault category, using a risk assessment model to quantitatively analyze the impact of the fault on system safety and operational continuity, and combining historical fault data and equipment aging status information to calculate the fault hazard level score, obtaining a five-level hazard classification result from mild to severe; S4. Obtain a differentiated protection domain configuration plan. Based on the hazard level score, a dynamic protection domain partitioning algorithm is used to determine the range of devices affected by the fault and the protection boundary. If the hazard level is below a preset safety threshold, a local protection area is delineated. If it exceeds the threshold, the protection range is expanded to include associated subsystems, thereby obtaining a differentiated protection domain configuration plan. S5, precise control: through the protection domain configuration scheme, the adaptive isolation control algorithm is activated to implement precise control of the faulty component, and the control strategy is dynamically adjusted according to the fault location and impact range. If the fault is located in a non-critical path, soft isolation is implemented to maintain some functions. If it is located in a critical path, hard isolation is implemented to completely disconnect. S6. Determine the effectiveness of the isolation measures and the degree to which system stability is maintained; obtain system response data after the isolation control is executed, monitor the operating status changes and load redistribution of the remaining normal components, adjust control parameters through a real-time feedback mechanism, and determine the effectiveness of the isolation measures and the degree to which system stability is maintained; S7, dynamic optimization: Based on the system stability assessment results, dynamically optimize the operating mode and operating parameter configuration. If the system can still maintain basic operating capabilities, switch to a degraded operating mode to continue operations. If the system stability is severely impaired, trigger a safety shutdown procedure to determine the final fault handling plan and equipment status. S8. Improve the intelligent fault diagnosis and safety protection system; use a closed-loop feedback mechanism to continuously monitor the fault isolation effect and system operation performance, collect key data during the processing to update the fault diagnosis model parameters, optimize the accuracy of hazard level assessment and the rationality of protection range division, and obtain a continuously improved intelligent fault diagnosis and safety protection system.

[0012] The technical solution provided by the embodiment of the present invention may have the following beneficial effects: The present invention discloses a method for intelligent fault diagnosis and safety protection of a combine harvester. The method collects equipment operating parameters in real time through a multi-dimensional sensor network, constructs a fault feature vector, and adopts a multi-level diagnostic algorithm to identify and classify anomalies. Combined with a risk assessment model, the degree of fault impact is quantitatively analyzed and divided into five levels of danger. According to the danger level, the protection domain range is dynamically divided and adaptive isolation control is implemented. Through a closed-loop feedback mechanism, the diagnostic model and protection strategy are continuously optimized to achieve accurate fault identification, dynamic assessment of risk levels, intelligent division of protection ranges, and adaptive control of isolation measures, effectively ensuring the safe and stable operation of the combine harvester under complex working conditions and improving equipment reliability and operation continuity. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 The present invention is a flow chart of a combine harvester intelligent fault diagnosis and safety protection method. DETAILED DESCRIPTION

[0014] To further understand the content of the present invention, the present invention is described in detail with reference to the accompanying drawings and examples. The present application is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are intended only to illustrate the relevant invention and are not intended to limit the invention. It should also be noted that, for ease of description, only portions relevant to the invention are shown in the accompanying drawings.

[0015] like Figure 1 This embodiment provides a combine harvester intelligent fault diagnosis and safety protection method, which may specifically include the following steps: S1. Obtain real-time operating parameter data. Obtain real-time operating parameter data of each subsystem of the combine harvester. Collect engine speed, hydraulic pressure, vibration frequency, temperature change, and current fluctuation signals through a preset sensor network, establish a multi-dimensional equipment monitoring data matrix, and obtain the original fault feature vector containing time series characteristics.

[0016] Specifically, a sensor network is used to collect operational data from various combine harvester subsystems, encompassing multi-dimensional signals such as engine speed, hydraulic pressure, vibration frequency, temperature changes, and current fluctuations. This data is then used to construct an initial equipment monitoring dataset. Based on this initial equipment monitoring dataset, data preprocessing methods are used to clean and standardize the collected operational data to eliminate noise interference, resulting in a normalized data matrix. Time series-related features are extracted from this normalized data matrix, and the temporal trends of each signal are analyzed to determine a time series feature set. If the trend of a signal in the time series feature set exceeds a preset threshold, an anomaly detection mechanism is triggered to determine whether potential fault signatures exist. The fault signatures identified by the anomaly detection mechanism are then combined with a pre-established support vector machine model to classify the fault signatures and obtain a specific fault category identifier. Based on the fault category identifier, corresponding subsystem operational status assessment data is generated. Historical data comparison is then used to determine the current equipment operational health. This data is then combined with the time series features and fault category identifier to construct a dynamically updated feature vector for subsequent equipment monitoring and fault prediction.

[0017] Specifically, in the scenario of combining harvester operational data monitoring, the sensor network collects multi-dimensional signals from subsystems such as the engine, hydraulic system, and transmission to form an initial dataset. For example, engine speed is collected once a minute, ranging from 1500 to 2500 rpm; hydraulic pressure is recorded every second, with a normal value of 10-15 MPa; vibration frequency is acquired by an accelerometer, with a typical value of 5-50 Hz; temperature sensors monitor bearing temperature, with a normal range of 40-80°C; and current fluctuations are recorded by the electronic control unit, with a normal fluctuation of 5-10 A. This initial dataset includes timestamps and signal values, but may contain missing values ​​or noise. The applicant believes that in one possible implementation, data preprocessing first uses median filtering to eliminate spike noise in the vibration frequency. For example, the outlier value of 50 Hz is replaced by the average of the neighboring values, approximately 20 Hz. Missing values ​​are supplemented by linear interpolation. For example, if the hydraulic pressure at a certain moment is missing, 10.75 MPa can be interpolated from the previous and next values ​​of 10.5 MPa and 11 MPa. Normalization converts signals of different dimensions into matrices with a mean of 0 and a standard deviation of 1 to facilitate subsequent analysis. This preprocessing ensures data consistency and improves the accuracy of feature extraction. For the normalized data matrix, time series feature extraction focuses on the trend of signal changes. For example, by calculating the 5-minute mean and standard deviation of engine speed using a sliding window, we found that a slow decrease from 2000 rpm to 1800 rpm could indicate a fuel shortage. Short-term fluctuations in hydraulic pressure exceeding 0.5 MPa / s indicate a potential leak. Feature sets, including mean, variance, and trend slope, reflect the subsystem's operating status. This feature extraction helps capture dynamic changes and provides a basis for fault detection. The anomaly detection mechanism is based on preset thresholds. For another example, an anomaly alarm is triggered when the vibration frequency exceeds 40 Hz or the temperature exceeds 85°C. Preferably, statistical methods, such as Z-score detection, are used to flag a signal as an anomaly if it deviates from the mean by more than three standard deviations. This mechanism can quickly identify potential faults, reduce false alarms, and improve diagnostic efficiency. In one possible embodiment, the abnormal signal is input into a support vector machine model for fault classification. For example, a high vibration frequency combined with an abnormal temperature could be classified as bearing wear, while a sudden drop in hydraulic pressure could indicate a pump failure.The model is trained based on historical fault data, with a classification accuracy of over 90%. This classification method converts fault characteristics into specific categories, providing precise guidance for maintenance. It also generates operating status assessment data based on the fault category. For example, if bearing wear causes the health level to drop from 90% to 70%, the rate of health decline is analyzed by comparison with historical data. If it drops by 5% per week, it indicates that maintenance is needed in advance. This assessment provides a quantitative basis for equipment management and extends equipment life. The dynamically updated feature vector combines time series characteristics and fault categories. For example, the speed reduction trend, vibration abnormality and bearing wear indicators are integrated into feature vectors and input into the prediction model to determine the probability of failure within the next week. In addition, it should be noted that this dynamic update can reflect the equipment status in real time, optimize maintenance plans, and reduce downtime risks. This multi-dimensional, closed-loop monitoring system significantly improves the operational reliability and maintenance efficiency of combine harvesters.

[0018] S2. Determine the current fault category and preliminary severity level; based on the fault feature vector, use a multi-level fault diagnosis algorithm to perform pattern recognition and classification processing on the abnormal signal. If the detected parameters deviate from the normal operating range, calculate the deviation metric value and match it with a pre-established fault type library to determine the current fault category and preliminary severity level.

[0019] Abnormal signals are collected by sensors, feature vectors are extracted, and the principal component analysis algorithm is used to reduce the dimension of the feature vectors to obtain a low-dimensional feature set. If the parameters of the low-dimensional feature set deviate from the normal range, the Euclidean distance is calculated as the deviation measure. The pre-established fault library is matched using the k-nearest neighbor algorithm to determine the fault category. The preset severity table is queried based on the fault category to determine the preliminary severity. Historical fault data is obtained, and the support vector machine algorithm is used to optimize the fault library and update the matching accuracy.

[0020] For example, to collect abnormal signals from sensors on a combine harvester, multiple sensor types, such as speed sensors, pressure sensors, and vibration sensors, can be deployed on the engine, hydraulic system, and transmission components to acquire real-time operating data. Suppose, for example, that in the engine subsystem, sensors detect an abnormal speed fluctuation. The data stream shows that the rpm jumps from the normal value of 1800 to 2200 rpm, exceeding the expected range. This abnormal signal can be used as an initial feature, and its occurrence frequency and duration are recorded using timestamps, providing basic data for subsequent feature extraction. It is important to note that sensors require regular calibration to ensure data accuracy, and the acquisition frequency can be set to 10 times per second to capture transient changes.The applicant artificially proposes that in a possible implementation of this embodiment, when the principal component analysis algorithm is used to reduce the dimension of the feature vector, high-dimensional data such as speed, pressure, vibration frequency and other signals can be compressed into a low-dimensional feature set. For example, the original data contains 10 dimensions. The first 3 principal components are retained through principal component analysis, explaining 90% of the variance to generate a low-dimensional feature set. This dimensionality reduction can reduce the computational complexity while retaining key information. For example, the pressure data and vibration frequency of the hydraulic system may be highly correlated. After dimensionality reduction, the features of the common influence of the two can be highlighted, which is convenient for subsequent analysis. For example, when judging whether the parameters of the low-dimensional feature set deviate from the normal range, the degree of abnormality can be quantified by calculating the Euclidean distance. Assume that the baseline value of the low-dimensional feature set under normal operation is [2.5, 1.8, 0.9], while it is [3.2, 2.5, 1.3], the Euclidean distance between the two is calculated to be 0.98, which exceeds the preset threshold of 0.5, indicating that there is a potential fault. The applicant's method intuitively reflects the degree of data deviation and is convenient for quickly locating the problem. In one possible implementation, when the k-nearest neighbor algorithm matches the fault library, a database containing hundreds of fault modes can be established based on historical fault data. For example, assuming that fault modes such as "insufficient hydraulic pump pressure" and "excessive bearing vibration" are stored in the database, the k-nearest neighbor algorithm (k=5) is used to compare the similarity between the current feature vector and the historical sample to determine the most likely fault category, such as "hydraulic system leakage". This matching method can quickly locate the fault type and improve diagnostic efficiency. The applicant needs to explain that when querying the preset severity table, the impact range can be determined according to the fault category; for example, a hydraulic system leakage may be rated as "moderately serious" because it may cause a decrease in efficiency but not an immediate shutdown; the severity table can be based on According to historical experience, the faults of mild, moderate and severe faults correspond to maintenance times of 1 hour, 4 hours and more than 8 hours respectively, which makes it easier to prioritize the handling of critical faults. For example, when optimizing the fault library, the support vector machine algorithm can analyze historical fault data and adjust the classification boundaries to improve matching accuracy. For example, for the "bearing wear" fault that occurs repeatedly, the algorithm can optimize the model by adding new samples to reduce the misjudgment rate from 15% to 5%. This dynamic update can continuously improve the fault library and ensure more reliable diagnostic results. In one possible implementation method, combining the above steps, a complete fault diagnosis process can be constructed. For example, when the sensor detects an abnormal vibration frequency and the features deviate from the normal value after dimensionality reduction, the "loose drive shaft" fault is matched through k-nearest neighbor matching, and it is determined to be "severe" after querying the severity table. The fault library is updated through the support vector machine. This process forms a closed loop from data collection to fault classification, significantly improving the operation monitoring capability of the combine harvester.

[0021] S3. Obtain the hazard classification results. For the fault categories, quantitatively analyze the impact of the fault on system safety and operation continuity through the risk assessment model. Combined with historical fault data and equipment aging status information, calculate the fault hazard level score and obtain a five-level hazard classification result from mild to severe.

[0022] Obtain fault category data, extract fault features from historical fault records and equipment aging status information, generate a set of fault categories, use a support vector machine algorithm to classify the set of fault categories, and combine it with historical fault data to obtain a preliminary risk score for the fault category. Using a quantitative analysis model, based on the preliminary risk score and status information, calculate the weight of the fault's impact on system safety and operational continuity to obtain a comprehensive impact score. If the comprehensive impact score exceeds a preset threshold, the fault is classified as high-risk. If it is lower than the threshold, continue to analyze the historical fault frequency, determine the low-risk or medium-risk category, and map it to a five-level classification system based on the comprehensive impact score and fault category to generate a hazard level classification result. Extract high-risk fault features from the hazard level classification result, combine it with the equipment aging status, generate a priority ranking, determine the maintenance scheduling order, update the fault handling strategy database through priority ranking, and generate an optimized fault management plan. For example, in the field of combine harvester fault diagnosis, obtaining fault category data is a key link. Data on common problems such as engine overheating and hydraulic pipeline blockage can be extracted from historical fault records. At the same time, combined with equipment aging status information, such as bearing service life or oil deterioration degree, a set of multiple fault modes is generated for the classification process of the support vector machine algorithm.Specifically, the fault characteristics recorded in the historical data, such as abnormal vibration frequency or pressure drop amplitude, can be used as input to preliminarily divide different fault categories and assign a risk score to each type of fault. For example, the hydraulic fault score may be 7.5, while the transmission component fault score is 6.0. In the application of the quantitative analysis model, based on the preliminary risk score and equipment status information, the impact weight of the fault on system safety and operation continuity can be analyzed. For example, the preliminary score of the engine overheating fault is 8.0. Combined with the aging status of the equipment, it shows that the cooling system has been used for more than the expected life. The calculated comprehensive impact score is 9.2. If the preset threshold is 8.5, the fault is classified as a high-risk category; for faults with scores below the threshold, such as slight wear of a transmission component with a score of 6.8, the historical fault frequency is further analyzed. If the frequency is low, it can be classified as a low-risk category; for example, for the mapping of the five-level classification system, the comprehensive impact score can be combined with the fault category to generate a hazard level classification result. For example, if the score of hydraulic pipeline blockage is 8.8 , classified as high-risk, corresponding to the fourth level of the five-level classification, while minor wear corresponds to the second level. This grading method helps to clarify the urgency of the fault. Then, we extract key information from the characteristics of high-risk faults. For example, the specific manifestation of engine overheating is the temperature continuously exceeding 90 degrees Celsius. Combined with the aging status of the equipment, we generate a priority ranking and determine the maintenance scheduling order, such as prioritizing engine problems before handling hydraulic faults. For example, when updating the fault handling strategy database, the fault management plan can be optimized based on the priority ranking. Assume that the original strategy in the database is to handle faults in the order of occurrence time. After the update, the faults in the high-risk category are prioritized, such as resolving temperature anomalies first and then handling other low-risk faults. This adjustment can effectively improve the stability of equipment operation. Therefore, through the above method, each link from data extraction to strategy optimization is closely connected, ensuring that the fault management plan is more scientific and reasonable, while also ensuring the continuity and safety of the combine harvester during operation.

[0023] S4. Obtain a differentiated protection domain configuration plan. Based on the hazard level score, use a dynamic protection domain division algorithm to determine the equipment range and protection boundary affected by the fault. If the hazard level is lower than the preset safety threshold, a local protection area is delineated. If it exceeds the threshold, the protection range is expanded to the associated subsystems to obtain a differentiated protection domain configuration plan.

[0024] Using sensor data and historical fault records, a hazard level score is calculated to determine the likelihood and impact of a fault. Based on the hazard level score, a K-means clustering algorithm is used to divide dynamic protection domains, determine the device range affected by the fault, and determine the initial protection boundary. If the hazard level score is below a preset safety threshold, a local protection zone configuration is generated based on the device range. If it is above the threshold, topology analysis is used to determine the device list of associated subsystems and expand the protection range. A depth-first search algorithm is used to calculate the protection boundary for the expanded protection range, generating differentiated protection domain configurations. Protection rules are extracted from the differentiated protection domain configurations to generate device isolation policies and access control lists. Network traffic monitoring is used to obtain real-time device status and determine whether the protection domain configuration meets fault isolation requirements. If the real-time device status indicates an anomaly, the protection domain configuration is adjusted based on the anomaly type, and the device isolation policy is updated to achieve an optimized configuration.

[0025] For example, in the field of combine harvester fault management, a hazard level score is calculated using sensor data and historical fault records to identify the likelihood of a fault and its impact on the equipment. Sensor data, which can include vibration frequency, oil pressure, or temperature readings, is combined with historical fault records, such as the frequency of blade wear or electrical short circuits over the past year, to generate a comprehensive score. Suppose a sensor detects an abnormal blade vibration frequency, and historical records show that this fault occurs monthly. The score is assigned to 7.2, which is below the safety threshold of 8.0, indicating that the risk is manageable. For example, in one possible implementation, a K-means clustering algorithm is used to divide dynamic protection domains to determine the equipment affected by the fault and the initial protection boundaries. The clustering algorithm analyzes features in the sensor data, such as vibration amplitude and duration, to divide the fault into different impact areas. For example, abnormal blade vibration may only affect the cutting system. The clustering results generate a local protection domain covering the blade and its drive motor, with the initial protection boundary limited to the cutting module. As can be seen, if the hazard level score is below the threshold, such as 7.2, the system generates a configuration plan for the local protection zone. Specifically, for blade vibration, the configuration plan may include instructions to reduce the motor speed or pause the blade operation. The protection area only involves the cutting system to avoid affecting the operation of the entire machine. If the score is above the threshold, such as a hydraulic system fault score of 8.5, topology analysis is used to identify associated subsystems, such as the hydraulic pump and pipelines, and the protection scope is expanded to the entire hydraulic module. For example, a depth-first search algorithm is used to calculate the protection boundary and generate a differentiated protection domain configuration. For hydraulic system faults, the algorithm analyzes the device topology, identifies the path from the hydraulic pump to the pipeline, and generates a protection boundary covering all related components. Differentiated configurations may include limiting the hydraulic pump's output power or switching to a backup pipeline. Protection rules are extracted from this to form a device isolation strategy, such as restricting communication between the hydraulic module and other systems and generating an access control list to only allow critical maintenance signals to pass. In one possible implementation, real-time device status is obtained through network traffic monitoring to determine whether the protection domain configuration meets fault isolation requirements. For example, if real-time monitoring shows a continuous drop in hydraulic system pressure, indicating that the protection domain configuration needs to be adjusted, the system dynamically updates the isolation strategy, such as further restricting the hydraulic module's external connections and optimizing the configuration to ensure that the fault does not spread to other subsystems. This approach improves the accuracy of fault management and the stability of equipment operation through real-time feedback and dynamic adjustment.

[0026] S5. Precise control: Through the protection domain configuration scheme, the adaptive isolation control algorithm is started to implement precise control of the faulty component, and the control strategy is dynamically adjusted according to the fault location and impact range. If the fault is located in a non-critical path, soft isolation is implemented to maintain some functions. If it is located in a critical path, hard isolation is implemented to completely disconnect.

[0027] By analyzing the protection domain configuration, the operating status data of the faulty component is obtained from the system to determine the fault location and impact range. If the fault location is on a non-critical path, a soft isolation strategy is adopted. By adjusting the operating parameters of the functional modules, some functions are retained to obtain a stable system state. Based on the retained functions, the current performance indicators of the system are obtained to determine whether they meet the preset threshold requirements. If the performance indicators meet the threshold requirements, the isolation strategy is dynamically adjusted through the adaptive isolation control algorithm to determine the optimized operating configuration. The operating data of the critical path is extracted from the optimized operating configuration to determine whether there is a potential fault risk. If there is a fault risk on the critical path, a hard isolation operation is performed to obtain the isolated system status by disconnecting the relevant connections. Based on the isolated system status, the latest system log data is obtained. By analyzing the logs, the stability and isolation effect of the system are determined.

[0028] S6. Determine the effectiveness of the isolation measures and the degree to which the system stability is maintained. Obtain system response data after the isolation control is executed. Monitor the operating status changes and load redistribution of the remaining normal components. Adjust the control parameters through the real-time feedback mechanism to determine the effectiveness of the isolation measures and the degree to which the system stability is maintained.

[0029] By extracting the response data after isolation control from the system log and analyzing the changes in the operating status of each component, preliminary status change trend data is obtained. Based on the status change trend data, a preset load distribution model is used to calculate the load redistribution ratio of the remaining normal components and determine the distribution status after load adjustment. The distribution status data after load adjustment is obtained, and combined with the real-time feedback mechanism, the operating parameter fluctuations of each component are monitored to determine whether the parameters exceed the preset threshold range. If the parameters exceed the preset threshold range, the control parameters are adjusted to optimize the system operation configuration and obtain the adjusted parameter optimization results. Based on the parameter optimization results, the system response data and stability maintenance status are analyzed to evaluate the impact of the isolation measures on the overall system and determine the evaluation data of the measures. Based on the evaluation data of the measures' effectiveness, combined with real-time feedback on the operating status and load distribution, the system configuration strategy is dynamically adjusted to obtain the final stability maintenance plan.

[0030] For example, in the scenario of system log analysis, extracting response data after isolation control can provide a preliminary understanding of changes in the operating status of various components. For example, suppose that in a distributed control system, after isolation, the response time of a non-critical component increases from 200 milliseconds to 500 milliseconds, while the response times of other components remain unaffected. This state change trend data provides a basis for subsequent load distribution. Analysis shows that the isolation measure reallocated some resources, affecting the efficiency of specific components. For example, in the application of load distribution models, the load proportion of remaining normal components can be calculated based on state change trend data. In one possible implementation, let's assume that there are five key components in the system. After one is isolated, the remaining four components must share its load. The original load proportion per component is 20%, which is adjusted to 25% per component after isolation. This redistribution ensures the continuity of overall system operation while avoiding the risk of overloading a single component. For example, when monitoring operating parameter fluctuations, combined with real-time feedback mechanisms, it can be discovered that after load adjustments, the temperature or resource utilization of certain components exceeds a preset threshold. For example, if the threshold is 80%, and after adjustments, the resource utilization of a particular component reaches 85%, timely adjustment of control parameters is necessary. By lowering the task priority of that component or offloading some tasks to other components, the utilization rate can be reduced to 75%, thereby maintaining system stability. For example, after optimizing the system operating configuration, analyzing the relationship between parameter optimization results and system response data can assess the impact of isolation measures. Suppose that after optimization, the overall system response time drops from 500 milliseconds to 300 milliseconds, indicating that isolation measures ensure stability without significantly sacrificing efficiency. This evaluation data provides important reference for subsequent policy adjustments. For example, when dynamically adjusting system configuration policies, combining real-time feedback on operating status and load distribution can help formulate a final strategy for maintaining stability. Suppose that after multiple adjustments, the system failure rate in high-load scenarios drops from 5% to 1%, while resource utilization increases by 10%. This approach ensures system adaptability in complex environments and extends the equipment's operating life. For example, when assessing the impact of isolation measures on the overall system, we can analyze their effects from multiple perspectives. Let's assume that after isolating non-critical paths, the system can still maintain 80% functionality, while isolating critical paths requires complete disconnection to avoid cascading failures. By comparing the response data and stability indicators of the two isolation methods, we can gain a more comprehensive understanding of the applicable scenarios of the measures. This multi-faceted analysis helps develop more flexible response strategies and enhance the system's risk resilience.

[0031] S7, dynamic optimization; based on the system stability assessment results, dynamically optimize the operating mode and operating parameter configuration. If the system can still maintain basic operating capabilities, switch to a degraded operating mode to continue operating. If the system stability is seriously damaged, trigger the safety shutdown procedure to determine the final fault handling plan and equipment status.

[0032] Obtain system operation data, collect various indicator data related to system stability through the real-time monitoring module, use the preset threshold range for preliminary comparison, judge whether the system is in the normal operating range, and obtain the stability assessment result. For the stability assessment result, use the logic judgment module to analyze the specific category of the assessment result. If the assessment result shows that the system is in the normal range, maintain the current operation mode and operating parameters, and determine the instructions for continuing the operation. If the assessment result shows that the system is close to an unstable state, dynamically adjust the operating parameters through the parameter optimization tool, obtain the adjusted parameter configuration, and judge whether the system has recovered to the basic capacity range. If the system has recovered to the basic capacity range after adjustment, switch to the reduced capacity mode. In the downgraded operation mode, the mode switching module loads the downgraded operation strategy and determines the operation instructions under the downgraded operation. If the system still fails to restore basic capabilities after adjustment and the assessment results show serious damage, the safety shutdown program is triggered through the safety protection module to obtain the shutdown execution status and determine whether the equipment has entered a safe state. For the equipment status after the safe shutdown, the fault diagnosis tool is used to analyze the system log and shutdown data, obtain the fault cause classification, and determine the priority ranking of the fault handling plan; according to the priority ranking of the fault handling plan, the pre-established fault recovery database is used to match the corresponding repair strategy, obtain the repair process instructions, and then determine whether the equipment status can be restored to the normal operating range.

[0033] S8. Improve the intelligent fault diagnosis and safety protection system; use a closed-loop feedback mechanism to continuously monitor the fault isolation effect and system operation performance, collect key data during the processing to update the fault diagnosis model parameters, optimize the accuracy of hazard level assessment and the rationality of protection range division, and obtain a continuously improved intelligent fault diagnosis and safety protection system.

[0034] Real-time operation data is obtained from the system operation status through a closed-loop feedback mechanism to determine the abnormal points of the system operation status. If the abnormal points exceed the preset threshold, key data are extracted from the abnormal points to obtain a key data set. The key data set is classified using a support vector machine algorithm to determine the fault type and severity. The fault diagnosis model parameters are updated according to the fault type and severity to obtain an optimized fault diagnosis model. The hazard level is evaluated using the optimized fault diagnosis model to determine the hazard level distribution. The cluster analysis algorithm is used to divide the protection range of the hazard level distribution to obtain a protection range set. The configuration of the safety protection system is adjusted according to the protection range set to obtain a dynamically updated safety protection strategy.

[0035] For example, the application of a closed-loop feedback mechanism in system operational status monitoring can be achieved through real-time sensor data collection. Sensors collect operational data such as temperature, pressure, and vibration every second, forming a data stream. Suppose, during operation of an industrial control system, a temperature sensor detects a value exceeding 85 degrees Celsius, exceeding the normal range of 80 degrees Celsius. The closed-loop feedback mechanism immediately records this anomaly and transmits the data to the analysis module. The core of this mechanism lies in continuous monitoring and feedback, ensuring that anomalies are quickly detected. It is important to note that closed-loop feedback is not limited to a single metric but also includes comprehensive analysis of multidimensional data, such as the combined evaluation of vibration frequency and pressure values, to improve the accuracy of anomaly detection. For example, when extracting key data sets from anomalies, data filtering tools can be used to isolate temperature, pressure, and timestamps at the time of the anomaly to form a key data set. For example, let's assume that the system experiences three temperature anomalies within 10 minutes: 87 degrees Celsius, 89 degrees Celsius, and 86 degrees Celsius. Combined with the vibration frequency rising to 120 Hz, exceeding the normal range of 100 Hz, these data are extracted to form a multidimensional dataset containing time, temperature, and vibration. The purpose of key data sets is to provide precise input for subsequent classification, avoiding interference from irrelevant data. For example, when using a support vector machine algorithm to classify fault type and severity, the key data set can be input into the model, which then makes its judgment based on a pre-trained fault signature library. For example, a temperature anomaly combined with an increased vibration frequency might be classified as a "mechanical overload" fault with a severity of "intermediate." The support vector machine constructs a hyperplane to classify the data into three categories: normal, slightly abnormal, and severely abnormal. It should be noted that the classification results rely on the accuracy of feature selection, such as the correlation analysis between temperature and vibration, to ensure that the fault type judgment is closer to reality. For example, when updating the fault diagnosis model parameters, the model weights can be adjusted according to the classification results. We assume that if mechanical overload is frequently detected, the model will increase the weight of the temperature and vibration correlation, optimizing diagnostic accuracy. The optimized model can identify similar faults more quickly. For example, the next time 88 degrees Celsius and 115 Hz vibration are detected, the conclusion of "intermediate mechanical overload" will be directly output. This dynamic update ensures that the model adapts to changes in system operation. For example, the hazard level assessment is performed using the optimized model to generate a hazard level distribution. We assume that the assessment results show that 30% of the anomalies are high-risk, 50% are medium-risk, and 20% are low-risk. The hazard level distribution provides a basis for subsequent protection strategies. Here, the applicant needs to explain that the assessment is not only based on current data, but also combines historical fault trends to ensure comprehensiveness. For example, when cluster analysis is used to categorize protection zones, the K-means algorithm can be used to classify risk levels into three protection zones: high, medium, and low. For example, if the high-risk zone includes temperatures exceeding 90 degrees Celsius and vibrations exceeding 130 Hz, cluster analysis will classify such data as requiring immediate intervention. This categorization ensures highly targeted security policies.For example, when adjusting the configuration of the safety protection system, the system can automatically reduce the operating power in high-risk scenarios based on the protection range set, such as reducing the motor speed from 1000 rpm to 800 rpm, and generate a dynamically updated safety protection strategy. This strategy can effectively extend the life of the equipment and reduce the risk of failure. In addition, the applicant needs to explain that dynamic adjustment can also be combined with predictive maintenance to provide early warning of potential failures and further improve system reliability.

[0036] The above describes specific embodiments of the present invention. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art may make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. The embodiments of this application and the features in the embodiments may be combined with each other in any manner unless there is a conflict.

Claims

1. A combine harvester intelligent fault diagnosis and safety protection method, characterized in that: The method comprises: S1. Obtain real-time operating parameter data: Obtain real-time operating parameter data for each subsystem of the combine harvester, collect engine speed, hydraulic pressure, vibration frequency, temperature change, and current fluctuation signals through a preset sensor network, establish a multi-dimensional equipment monitoring data matrix, and obtain the original fault feature vector containing time series characteristics; S2. Determine the current fault category and preliminary severity level; based on the fault feature vector, use a multi-level fault diagnosis algorithm to perform pattern recognition and classification processing on the abnormal signal. If the detected parameter deviates from the normal operating range, calculate the deviation metric value and match it with a pre-established fault type library to determine the current fault category and preliminary severity level; S3. Obtaining a hazard classification result; for each fault category, using a risk assessment model to quantitatively analyze the impact of the fault on system safety and operational continuity, and combining historical fault data and equipment aging status information to calculate the fault hazard level score, obtaining a five-level hazard classification result from mild to severe; S4. Obtain a differentiated protection domain configuration plan. Based on the hazard level score, a dynamic protection domain partitioning algorithm is used to determine the range of devices affected by the fault and the protection boundary. If the hazard level is below a preset safety threshold, a local protection area is delineated. If it exceeds the threshold, the protection range is expanded to include associated subsystems, thereby obtaining a differentiated protection domain configuration plan. S5. Precise control: Using the protection domain configuration scheme, an adaptive isolation control algorithm is activated to precisely control the faulty component. The control strategy is dynamically adjusted based on the fault location and impact range. If the fault is located in a non-critical path, soft isolation is implemented to maintain some functions. If the fault is located in a critical path, hard isolation is implemented to completely disconnect the component. S6. Determine the effectiveness of the isolation measures and the degree to which the system stability is maintained; obtain system response data after the isolation control is executed, monitor the changes in the operating status of the remaining normal components and the load redistribution, adjust the control parameters through the real-time feedback mechanism, and determine the effectiveness of the isolation measures and the degree to which the system stability is maintained.

2. The intelligent fault diagnosis and safety protection method for a combine harvester according to claim 1, characterized in that: The following steps are also included: S7, dynamic optimization: Based on the system stability assessment results, dynamically optimize the operating mode and operating parameter configuration. If the system can still maintain basic operating capabilities, switch to a degraded operating mode to continue operations. If the system stability is severely impaired, trigger a safety shutdown procedure to determine the final fault handling plan and equipment status. S8. Improve intelligent fault diagnosis and safety protection system; A closed-loop feedback mechanism is used to continuously monitor the fault isolation effect and system operation performance, collect key data in the processing process to update the fault diagnosis model parameters, optimize the accuracy of hazard level assessment and the rationality of protection range division, and obtain a continuously improved intelligent fault diagnosis and safety protection system.

3. The intelligent fault diagnosis and safety protection method for a combine harvester according to claim 1, characterized in that: The S1 step includes: The sensor network acquires operational data from various subsystems of the combine harvester, including engine speed, hydraulic pressure, vibration frequency, temperature changes, and current fluctuations, to construct an initial equipment monitoring data set. Based on the initial equipment monitoring data set, the collected operating data is cleaned and standardized using data preprocessing methods to eliminate noise interference and obtain a standardized data matrix; For the normalized data matrix, extract the time series related features, analyze the changing trend of each signal in the time dimension, and determine the time series feature set; If the change trend of a signal in the time series feature set exceeds the preset threshold range, the anomaly detection mechanism is triggered to determine whether there are potential fault characteristics; The fault features identified by the anomaly detection mechanism are combined with the pre-established support vector machine model to classify the fault features and obtain a specific fault category identification; Generate corresponding subsystem operation status assessment data based on fault category identification, and use historical data comparison method to determine the health of the current equipment operation status; Obtain the health data of the current equipment operating status, combine it with time series characteristics and fault category identification, and construct a dynamically updated feature vector for subsequent equipment monitoring and fault prediction.

4. The intelligent fault diagnosis and safety protection method for a combine harvester according to claim 1, characterized in that: The S2 step includes: Collect abnormal signals through sensors and extract feature vectors; The principal component analysis algorithm is used to reduce the dimension of the feature vector to obtain a low-dimensional feature set; If the low-dimensional feature set parameters deviate from the normal interval, the Euclidean distance is calculated as the deviation measure; Use the k-nearest neighbor algorithm to match the pre-established fault library and determine the fault category; Query the preset severity table according to the fault type to determine the preliminary severity; Obtain historical fault data, use support vector machine algorithm to optimize the fault library, and update the matching accuracy.

5. The intelligent fault diagnosis and safety protection method for a combine harvester according to claim 1, characterized in that: The S3 step includes: Obtain fault category data, extract fault features from historical fault records and equipment aging status information, and generate a fault category set; The support vector machine algorithm is used to classify the fault category set and, combined with historical fault data, a preliminary risk score for the fault category is obtained; Through the quantitative analysis model, based on the preliminary risk score and status information, the impact weight of the fault on system safety and operation continuity is calculated to obtain a comprehensive impact score; If the comprehensive impact score exceeds the preset threshold, the fault is classified as high-risk; If it is below the threshold, continue to analyze the historical fault frequency to determine the low-risk or medium-risk category; Based on the comprehensive impact score and fault category, they are mapped to a five-level classification system to generate a hazard level classification result; Extract high-risk fault characteristics from the hazard level classification results, combine them with the equipment aging status, generate priority rankings, and determine the maintenance scheduling sequence; By prioritizing, the fault handling strategy database is updated to generate an optimized fault management plan.

6. The intelligent fault diagnosis and safety protection method for a combine harvester according to claim 1, characterized in that: The S4 step includes: Calculate the hazard level score based on sensor data and historical fault records to determine the probability and impact of a fault; Based on the hazard level score, the K-means clustering algorithm is used to divide the dynamic protection domain to obtain the range of equipment affected by the fault and the initial protection boundary; If the hazard level score is lower than the preset safety threshold, a configuration plan for a local protection area is generated based on the equipment range; If it is higher than the threshold, the device list of the associated subsystem is determined through topology analysis, and the protection scope is expanded; For the expanded protection range, a depth-first search algorithm is used to calculate the protection boundary and generate differentiated protection domain configurations; Extract protection rules from differentiated protection domain configurations to generate device isolation policies and access control lists; Obtain real-time device status through network traffic monitoring to determine whether the protection domain configuration meets fault isolation requirements; If the real-time device status shows an abnormality, the protection domain configuration is adjusted according to the abnormality type, and the device isolation policy is updated to obtain an optimized configuration plan.

7. The intelligent fault diagnosis and safety protection method for a combine harvester according to claim 1 is characterized in that The S5 step includes: By analyzing the protection domain configuration, the system obtains the operating status data of the faulty component and determines the fault location and impact range. If the fault location is on a non-critical path, a soft isolation strategy is used to adjust the operating parameters of the functional modules to retain some functions and achieve a stable system state. Based on the retained functions, obtain the current performance indicators of the system and determine whether they meet the preset threshold requirements; If the performance indicators meet the threshold requirements, the isolation strategy is dynamically adjusted through the adaptive isolation control algorithm to determine the optimized operation configuration; Extract critical path operation data from the optimized operation configuration to determine whether there is a potential failure risk; If there is a risk of failure on the critical path, a hard isolation operation is performed to obtain the isolated system state by disconnecting the relevant connections; Based on the isolated system status, obtain the latest system log data and determine the system stability and isolation effect by analyzing the logs.

8. The intelligent fault diagnosis and safety protection method for a combine harvester according to claim 1, characterized in that: The S6 step includes: By extracting the response data after isolation control from the system log, analyzing the operating status changes of each component, and obtaining preliminary status change trend data; Based on the status change trend data, the preset load distribution model is used to calculate the load redistribution ratio of the remaining normal components and determine the distribution status after load adjustment; Obtain the distribution status data after load adjustment, and combine it with the real-time feedback mechanism to monitor the fluctuation of the operating parameters of each component and determine whether the parameters exceed the preset threshold range; If the parameter exceeds the preset threshold range, the system operation configuration is optimized by adjusting the control parameter to obtain the adjusted parameter optimization result; Based on the parameter optimization results, analyze the system response data and stability maintenance status, evaluate the impact of isolation measures on the overall system, and determine the evaluation data of the measures; Based on the evaluation data of the measures’ effects and combined with real-time feedback on operating status and load distribution, the system configuration strategy is dynamically adjusted to obtain the final stability maintenance plan.

9. The intelligent fault diagnosis and safety protection method for a combine harvester according to claim 2, characterized in that: The S7 step includes: Obtain system operation data, collect various indicator data related to system stability through the real-time monitoring module, perform preliminary comparison using the preset threshold range, determine whether the system is in the normal operating range, and obtain stability assessment results; Based on the stability assessment results, a logic judgment module is used to analyze the specific category of the assessment results. If the assessment results show that the system is within the normal range, the current operating mode and operating parameters are maintained and instructions for continuing operation are determined; If the assessment results show that the system is close to an unstable state, the operating parameters are dynamically adjusted through the parameter optimization tool, and the adjusted parameter configuration is obtained to determine whether the system has recovered to the basic capacity range; If the system recovers to the basic capacity range after adjustment, it switches to the degraded operation mode, loads the degraded operation strategy through the mode switching module, and determines the operation instructions under the degraded operation; If the system still fails to recover its basic capabilities after adjustments and the assessment results indicate serious damage, the safety protection module triggers the safety shutdown procedure, obtains the shutdown execution status, and determines whether the device has entered a safe state; Based on the equipment status after safe shutdown, use fault diagnosis tools to analyze system logs and shutdown data, obtain fault cause classification, and determine the priority of fault handling solutions; According to the priority of the fault handling plan, the pre-established fault recovery database is used to match the corresponding repair strategy, obtain the repair process instructions, and determine whether the equipment status can be restored to the normal operating range.

10. The intelligent fault diagnosis and safety protection method for a combine harvester according to claim 2, characterized in that: The S8 step includes: Obtain real-time operating data from the system operating status through a closed-loop feedback mechanism to identify abnormal points in the system operating status; If the outlier exceeds the preset threshold, key data is extracted from the outlier to obtain a key data set; Use support vector machine algorithms to classify key data sets and determine fault types and severity; Update the fault diagnosis model parameters according to the fault type and severity to obtain an optimized fault diagnosis model; Evaluate the hazard level through the optimized fault diagnosis model and determine the hazard level distribution; Cluster analysis algorithm is used to divide the protection range of the hazard level distribution and obtain the protection range set; Adjust the configuration of the security protection system according to the protection range set to obtain dynamically updated security protection strategies.

Images