A high-pressure oil pump fault early warning method, system and device

Abstract

The application discloses a high-pressure oil pump fault early warning method, system and device, and relates to the technical field of ship diesel engine fault diagnosis. The early warning method comprises the following steps: in response to the high-pressure oil pump being in a steady pressure working condition, sample data of the high-pressure oil pump are acquired; the sample data comprise a front end temperature and rail pressure of the high-pressure oil pump; the sample data are input into a pre-trained early warning model to acquire detection parameters; and in response to the detection parameters being greater than a first threshold value, early warning information is generated. In this way, by monitoring the front end temperature and rail pressure of the high-pressure oil pump and inputting the data into the pre-trained early warning model, online fault early warning of the high-pressure oil pump can be realized, potential fault conditions can be discovered early, fault further deterioration can be avoided, downtime and maintenance costs can be reduced, and the reliability and stability of the diesel engine can be improved. Furthermore, by generating the detection parameters and the first threshold value, when the detection parameters exceed the first threshold value, the early warning information is generated, thereby improving the accuracy of fault detection and reducing false positives and false negatives.

Description

Technical Field

[0001] This application relates to the field of marine diesel engine fault diagnosis technology, specifically to a high-pressure oil pump fault early warning method, system and device. Background Technology

[0002] The high-pressure common rail system, as the fuel system for marine diesel engines, offers advantages such as high injection pressure, consistent injection patterns, precise injection timing, and flexible adjustment of key marine characteristics, including cyclic injection volume. A typical high-pressure common rail system includes a high-pressure fuel pump, common rail pipe, high-pressure fuel lines, injectors, an electronic control unit (ECU), various sensors, and actuators. The high-pressure fuel pump, as the core component of the system, compresses metered low-pressure fuel to establish high pressure, which is then delivered to other components of the system via the common rail pipe.

[0003] However, current diesel engines cannot provide online fault warnings for high-pressure fuel pumps. Once the high-pressure fuel pump fails, the high-pressure common rail system will be unable to establish high pressure normally, which seriously affects the reliability of diesel engine operation. Summary of the Invention

[0004] This application provides a high-pressure oil pump fault early warning method. By inputting the front-end temperature and rail pressure of the high-pressure oil pump into the early warning model, online fault early warning of the high-pressure oil pump can be achieved, solving the problem that the prior art cannot perform online fault early warning of high-pressure oil pumps. This application also provides a high-pressure oil pump fault early warning system and a high-pressure oil pump fault early warning device.

[0005] This application provides a high-pressure oil pump fault early warning method, including:

[0006] In response to the high-pressure oil pump being in a pressure-stabilizing condition, sample data of the high-pressure oil pump is acquired; the sample data includes the front-end temperature and rail pressure of the high-pressure oil pump.

[0007] The sample data is input into a pre-trained early warning model to obtain detection parameters;

[0008] A warning message is generated in response to the detection parameter being greater than a first threshold.

[0009] In some embodiments, the early warning model is pre-trained based on the following steps:

[0010] Acquire multiple historical data; the historical data includes the historical front-end temperature and historical rail pressure generated by the high-pressure oil pump during historical pressure stabilization operation.

[0011] The multiple historical data sets are divided into training datasets and test datasets according to a preset ratio;

[0012] The early warning model is trained based on the training dataset, the predetermined early warning model structure, the predetermined initial learning rate, and the predetermined maximum number of iterations.

[0013] The accuracy of the early warning model is verified based on the training dataset and the test dataset.

[0014] In response to the warning model failing the accuracy verification, the initial learning rate and the maximum number of iterations are redefined, and the steps of training the warning model based on the training dataset, the warning model structure, the initial learning rate, and the maximum number of iterations are returned until the warning model passes the accuracy verification, thus completing the training of the warning model.

[0015] In some embodiments, before the step of dividing the plurality of historical data into training datasets and test datasets according to a preset ratio, the method further includes:

[0016] Remove missing data points from the historical data;

[0017] Remove out-of-range data and over-alarm value data from the historical data;

[0018] Convert the historical data into dimensionless values.

[0019] In some embodiments, the step of validating the accuracy of the early warning model based on the training dataset and the test dataset includes:

[0020] Obtain the target neuron of the early warning model;

[0021] A second threshold is obtained based on the target neuron and the training dataset;

[0022] Input the test dataset into the early warning model and obtain the second distance dataset;

[0023] The proportion of abnormal samples in the second distance dataset is obtained based on the second threshold.

[0024] In response to the abnormal sample ratio being greater than or equal to a preset third threshold, the system outputs a result indicating that the early warning model has failed the accuracy verification.

[0025] In response to the abnormal sample ratio being less than the third threshold, the result of the early warning model passing the accuracy verification of the early warning model is output.

[0026] In some embodiments, the step of obtaining the second threshold based on the target neuron and the training dataset includes:

[0027] Obtain a first distance dataset, which includes multiple distance samples, the distance samples being obtained based on the training dataset and the target neuron;

[0028] The multiple distance samples in the first distance dataset are divided into multiple normal samples and multiple abnormal samples;

[0029] The second threshold is obtained based on the normal samples and the abnormal samples.

[0030] In some embodiments, the step of dividing the plurality of distance samples in the first distance dataset into a plurality of normal samples and a plurality of abnormal samples includes:

[0031] Determine the normal proportion of the first distance dataset; wherein the normal proportion is configured to be less than a second threshold;

[0032] The multiple distance samples are arranged from largest to smallest, and based on the normal ratio, the multiple distance samples are divided into multiple normal samples and multiple abnormal samples.

[0033] In some embodiments, the step of inputting the sample data into a pre-trained early warning model to obtain detection parameters in response to the high-pressure oil pump being in a pressure-stabilizing condition includes:

[0034] Obtain the minimum distance between the sample data and the target neuron in the early warning model;

[0035] The detection parameters are obtained based on the minimum distance and the second threshold.

[0036] In some embodiments, the step of acquiring sample data of the high-pressure oil pump in response to the high-pressure oil pump being in a stabilizing condition includes:

[0037] Obtain the rail pressure and the front end temperature;

[0038] The monotonicity of the front-end temperature is obtained in response to changes in the rail pressure;

[0039] In response to a change in the monotonicity of the front-end temperature, it is determined that the high-pressure oil pump is in the pressure-stabilizing condition;

[0040] The rail pressure and the front-end temperature when the high-pressure oil pump is in the pressure stabilization condition are determined as the sample data.

[0041] In some embodiments, it also includes:

[0042] In response to the preset operating time of the high-pressure oil pump, the sample data within the preset time period is acquired;

[0043] The early warning model was retrained based on the sample data.

[0044] In some embodiments, the first threshold is 1 or 0.9; the normal percentage is 95% or 99%; and the third threshold is 90%.

[0045] Accordingly, this application also provides a high-pressure oil pump fault early warning system, including:

[0046] The first acquisition module is configured to acquire real-time data sample data of the high-pressure oil pump in response to the high-pressure oil pump being in a pressure-stabilizing condition; the real-time data sample data includes the front-end temperature and rail pressure of the high-pressure oil pump.

[0047] The second acquisition module is configured to input the real-time data sample data into a pre-trained early warning model to acquire detection parameters;

[0048] The early warning module is configured to generate an early warning message in response to the detection parameter being greater than a first threshold.

[0049] Accordingly, this application also provides a high-pressure oil pump fault early warning device, including:

[0050] The host computer is used to execute the high-pressure oil pump fault early warning method as described in any of the above embodiments.

[0051] Compared with existing technologies, a high-pressure oil pump fault early warning method according to an embodiment of this application includes: in response to the high-pressure oil pump being in a pressure-stabilized operating condition, acquiring sample data of the high-pressure oil pump; the sample data includes the front-end temperature and rail pressure of the high-pressure oil pump; inputting the sample data into a pre-trained early warning model to acquire detection parameters; and in response to the detection parameters exceeding a first threshold, generating early warning information. Thus, by monitoring the front-end temperature and rail pressure of the high-pressure oil pump and inputting the data into a pre-trained early warning model, online fault early warning of the high-pressure oil pump can be achieved, allowing for early detection of potential faults, preventing further deterioration of the fault, reducing downtime and maintenance costs, and improving the reliability and stability of the diesel engine. Furthermore, the pre-trained early warning model can generate detection parameters based on the input front-end temperature and rail pressure data, and then judge the detection parameters by setting a first threshold. When the detection parameters exceed the first threshold, early warning information is generated, thereby improving the accuracy of fault detection and reducing false alarms and missed alarms.

[0052] It is understood that, compared with the prior art, the high-pressure oil pump fault early warning system provided in this application embodiment has all the technical features and beneficial effects of the above-mentioned high-pressure oil pump fault early warning method, which will not be repeated here.

[0053] It is understood that, compared with the prior art, the high-pressure oil pump fault early warning device provided in this application embodiment has all the technical features and beneficial effects of the above-mentioned high-pressure oil pump fault early warning method, which will not be repeated here. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a first flowchart illustrating a high-pressure oil pump fault early warning method provided in an embodiment of this application;

[0056] Figure 2 This is a second flowchart illustrating a high-pressure oil pump fault early warning method provided in an embodiment of this application;

[0057] Figure 3 A schematic diagram of the training process of the early warning model in a high-pressure oil pump fault early warning method provided in this application embodiment;

[0058] Figure 4 This is a schematic diagram of the structure of a high-pressure oil pump fault early warning system provided in an embodiment of this application;

[0059] Figure 5 This is a schematic diagram of a high-pressure oil pump fault early warning device provided in an embodiment of this application. Detailed Implementation

[0060] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0061] This application provides a high-pressure oil pump fault early warning method, please refer to it as well. Figure 1 and Figure 2 , Figure 1 This illustration shows a flowchart of a high-pressure oil pump fault early warning method provided in an embodiment of this application; Figure 2 This illustration shows a second flowchart of a high-pressure oil pump fault early warning method provided in an embodiment of this application. The first embodiment of this application provides a high-pressure oil pump fault early warning method, including steps 101 to 103.

[0062] Step 101: In response to the high-pressure oil pump being in a pressure-stabilizing condition, sample data of the high-pressure oil pump is acquired; the sample data includes the front-end temperature and rail pressure of the high-pressure oil pump. Specifically, this embodiment of the application analyzes the performance of the high-pressure oil pump by collecting sample data, and evaluates the operating efficiency and stability of the high-pressure oil pump by monitoring changes in front-end temperature and rail pressure, thereby monitoring and diagnosing the operating status of the high-pressure oil pump. If the sample data shows that the temperature is too high or the rail pressure fluctuates significantly, it may be necessary to adjust or optimize the operating parameters of the high-pressure oil pump to improve its performance and reliability.

[0063] Step 102: Input the sample data into the pre-trained early warning model to obtain detection parameters. Specifically, inputting the sample data into the pre-trained early warning model and obtaining detection parameters allows the detection parameters to reflect the operating status of the high-pressure oil pump.

[0064] Step 103: In response to the detection parameter exceeding a first threshold, a warning message is generated. Specifically, the first threshold is a preset value used to compare with the detection parameter to determine whether the detection parameter is normal or abnormal. When the detection parameter is abnormal, the generated warning message can promptly react to and handle potential faults in the high-pressure oil pump, reducing potential risks and losses to the high-pressure oil pump, preventing the high-pressure oil pump from continuing to operate under abnormal conditions, thereby protecting the safety and reliability of the diesel engine.

[0065] Please see Figure 3 , Figure 3This illustration shows a schematic diagram of the training process of the early warning model in a high-pressure oil pump fault early warning method provided in an embodiment of this application. In some embodiments, the early warning model is pre-trained based on the following steps: acquiring multiple historical data; the historical data includes historical front-end temperature and historical rail pressure generated by the high-pressure oil pump during historical stable operation; dividing the multiple historical data into training dataset and test dataset according to a preset ratio; training the early warning model based on the training dataset, a pre-determined early warning model structure, a pre-determined initial learning rate, and a pre-determined maximum number of iterations; verifying the accuracy of the early warning model based on the training dataset and test dataset; in response to the early warning model failing the accuracy verification, re-determining the initial learning rate and the maximum number of iterations, and returning to the steps of training the early warning model based on the training dataset, the early warning model structure, the initial learning rate, and the maximum number of iterations, until the early warning model passes the accuracy verification, thus completing the training of the early warning model. Specifically, the preset ratio of the training dataset to the test dataset can be, but is not limited to, 5:5, 6:4, 7:3, or 8:2. Thus, by using a training dataset, the early warning model learns from multiple historical data points about the historical operation and trends of the high-pressure oil pump, thereby improving the model's accuracy and enabling it to better predict potential failures or anomalies in the high-pressure oil pump. Furthermore, if the early warning model fails accuracy verification, the initial learning rate and maximum number of iterations need to be redefined to optimize the model's training parameters, making it better suited to the high-pressure oil pump's operating conditions. By continuously and repeatedly verifying the model's accuracy, adjusting the training parameters, and retraining the model, the accuracy and reliability of the prediction model can be improved.

[0066] In some embodiments, before dividing multiple historical data sets into training and testing datasets according to a preset ratio, the method further includes: removing missing data points from the historical data; removing out-of-range data and alarm-value data from the historical data; and converting the historical data into dimensionless values. Specifically, since some data points may not be recorded in the historical data, resulting in missing data points, these missing data points can negatively impact the training and testing of the early warning model. Therefore, it is necessary to remove missing data points to improve the accuracy of the early warning model. Secondly, due to sensor malfunctions, data acquisition errors, or other reasons, outliers may also exist in the historical data, i.e., data points exceeding the normal range, which will also affect the training and testing results of the prediction model. Therefore, it is also necessary to remove out-of-range data and alarm-value data to improve the reliability of the early warning model. For example, out-of-range data refers to values ​​that exceed the sensor's measurement range. For instance, if the rail pressure sensor's range is 0-3000 bar, and the rail pressure parameter exceeds 3000 bar, then the record is considered an outlier sample. Parameters exceeding the alarm value refer to those exceeding the alarm value by a certain multiple. For example, the normal upper limit of rail pressure is approximately 1800 bar; records exceeding this limit by more than double, i.e., 3600 bar, are considered outlier samples. Furthermore, historical data includes front-end temperature and rail pressure, which have different dimensions. In the early warning model, this difference in dimensions may degrade its performance. Therefore, it is necessary to convert historical data into dimensionless values ​​to eliminate the dimensional difference between front-end temperature and rail pressure, thereby improving the training and testing accuracy of the early warning model.

[0067] In the embodiments of this application, historical data is converted into dimensionless numerical values ​​based on the following formula.

[0068]

[0069] Where, x z-score x is a dimensionless numerical value. i For historical data, x min x is the minimum value in the historical data. max This represents the maximum value from historical data.

[0070] In some embodiments, the step of verifying the accuracy of an early warning model based on a training dataset and a test dataset includes: obtaining the target neuron of the early warning model; obtaining a second threshold based on the target neuron and the training dataset; inputting the test dataset into the early warning model and obtaining a second distance dataset; obtaining the proportion of abnormal samples in the second distance dataset based on the second threshold; outputting a result indicating that the early warning model has failed the accuracy verification in response to the proportion of abnormal samples being greater than or equal to a preset third threshold; and outputting a result indicating that the early warning model has passed the accuracy verification in response to the proportion of abnormal samples being less than the third threshold. Specifically, the structure of the early warning model in this application is a SOM network, and the target neuron refers to the optimal neuron in the SOM network. Thus, by evaluating the training results of the early warning model through accuracy verification, the accuracy and adjustability of the early warning model are improved.

[0071] In some embodiments, the step of obtaining a second threshold based on a target neuron and a training dataset includes: obtaining a first distance dataset, which includes multiple distance samples obtained based on the training dataset and the target neuron; dividing the multiple distance samples in the first distance dataset into multiple normal samples and multiple abnormal samples; and obtaining a second threshold based on the normal samples and abnormal samples. Thus, by dividing the samples in the first distance dataset into normal samples and abnormal samples, this application obtains a second threshold based on the characteristics and distribution of actual data, thereby characterizing the normal ratio of normal samples and abnormal samples, preparing for subsequent validation on test datasets, and improving the accuracy of the early warning model.

[0072] In some embodiments, the step of dividing multiple distance samples in a first distance dataset into multiple normal samples and multiple abnormal samples includes: determining the normal proportion of the first distance dataset; wherein the normal proportion is configured to be less than a second threshold; arranging the multiple distance samples from largest to smallest, and dividing the multiple distance samples into multiple normal samples and multiple abnormal samples based on the normal proportion. Specifically, this application first determines the normal proportion p to be 95% or 99%, and initially identifies the distance samples with a larger (1-p) proportion of values ​​in the distance dataset as abnormal samples, and the distance samples with a smaller p proportion of values ​​as normal samples. Then, a second threshold is determined based on the maximum value among the normal samples and the minimum value among the abnormal samples to ensure that the normal proportion is less than the second threshold. For example, the average of the maximum value among the normal samples and the minimum value among the abnormal samples is used as the second threshold. In this way, by determining the normal proportion, this application reasonably divides the distance samples into normal and abnormal samples, thereby ensuring the accuracy of the early warning model in identifying normal samples and improving the reliability of the early warning model. Secondly, by arranging the distance samples from largest to smallest and dividing them into normal and abnormal samples according to the normal proportion, abnormal samples can be better distinguished, thereby reducing the false alarm rate and improving the accuracy and reliability of the early warning model. Meanwhile, by dividing distance samples into normal and abnormal samples, information about the sample distribution can be obtained, which helps to further analyze and understand the characteristics of the distance dataset and the response of the early warning model to different samples, so as to improve the design and optimization of the early warning model.

[0073] In some embodiments, the first threshold is 1 or 0.9; the normal percentage is 95% or 99%; and the third threshold is 90%.

[0074] In some embodiments, the distance sample is obtained based on the following formula:

[0075]

[0076] Among them, D i For distance samples, a i and b i The data is historical, and n1 and n2 are the target neurons.

[0077] In some embodiments, in response to the high-pressure oil pump being in a stabilizing condition, the step of inputting sample data into a pre-trained early warning model to obtain detection parameters includes: obtaining the minimum distance between the sample data and the target neuron in the early warning model; and obtaining detection parameters based on the minimum distance and a second threshold. Specifically, when the detection parameter is greater than the first threshold, that is, the distance between the t-th sample data and the target neuron (i.e., the optimal neuron) is greater than the maximum distance between historical data in the training dataset and the corresponding target neuron, it is determined that the current data may be abnormal data. The first threshold can generally be set to 1.0 or 0.9. When the detection parameter is greater than 1, it means that the distance between the t-th sample data and the target neuron is greater than the maximum distance between the training data and the target neuron, that is, the t-th sample data has a large difference from the training data and may be abnormal data. The larger the detection parameter, the greater the degree of abnormality of the t-th sample data. In practical applications, there may be some unstable operating condition data in the training data. Setting the threshold to 1 will cause some abnormal data to not be detected in time. Therefore, a threshold slightly less than 1 can also be selected, such as 0.9, 0.8, or others, to provide the accuracy of the judgment. It can be adjusted according to the usage in engineering applications.

[0078] Specifically, this application obtains the detection parameters based on the following formula.

[0079]

[0080] Where, τ t For the detection parameter, r t is the minimum distance, and w is the second threshold.

[0081] In some embodiments, the step of acquiring sample data of the high-pressure oil pump in response to the high-pressure oil pump being in a stable pressure condition includes: acquiring rail pressure and front-end temperature; acquiring the monotonicity of the front-end temperature in response to a change in rail pressure; determining that the high-pressure oil pump is in a stable pressure condition in response to a change in the monotonicity of the front-end temperature; and determining the rail pressure and front-end temperature when the high-pressure oil pump is in a stable pressure condition as sample data. For example, when the rail pressure is rising, the front-end temperature is also rising, which is a variable operating condition; when the front-end temperature changes from rising to falling, it is determined that the current variable operating condition stage has ended, and the moment when the rising temperature changes to falling temperature is the end time of the variable operating condition, entering a steady-state operating condition. In this embodiment of the application, by acquiring the two parameters of rail pressure and front-end temperature, key information about the working state of the high-pressure oil pump can be obtained. By observing the monotonicity change of the front-end temperature, it can be determined whether the high-pressure oil pump is in a stable pressure condition, thereby filtering out sample data that meets the requirements and providing an accurate data foundation for subsequent analysis and modeling.

[0082] In some embodiments, the method further includes: acquiring sample data within a preset time period in response to a preset operating time of the high-pressure oil pump; and retraining the early warning model based on the sample data. Specifically, after the high-pressure oil pump has been running for a preset time, all steady-state operating condition data records of the high-pressure oil pump and a certain proportion of steady-state operating condition data of other high-pressure oil pumps of the same model are periodically selected. The preset time can be determined according to actual conditions, referring to the break-in period of the high-pressure oil pump of this model, for example, 100 hours. The duration of the periodicity can also be determined according to actual conditions, for example, 50 hours. Then, the early warning model (i.e., the SOM network) is retrained according to the training steps of the early warning model. When dividing the training samples and test samples, the proportion of training data and test data in the data of the high-pressure oil pump and other high-pressure oil pumps of the same model is kept consistent. After the early warning model is trained, it is compared with the abnormal proportion of the original test data. If the abnormal proportion of the test data of the current early warning model is less than the abnormal proportion of the original test data, the SOM network model, the third threshold, the initial learning rate, and the maximum number of iterations are updated. If the abnormal proportion of the test data of the current SOM network is not less than the abnormal proportion of the original test data, the SOM network model, the third threshold, the initial learning rate, and the maximum number of iterations are not updated. Thus, by periodically selecting steady-state operating data records of the high-pressure oil pump and data from other high-pressure oil pumps of the same model, the dataset of the early warning model can be kept updated, enabling it to adapt to changes in actual conditions. Secondly, retraining the early warning model can optimize its performance by using new sample data. By comparing the anomaly rate with the original test data, it can be determined whether to update the parameters and thresholds of the early warning model, giving it better adaptability and enabling it to adapt to different data distributions and changes.

[0083] An embodiment of this application provides a high-pressure oil pump fault early warning method, comprising: in response to the high-pressure oil pump being in a pressure-stabilized operating condition, acquiring sample data of the high-pressure oil pump; the sample data including the front-end temperature and rail pressure of the high-pressure oil pump; inputting the sample data into a pre-trained early warning model to acquire detection parameters; and in response to the detection parameters exceeding a first threshold, generating early warning information. Thus, by monitoring the front-end temperature and rail pressure of the high-pressure oil pump and inputting the data into a pre-trained early warning model, online fault early warning of the high-pressure oil pump can be achieved, enabling early detection of potential faults, preventing further deterioration of the fault, reducing downtime and maintenance costs, and improving the reliability and stability of the diesel engine. Furthermore, the pre-trained early warning model can generate detection parameters based on the input front-end temperature and rail pressure data, and then judge the detection parameters by setting a first threshold. When the detection parameters exceed the first threshold, early warning information is generated, thereby improving the accuracy of fault detection and reducing false alarms and missed alarms.

[0084] Accordingly, please refer to Figure 4 , Figure 4This illustration shows a structural diagram of a high-pressure oil pump fault early warning system provided in an embodiment of this application. The system includes a first acquisition module, a second acquisition module, and an early warning module. The first acquisition module is configured to acquire real-time data sample data of the high-pressure oil pump in response to the high-pressure oil pump being in a pressure-stabilizing condition. The real-time data sample data includes the front-end temperature and rail pressure of the high-pressure oil pump. The second acquisition module is configured to input the real-time data sample data into a pre-trained early warning model to acquire detection parameters. The early warning module is configured to generate early warning information in response to the detection parameters exceeding a first threshold.

[0085] It is understood that, compared with the prior art, the high-pressure oil pump fault early warning system provided in this application has all the technical features and beneficial effects of the above-mentioned high-pressure oil pump fault early warning method, which will not be repeated here.

[0086] Accordingly, please refer to Figure 5 , Figure 5 This illustration shows a structural diagram of a high-pressure oil pump fault early warning device provided in an embodiment of this application. The device includes a host computer, which executes the high-pressure oil pump fault early warning method as described in any of the above embodiments. The host computer is a control system host computer. Furthermore, the early warning device includes a human-machine interface unit and a signal acquisition and processing unit. The human-machine interface unit enables user interaction with the early warning device, such as setting early warning parameters, viewing early warning information, and performing fault diagnosis. The human-machine interface unit may include a display screen, keyboard, mouse, or touchscreen, allowing users to easily input commands, view results, and perform operations. The signal acquisition and processing unit collects and processes real-time data of key parameters such as the front-end temperature and rail pressure of the high-pressure oil pump, and performs signal processing, such as filtering, amplification, and digitization. The signal acquisition and processing unit can also preprocess the collected data, such as through calibration and normalization, to ensure the accuracy and reliability of the data. Through the signal acquisition and processing unit, the early warning device can acquire high-quality data for subsequent fault prediction and early warning analysis.

[0087] It is understood that, compared with the prior art, the high-pressure oil pump fault early warning device provided in this application embodiment has all the technical features and beneficial effects of the above-mentioned high-pressure oil pump fault early warning method, which will not be repeated here.

[0088] The above provides a detailed description of a high-pressure oil pump fault early warning method, system, and device provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the technical solutions and core ideas of this application. Those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A method for early warning of high-pressure oil pump failure, characterized in that, include: Obtain the rail pressure and the front end temperature of the high-pressure oil pump; The monotonicity of the front-end temperature is obtained in response to changes in the rail pressure. In response to a change in the monotonicity of the front-end temperature, the high-pressure oil pump is determined to be in a pressure-stabilizing condition. If both the rail pressure and the front-end temperature are rising, it is considered a variable condition. When the front-end temperature changes from rising to falling, the current variable condition phase is considered to have ended, with the moment of change being the end of the variable condition, and the pump enters the pressure-stabilizing condition. In response to the high-pressure oil pump being in the pressure-stabilizing condition, sample data of the high-pressure oil pump is acquired, and the rail pressure and front-end temperature when the high-pressure oil pump is in the pressure-stabilizing condition are determined as the sample data. The sample data is input into a pre-trained early warning model to obtain detection parameters; A warning message is generated in response to the detection parameter being greater than a first threshold; The early warning model is pre-trained based on the following steps: acquiring multiple historical data; the historical data includes the historical front-end temperature and historical rail pressure generated by the high-pressure oil pump during historical pressure stabilization operation; dividing the multiple historical data into training dataset and test dataset according to a preset ratio; and training the early warning model based on the training dataset, a pre-determined early warning model structure, a pre-determined initial learning rate, and a pre-determined maximum number of iterations. Obtain the target neuron of the early warning model; A second threshold is obtained based on the target neuron and the training dataset; the test dataset is input into the early warning model, and a second distance dataset is obtained; The proportion of abnormal samples in the second distance dataset is obtained based on the second threshold. In response to the abnormal sample ratio being greater than or equal to a preset third threshold, the system outputs a result indicating that the early warning model has failed the accuracy verification; it then redetermines the initial learning rate and the maximum number of iterations, and returns to the steps of training the early warning model based on the training dataset, the early warning model structure, the initial learning rate, and the maximum number of iterations, until the early warning model passes the accuracy verification, thus completing the training of the early warning model. In response to the abnormal sample ratio being less than the third threshold, the result of the early warning model passing the accuracy verification of the early warning model is output. The dataset of the early warning model is kept updated, and the sample data within the preset time is obtained in response to the preset operation time of the high-pressure oil pump; The early warning model is retrained based on the sample data to enable it to adapt to changes in the actual situation. After the early warning model is retrained, it is compared with the proportion of abnormal samples in the original test data. If the proportion of abnormal samples in the test data of the current early warning model is less than the proportion of abnormal samples in the original test data, the SOM network model, the third threshold, the initial learning rate, and the maximum number of iterations are updated. If the proportion of abnormal samples in the current test data of the SOM network is not less than the proportion of abnormal samples in the original test data, the SOM network model, third threshold, initial learning rate, and maximum number of iterations will not be updated.

2. The high-pressure oil pump fault early warning method as described in claim 1, characterized in that, Before the step of dividing the multiple historical data into training and testing datasets according to a preset ratio, the method further includes: Remove missing data points from the historical data; Remove out-of-range data and over-alarm value data from the historical data; Convert the historical data into dimensionless values.

3. The high-pressure oil pump fault early warning method as described in claim 2, characterized in that, The step of obtaining the second threshold based on the target neuron and the training dataset includes: Obtain a first distance dataset, which includes multiple distance samples, the distance samples being obtained based on the training dataset and the target neuron; The multiple distance samples in the first distance dataset are divided into multiple normal samples and multiple abnormal samples; The second threshold is obtained based on the normal samples and the abnormal samples.

4. The high-pressure oil pump fault early warning method as described in claim 3, characterized in that, The step of dividing the multiple distance samples in the first distance dataset into multiple normal samples and multiple abnormal samples includes: Determine the normal proportion of the first distance dataset; wherein the normal proportion is configured to be less than a second threshold; The multiple distance samples are arranged from largest to smallest, and based on the normal ratio, the multiple distance samples are divided into multiple normal samples and multiple abnormal samples.

5. The high-pressure oil pump fault early warning method as described in claim 1, characterized in that, The step of inputting the sample data into a pre-trained early warning model to obtain detection parameters in response to the high-pressure oil pump being in a pressure-stabilizing condition includes: Obtain the minimum distance between the sample data and the target neuron in the early warning model; The detection parameters are obtained based on the minimum distance and the second threshold.

6. The high-pressure oil pump fault early warning method as described in claim 4, characterized in that, The first threshold is 1 or 0.9; the normal percentage is 95% or 99%; and the third threshold is 90%.

7. A high-pressure oil pump fault early warning system, characterized in that, include: The first acquisition module is configured to acquire the rail pressure and the front end temperature of the high-pressure oil pump; The monotonicity of the front-end temperature is obtained in response to changes in the rail pressure. In response to a change in the monotonicity of the front-end temperature, the high-pressure oil pump is determined to be in a pressure-stabilizing condition. If both the rail pressure and the front-end temperature are rising, it is considered a variable condition. When the front-end temperature changes from rising to falling, the current variable condition phase is considered to have ended, with the moment of change being the end of the variable condition, and the pump enters the pressure-stabilizing condition. In response to the high-pressure oil pump being in the pressure-stabilizing condition, real-time data sample data of the high-pressure oil pump is acquired, and the rail pressure and front-end temperature when the high-pressure oil pump is in the pressure-stabilizing condition are determined as the sample data. The second acquisition module is configured to input the real-time data sample data into a pre-trained early warning model to acquire detection parameters; The early warning module is configured to generate an early warning message in response to the detection parameter being greater than a first threshold. The early warning model is pre-trained based on the following steps: acquiring multiple historical data; the historical data includes the historical front-end temperature and historical rail pressure generated by the high-pressure oil pump during historical pressure stabilization operation; dividing the multiple historical data into training dataset and test dataset according to a preset ratio; and training the early warning model based on the training dataset, a pre-determined early warning model structure, a pre-determined initial learning rate, and a pre-determined maximum number of iterations. Obtain the target neuron of the early warning model; A second threshold is obtained based on the target neuron and the training dataset; the test dataset is input into the early warning model, and a second distance dataset is obtained; The proportion of abnormal samples in the second distance dataset is obtained based on the second threshold. In response to the abnormal sample ratio being greater than or equal to a preset third threshold, the system outputs a result indicating that the early warning model has failed the accuracy verification; it then redetermines the initial learning rate and the maximum number of iterations, and returns to the steps of training the early warning model based on the training dataset, the early warning model structure, the initial learning rate, and the maximum number of iterations, until the early warning model passes the accuracy verification, thus completing the training of the early warning model. In response to the abnormal sample ratio being less than the third threshold, the result of the early warning model passing the accuracy verification of the early warning model is output. The dataset of the early warning model is kept updated, and the sample data within the preset time is obtained in response to the preset operation time of the high-pressure oil pump; The early warning model is retrained based on the sample data to enable it to adapt to changes in the actual situation. After the early warning model is retrained, it is compared with the proportion of abnormal samples in the original test data. If the proportion of abnormal samples in the test data of the current early warning model is less than the proportion of abnormal samples in the original test data, the SOM network model, the third threshold, the initial learning rate, and the maximum number of iterations are updated. If the proportion of abnormal samples in the current test data of the SOM network is not less than the proportion of abnormal samples in the original test data, the SOM network model, third threshold, initial learning rate, and maximum number of iterations will not be updated.

8. A high-pressure oil pump fault early warning device, characterized in that, include: A host computer, which is used to execute the high-pressure oil pump fault early warning method as described in any one of claims 1 to 6.

Images