Vehicle part service life prediction method based on user driving habits
By building a random forest model and real-time monitoring technology, combining vehicle network data, dynamically predicting vehicle component life is solved, and the problem of neglecting driving habits and environmental changes in traditional methods is achieved, more accurate component life prediction and personalized maintenance suggestions are achieved, and more accurate component life prediction and personalized maintenance suggestions are improved. The safety and economics of the vehicle.
Patent Information
- Application Number
- CN202410974975.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-19
- Publication Date
- 2024-11-19
AI Technical Summary
Traditional vehicle components life prediction methods ignore dynamic changes in user driving habits and usage environment, resulting in insufficient accuracy of prediction results and inability to promptly reflect the health status of vehicle components, which may lead to failure and safety hazards.
Using a vehicle component life prediction method based on user driving habits, a random forest model is constructed through characteristic processing, standardization and cleaning of training data, combined with vehicle network technology to monitor vehicle status in real time, dynamically predict component life, and consider driving behavior and environmental factors. Influence.
Real-time monitoring and accurate prediction of the health status of vehicle components is achieved, personalized maintenance suggestions are provided, and the reliability, safety and economy of the vehicle are improved, and the differences in usage modes of different users are adapted to different users.
Smart Images

Figure CN118981697A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of vehicle detection and analysis, and in particular relates to a method for predicting the life of vehicle components based on user driving habits. Background Art
[0002] With the progress of the times and the continuous upgrading of the automobile industry, today's automobile parts have become more and more sophisticated and intelligent. However, the more sophisticated the machinery and its parts, the more likely they are to affect the driving quality of the vehicle and the safety of the driver, especially for machines such as engines and gearboxes that are in high-speed motion for a long time. The wear and tear of different parts will vary due to personal driving habits, climate differences, frequency of car use, daily driving environment and other factors. Traditional vehicle maintenance is usually performed based on fixed time intervals or mileage, ignoring the impact of the above factors on the wear and tear of vehicle parts.
[0003] Traditional methods for predicting the life of vehicle components mainly rely on the manufacturer's factory test model, which is based on fixed time intervals or mileage to maintain or replace parts. However, these methods often ignore the user's changing usage patterns and driving habits during actual driving, resulting in insufficient accuracy of the prediction results. In the real world, the driving behavior of electric vehicle users, such as acceleration, braking, and ambient temperature, will have a significant impact on the performance and life of vehicle components. Existing maintenance prediction methods fail to fully consider these dynamic factors. Therefore, it is possible that failures that can be avoided by replacing parts in advance may suddenly occur during driving, resulting in a large amount of financial losses, and even the owner is more likely to lose his life in an accident. Therefore, more accurate prediction of the life of vehicle components is crucial.
[0004] Vehicle component life prediction methods usually rely on specific driving models or empirical models, which are often based on data under laboratory conditions. Due to the differences between laboratory environments and actual usage environments, the accuracy and reliability of these models in real-world scenarios are limited.
[0005] Under the existing repair and maintenance rules, cars are generally repaired, inspected or maintained at a repair station after the owner discovers a car failure. However, each owner's driving habits and car usage environment are different, and it is impossible to effectively guarantee the driving safety of the vehicle every time.
[0006] Many traditional maintenance prediction calculation methods are static and fail to update and reflect the health changes of components in real time during use. Such static methods lag behind the rapid changes in the health status of vehicle components and cannot provide timely and accurate health status assessment. Summary of the invention
[0007] In view of the above problems, the present invention provides a method for predicting the life of vehicle components based on user driving habits.
[0008] The vehicle component life prediction method based on user driving habits provided by the present invention comprises the steps of:
[0009] 1. Characterization processing of training data;
[0010] 2. Perform data cleaning on the standardized data obtained through the processing in step 1;
[0011] 3. Divide the standardized training data obtained after processing in step 2 into a training set and a test set;
[0012] 4. Feature data division, including maximum and minimum normalization processing and feature standardization of the data in the training set and the test set;
[0013] 5. Build a random forest model;
[0014] 6. Using the training set to train the random forest model;
[0015] 7. Use the test set to test the random forest model;
[0016] 8. Application of the random forest model.
[0017] further,
[0018] The step 1 includes determining the type of the feature data to be extracted, the feature data including vehicle data, driving behavior data, vehicle operating environment data and a combination thereof;
[0019] The step 1 comprises the steps of:
[0020] 1a. Standardization of the vehicle data, each piece of the vehicle data includes the frame number, model, year, manufacturer, component name, replacement time, maintenance time, and component material;
[0021] 1b. Standardization of the driving behavior data, each of which includes the following data: vehicle frame number, user identifier, average vehicle speed, maximum vehicle speed, average engine speed, maximum engine speed, average pedal position, total mileage, average single mileage, number of accelerations, average number of accelerations per kilometer, number of brakes, average number of brakes per kilometer, idling time, average idling time per time, number of speeding, average number of speeding per kilometer, instantaneous fuel consumption, average fuel consumption per 100 kilometers;
[0022] 1c. Standardization of the vehicle operating environment data, each piece of the vehicle operating environment data includes: frame number, user's city, ambient temperature, ambient humidity, and road conditions.
[0023] further,
[0024] The step 2 comprises:
[0025] 2a. Missing value processing:
[0026] For any vehicle data A in the vehicle data, if the vehicle frame number is missing, directly delete the vehicle data A;
[0027] For any driving behavior data B in the driving behavior data, if the driving behavior data lacks the vehicle frame number and the user identifier, directly delete the driving behavior data B;
[0028] For any piece of vehicle operating environment data C in the vehicle operating environment data, if the vehicle frame number is missing, the vehicle operating environment data C is directly deleted.
[0029] 2b. Outlier processing: Use KMeans algorithm to filter out abnormal data;
[0030] 2c. Feature encoding: For non-numeric feature data, one-hot encoding or label encoding is used.
[0031] further,
[0032] The step 2b comprises the steps of:
[0033] 2b1: Feature data preparation: Collect feature data to be processed, including driving behavior data and environmental data, including vehicle frame number, user identifier, average vehicle speed, maximum vehicle speed, average speed, maximum speed, average pedal position, total mileage, average single mileage, number of accelerations, average number of accelerations per kilometer, number of brakes, average number of brakes per kilometer, idling time, average idling time per time, number of speeding, average number of speeding per kilometer, instantaneous fuel consumption, average fuel consumption per 100 kilometers, ambient temperature, ambient humidity, and road conditions;
[0034] 2b2: Characteristic data standardization: The characteristic data to be processed are standardized so that the values of the characteristic data to be processed are within the same range. The standardization formula used in this step is as follows:
[0035]
[0036] In the above formula, z is the value of the feature data to be processed after standardization, x0 is the original feature value, μ is the mean of the feature data to be processed, and σ is the standard deviation of the feature data to be processed;
[0037] 2b3: KMeans clustering: Select an appropriate cluster number kkk, and input the standardized feature data to be processed into the KMeans algorithm as the corresponding data point. The KMeans algorithm includes the following steps:
[0038] 2b3a, Initialization: Randomly select kkk points from the data points as initial cluster centers;
[0039] 2b3b. Assign the data points: assign each data point to the cluster center closest to it, and calculate the distance from each data point to all cluster centers:
[0040]
[0041] In the above formula, x and y are two data points, n is the dimension of the feature, and i represents the dimension of the i-th feature;
[0042] 2b3c. Update the cluster center: recalculate the center of each cluster, the formula is:
[0043]
[0044] In the above formula, C j is the new center of the jth cluster, S j is the set of data points belonging to the jth cluster, x S,j For the data, |S j |For S j The number of data points;
[0045] Repeat steps 2b2 and 2b3 until the position change of the cluster center does not exceed a preset position threshold or reaches a preset maximum number of iterations;
[0046] 2b4: Calculate the distance between each data point and the center of the cluster to which it belongs:
[0047]
[0048] Among them, C j is the center of the cluster to which the data point x belongs;
[0049] 2b5: Calculate the mean and standard deviation of the distances from all data points to their cluster centers using the following formula:
[0050]
[0051] In the above two formulas, μ d is the mean, σ d is the standard deviation, N is the cluster center C j The data point x in the cluster i1 Total number of;
[0052] 2b6: Set the threshold H of outliers:
[0053] H=μ d +k·σ d ,
[0054] Where k is a multiple, either 2 or 3:
[0055] 2b7: Filter abnormal data: Mark the data points exceeding the threshold value H as abnormal data, and the threshold value H satisfies:
[0056] d(x,C j )>H;
[0057] 2b8: Manual confirmation and processing: Manually confirm the abnormal data screened out, select a k value, wherein the proportion of the abnormal data is controlled to be 5% or less of the total data, and delete or correct the abnormal data.
[0058] further,
[0059] In step 2b2, for non-numerical feature data: vehicle model, year, manufacturer, component name, component material, one-hot encoding or label encoding is used.
[0060] further,
[0061] In step 3, 70% of the standardized training data obtained after processing in step 2 is used as the training set, and 30% of the standardized training data obtained after processing in step 2 is used as the test set.
[0062] further,
[0063] In step 4,
[0064] 4a. The following characteristic data obtained in step 1: average vehicle speed, maximum vehicle speed, average speed, maximum speed, average pedal position, total mileage, average single mileage, number of accelerations, average number of accelerations per kilometer, number of brakes, average number of brakes per kilometer, idling time, average idling time per time, number of speeding, average number of speeding per kilometer, instantaneous fuel consumption, average fuel consumption per 100 kilometers are normalized to the maximum and minimum using the following formula:
[0065]
[0066] In the above formula, x1 is the normalized feature data value, x0 is the original feature value, and x min and x max are respectively the minimum and maximum values of the characteristic data;
[0067] 4b. Standardization of characteristic data: The characteristic data obtained after processing in step 4a is scaled to a range with a mean of 0 and a standard deviation of 1 using the method in step 2b2.
[0068] further,
[0069] In step 5,
[0070] The initial parameters of the random forest model: the number of trees n_estimators, the maximum depth max_depth:
[0071] on_estimators = 100,
[0072] omax_depth=None.
[0073] further,
[0074] In step 7,
[0075] During the test, for the prediction object component A, the difference between the component life in the prediction result obtained by the test and the actual component life is classified and compared according to the three dimensions of vehicle data, driving behavior data, and vehicle operating environment data, and the error mean or standard deviation of each type of data is calculated. If the mean square error MSE of the difference between the prediction result and the actual measurement is large or the R square is not close to 1, the number of trees n_estimators and the maximum depth max_depth of the random forest model are unreasonable, and the parameters are adjusted and steps 2 to 7 are repeated for retraining.
[0076] The model was optimized while testing: the parameters were adjusted by grid search and cross-validation, and the best parameter combination was found: n_estimators = 200 max_depth = 20,
[0077] The classification comparison comprises the steps of:
[0078] 7a. Data grouping: group the data according to different classification standards to obtain at least three groups of data: vehicle type in vehicle data, average vehicle speed in driving behavior data, and ambient temperature in operating environment data: vehicle type data group, average vehicle speed data group, and ambient temperature data group.
[0079] in,
[0080] The data element in the vehicle model data group includes three sub-elements: a vehicle model sub-element storing a vehicle model, a component A life actual value element storing an actual value of component A life, and a component A life prediction value element storing a predicted value of component A life. The vehicle models include: SUV, compact car, and truck. The data elements of vehicle models in the vehicle model data group that are all SUVs constitute an SUV group, the data elements of vehicle models that are all compact cars constitute a compact car group, and the data elements of vehicle models that are all trucks constitute a truck group. Thus, the vehicle model data group includes three data groups: an SUV group, a compact car group, and a truck group.
[0081] The data element in the average vehicle speed data group includes three sub-elements: an average vehicle speed sub-element storing the average vehicle speed, the component A life actual value element storing the actual value of component A life, and the component A life prediction value element storing the predicted value of component A life. The average vehicle speed is divided into three gears: low speed, medium speed, and high speed. Low speed refers to an average vehicle speed of 0 to 40 km / h, medium speed refers to an average vehicle speed of 40 km / h to 70 km / h, and high speed refers to an average vehicle speed of more than 70 km / h. In the average vehicle speed data group, the data elements whose average vehicle speeds are all low speeds constitute a low speed group, the data elements whose average vehicle speeds are all medium speeds constitute a medium speed group, and the data elements whose average vehicle speeds are all high speeds constitute a high speed group. Thus, the average vehicle speed data group includes three data groups: a low speed group, a medium speed group, and a high speed group.
[0082] The data element in the large environmental temperature data group includes three sub-elements: an environmental temperature sub-element storing the environmental temperature, an actual life value element of component A storing the actual life value of component A, and a predicted life value element of component A storing the predicted life value of component A. The environmental temperature is divided into three levels: low temperature, medium temperature, and high temperature. Low temperature refers to an environmental temperature of 0-15°C, medium temperature refers to an environmental temperature of 15°C to 30°C, and high temperature refers to an environmental temperature of 30°C or above. In the large environmental temperature data group, the data elements whose environmental temperatures are all low temperatures constitute a low temperature group, the data elements whose environmental temperatures are all medium temperatures constitute a medium temperature group, and the data elements whose environmental temperatures are all high temperatures constitute a high temperature group. Thus, the large environmental temperature data group includes three data groups: a low temperature group, a medium temperature group, and a high temperature group.
[0083] 7b. Calculate the mean square error (MSE) of each data group.
[0084] For each of the data sets, the mean square error MSE is calculated:
[0085]
[0086] Wherein, M is the number of data elements in any of the data groups, which is an integer not less than 1, and i2 represents the i2th data element in any of the data groups;
[0087] 7c. Compare and analyze the data of each major group.
[0088] For the large group of vehicle model data, the large group of average vehicle speed data, and the large group of ambient temperature data, the mean square error (MSE) of each of the data groups is compared respectively, and the data group whose mean square error (MSE) exceeds the expected threshold is found, and then the corresponding model parameters are optimized.
[0089] further,
[0090] For the vehicle data other than the vehicle model, the driving behavior data other than the average vehicle speed, and the operating environment data other than the ambient temperature, according to the grouping comparison steps 7a to 7c, grouping rules are formulated for comparison based on specific operating conditions.
[0091] The vehicle component life prediction method based on user driving habits provided by the present invention utilizes Internet of Vehicles technology and real-time data collection to achieve real-time monitoring and dynamic prediction of vehicle status, overcomes the lag of traditional static methods, and can accurately reflect changes in the health status of vehicle components during actual use. The present invention combines the user's specific driving habits, such as acceleration, braking frequency, vehicle usage scenarios, etc., to provide personalized component life prediction. This personalized analysis can reflect the differences in usage patterns of different users, thereby providing a more realistic health status assessment. The present invention improves the accuracy of prediction by monitoring changes in external factors such as ambient temperature, humidity, and vibration in real time, and comprehensively analyzing their impact on component life.
[0092] The present invention can effectively capture the impact of driving behavior on automobile components and provide more accurate and practical component life prediction results and maintenance suggestions. This innovative method not only conforms to the trend of intelligent and personalized automobile maintenance, but also helps to improve the reliability, safety and economy of vehicles.
[0093] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0094] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0095] Figure 1A flow chart of a method for predicting the life of vehicle components based on user driving habits according to an embodiment of the present invention is shown. DETAILED DESCRIPTION
[0096] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0097] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which this application belongs; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" in the specification and claims of this application and the above-mentioned figure descriptions and any variations thereof are intended to cover non-exclusive inclusions. The terms "first", "second", "third", etc. in the specification and claims of this application or the above-mentioned figures are used to distinguish different objects, not to describe a specific order or a primary and secondary relationship. "Multiple" appearing in this application refers to more than two (including two).
[0098] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.
[0099] Figure 1 The flowchart of the vehicle component life prediction method based on user driving habits provided by the present invention is shown. Figure 1 The vehicle component life prediction method based on user driving habits includes the following steps.
[0100] Step 1: Characterization of training data. This includes determining the type of feature data (which can be abbreviated as features) to be extracted. These feature data may include vehicle data, driving behavior data, vehicle operating environment data, and combinations thereof. The vehicle frame number is used as a unique identifier for the vehicle to associate various types of data as all data for the same vehicle. After processing in this step, the training data can be processed and used in the same subsequent manner. This step includes the following steps:
[0101] 1a. Standardization of vehicle data.
[0102] Each piece of vehicle data includes the frame number, model, year, manufacturer, component name, replacement time, maintenance time, and component material.
[0103] 1b. Standardization of driving behavior data.
[0104] It includes sorting out the driving data reported by the hardware equipment and extracting the driving behavior data. Each driving behavior data includes the following data: frame number, user identifier (ID), average vehicle speed, maximum vehicle speed, average engine speed, maximum engine speed, average pedal position, total mileage, average single mileage, number of accelerations, average number of accelerations per kilometer, number of brakes, average number of brakes per kilometer, idling time, average idling time each time, number of speeding, average number of speeding per kilometer, instantaneous fuel consumption, and average fuel consumption per 100 kilometers.
[0105] Idle time refers to the time when the engine is running but the vehicle is not moving. In order to ensure the accuracy of the data and avoid excessive data volume, effectively reduce the data volume, ensure that the recorded data is meaningful and can reflect the actual driving behavior, the following conditions or restrictions are set when recording idle time:
[0106] 1. Time threshold: Only record the idling time that exceeds a certain time threshold. For example, if you set the threshold to 30 seconds, the idling time will be recorded only when it exceeds 30 seconds, which can avoid recording short parking waiting time.
[0107] 2. Speed threshold: When the vehicle speed is below a certain threshold (e.g., 0 km / h), the engine running time is recorded as idling time. This ensures that idling time is only recorded when the vehicle is completely stopped.
[0108] 3. Position change: Idle time is recorded at a relatively fixed position (for example, when the GPS coordinates change very little) to ensure that the idle time does not include low-speed driving time. If the vehicle is started and the vehicle's GPS coordinates change within the next 5 seconds does not exceed the threshold of 5 meters (m), the vehicle is considered to be in an idle state from the start time until the vehicle coordinates change exceeds the threshold of 5m.
[0109] 4. Exclude special situations: You can set some rules to exclude idling time that does not need to be recorded, such as common situations such as short stops at traffic lights.
[0110] 5. Segmented recording: If the continuous idling time exceeds a certain length (for example, 10 minutes), it can be recorded in segments or averaged to avoid a single data point with too long a recording time.
[0111] The instantaneous fuel consumption refers to the fuel consumption rate of the vehicle at a certain moment, usually expressed in liters per 100 kilometers (L / 100km) or miles per gallon (MPG). In order to ensure the accuracy of the data and avoid excessive data volume, effectively reduce the amount of recorded instantaneous fuel consumption data, ensure that the data is representative and practical, and improve the efficiency of data processing and the accuracy of model prediction, the following conditions and restrictions are set when recording instantaneous fuel consumption:
[0112] 1. Time interval: Set a fixed time interval to record instantaneous fuel consumption. For example, record once every 5 seconds, 10 seconds or 30 seconds. This can avoid continuously recording instantaneous fuel consumption data, thereby reducing the amount of data.
[0113] 2. Speed threshold: Instantaneous fuel consumption is recorded only when the vehicle speed exceeds a certain threshold. For example, instantaneous fuel consumption is recorded only when the vehicle speed exceeds 10 km / h to avoid recording instantaneous fuel consumption data at low speed or idling.
[0114] 3. Fuel consumption change threshold: Record only when the instantaneous fuel consumption change exceeds a certain threshold. For example, if the fuel consumption change threshold is set to 0.1L / 100km, it will only be recorded when the instantaneous fuel consumption change exceeds 0.1L / 100km. This can filter out some minor changes and reduce the amount of data.
[0115] 4. Event triggering: Record instantaneous fuel consumption data when a specific driving event (such as acceleration, braking, turning, etc.) occurs. This method can focus on recording driving behaviors that have a greater impact on fuel consumption, thereby reducing unnecessary data.
[0116] 5. Data smoothing: Smoothing the instantaneous fuel consumption data (e.g., moving average method) to reduce data fluctuations and noise, while recording the smoothed data can reduce the amount of data and improve data stability.
[0117] 6. Location change: Combined with GPS location data, record the instantaneous fuel consumption in different road sections or different driving environments. This can analyze the relationship between fuel consumption and driving environment in a targeted manner, while avoiding excessive duplication of data.
[0118] The hardware equipment mainly refers to the equipment used to collect vehicle driving data, including but not limited to the following:
[0119] 1. On-Board Diagnostics (OBD) devices: OBD devices can be plugged into the vehicle's OBD port to collect vehicle operating status and fault information by reading the vehicle's electronic control unit (ECU) data. Common OBD devices include OBD-II adapters, which can be connected to mobile phone applications or other data recording systems.
[0120] 2. Global Position System (GPS) device: GPS device can record vehicle location, speed, driving route and other information to help analyze driving behavior and driving habits.
[0121] 3. Digital Video Record (DVR): The digital video recorder can record video and audio information during driving. Some advanced recorders also have the function of speed measurement and location information recording.
[0122] 4. Telematics Devices: These devices are usually provided by vehicle manufacturers or third parties. They can monitor various vehicle data in real time, including speed, fuel consumption, driving behavior, etc., and transmit the data to the cloud via wireless networks for analysis and storage.
[0123] 5. Tire Pressure Monitoring System (TPMS): TPMS can monitor the tire pressure and temperature in real time and provide safety warning information.
[0124] 6. Driving behavior recorder: A device specifically used to record driving behavior, which can monitor driving operations such as acceleration, braking, turning, and perform data analysis.
[0125] Currently, cars with newer platforms will integrate the above devices or send the integrated data to cloud storage.
[0126] 1c. Standardization of vehicle operating environment data.
[0127] Each piece of vehicle operating environment data includes: vehicle frame number, user's city, ambient temperature, ambient humidity, and road conditions. The ambient temperature, ambient humidity, and road condition data must meet the following conditions and restrictions to ensure the accuracy and validity of the data and avoid excessive data volume:
[0128] Ambient temperature:
[0129] 1. Time interval: Set a fixed time interval to record the ambient temperature, such as once every hour, which can reduce the amount of data.
[0130] 2. Position change: Record temperature data when the vehicle position changes significantly (for example, exceeds a certain distance). Specifically, calculate the distance between the current vehicle position and the last recorded vehicle position. If the distance exceeds a set threshold such as 10km, record the current temperature data and position. This can capture temperature changes in different areas.
[0131] 3. Temperature change threshold: Record only when the ambient temperature changes by more than a certain threshold, for example, record once when the temperature changes by more than 1°C, so as to avoid recording frequent small changes.
[0132] 4. Event triggering: Record temperature data when a specific event occurs, such as when a vehicle starts, stops, or enters a specific area.
[0133] Ambient humidity:
[0134] 1. Time interval: You can also set a fixed time interval to record the ambient humidity, for example, once every hour.
[0135] 2. Position change: Record humidity data when the vehicle position changes significantly. Specifically, calculate the distance between the current vehicle position and the last recorded vehicle position. If the distance exceeds a set threshold, such as 10km, record the current humidity data and position to ensure that humidity differences in different areas are captured.
[0136] 3. Humidity change threshold: record when the ambient humidity changes by more than a certain threshold, for example, record once when the humidity changes by more than 5%.
[0137] 4. Event triggering: Record humidity data when a specific event occurs, such as when a vehicle enters or leaves a specific environment. The specific environment includes but is not limited to tunnels, underground parking lots, car washers, mountainous areas, unpaved roads, and coastlines; the specific event is triggered by GPS fences and sensor detection.
[0138] Road conditions:
[0139] 1. Road condition category: Use predefined road condition categories (such as highways, urban roads, rural roads, etc.) to record road condition information, and record once each time you enter a new road condition category.
[0140] 2. Time interval: You can set a fixed time interval such as 10 minutes to record traffic information, but this needs to be combined with changes in traffic conditions.
[0141] 3. Event triggering: Recording traffic data when a specific event occurs, such as when a vehicle enters an urban road from a highway or when a vehicle enters a rural road from an urban road.
[0142] 4. Location change: Combined with GPS location data, the vehicle records road condition information when it passes through a specific road section or area, such as when it passes through an area with poor road conditions. Among them, poor road conditions include the following road conditions:
[0143] 1. Road surface conditions:
[0144] Potholes: There are many potholes on the road, which causes frequent vehicle bumps;
[0145] Cracks: There are obvious cracks on the road surface, which affect driving stability;
[0146] Uneven: The road surface is uneven with ups and downs.
[0147] 2. Traffic congestion:
[0148] High-density traffic: When vehicles travel in high-density traffic areas, their speed is significantly reduced;
[0149] Frequent braking: Due to severe traffic conditions, the vehicle needs to brake and start frequently.
[0150] 3. Construction area:
[0151] Road construction: vehicles passing through the construction area, resulting in partial closure of the road or temporary diversion;
[0152] Obstacle: There are construction equipment or materials on the road, which affects normal driving.
[0153] 4. Weather factors:
[0154] Water accumulation: Due to rainfall, there is water accumulation on the road, increasing driving risks;
[0155] Snow and ice cover: During winter, roads may be covered with snow and ice, making them slippery.
[0156] 5. Speed limits and speed limit changes:
[0157] Low-speed limit zone: Road speed limits are significantly reduced, such as school areas, residential areas, etc.
[0158] Frequent speed limit changes: Road speed limit signs change frequently, requiring frequent deceleration or acceleration.
[0159] 6. External data sources:
[0160] Map service data: Use real-time traffic data provided by map services, such as Google Maps, Baidu Maps, etc., to mark areas with “poor traffic conditions”.
[0161] The above road condition status can be determined by technicians through vehicle sensor data or external data sources (such as road condition information provided by map services).
[0162] Comprehensive consideration:
[0163] 1. Data smoothing and aggregation: Perform smoothing (such as moving average method) and aggregation (such as aggregation by time period or region) on ambient temperature, humidity and road condition data to reduce data volume and noise and improve data validity.
[0164] 2. Correlation analysis: Combined with other data (such as vehicle speed, mileage, etc.) for comprehensive analysis to extract meaningful environmental features to reduce unnecessary duplicate records.
[0165] The above conditions and restrictions ensure that the recording of ambient temperature, humidity and road condition data is representative and practical, while effectively controlling the amount of data.
[0166] Step 2: Clean the standardized data obtained in step 1 to ensure that the standardized training data obtained can ensure its quality and meet the needs of subsequent models. This step includes the following steps:
[0167] 2a. Missing value processing.
[0168] All training data revolves around vehicles, driving behaviors, and driving behaviors or driving habits. For a certain vehicle data A, the frame number is the unique identification data. If the frame number is missing, the vehicle data A will be deleted directly. The frame number and user ID in the driving behavior data are combined to uniquely identify the data. If a certain driving behavior data B lacks the frame number and user ID, the driving behavior data B will also be deleted directly. For a certain vehicle operating environment data C, if it lacks the frame number, the vehicle operating environment data C will also be deleted directly.
[0169] 2b. Outlier processing.
[0170] Since the driving information, i.e. the driving behavior data, is reported directly by the hardware device, the data range needs to be detected to exclude obviously abnormal values. In this step, we use the KMeans algorithm to preliminarily screen abnormal data. K-means is a clustering algorithm that can be used for outlier detection. The basic idea is to cluster the data first, and then determine whether the point is an outlier by calculating the distance between each point and the center of its cluster. Usually, outliers (outliers) are far away from the center of the cluster. When using the K-means clustering algorithm for outlier detection, setting the distance threshold is a key step. The main purpose of setting the threshold is to distinguish between normal points and abnormal points. Here we use the mean plus standard deviation method based on statistical characteristics.
[0171] Among them, the use of KMeans algorithm to filter abnormal data includes the following steps:
[0172] 2b1: Feature data preparation.
[0173] First, collect the characteristic data that needs to be processed, including driving behavior data and environmental data. These data should specifically include vehicle frame number, user ID, average vehicle speed, maximum vehicle speed, average speed, maximum speed, average pedal position, total mileage, average single mileage, number of accelerations, average number of accelerations per kilometer, number of brakes, average number of brakes per kilometer, idling time, average idling time each time, number of speeding times, average number of speeding times per kilometer, instantaneous fuel consumption, average fuel consumption per 100 kilometers, ambient temperature, ambient humidity, road conditions, etc.
[0174] 2b2: Feature data standardization.
[0175] The feature data to be processed are standardized so that the values of each feature data are in the same range. For non-numerical feature data such as vehicle model, year, manufacturer, component name, component material, etc., one-hot encoding or label encoding can be used so that the model can process it. The standardization formula used in this step is as follows:
[0176]
[0177] In the above formula, z is the value of the standardized feature data, x0 is the original feature data value, μ is the mean of the feature data, and σ is the standard deviation of the feature data.
[0178] This step scales the eigenvalues to a fixed range, usually [0,1]. For example, the average speed is represented by a number in the range [0,1]. For example, the average speeds are: 0km / h, 75km / h, 100km / h, 150km / h. After encoding, the corresponding eigenvalues are: 0, 0.5, 0.67, 1.
[0179] The range difference of feature values will affect the training efficiency of machine learning models. Through feature encoding, the magnitude difference between different features can be eliminated, thereby accelerating the convergence of the model and improving training efficiency.
[0180] 2b3: KMeans clustering.
[0181] Select a suitable clustering number kkk, input the standardized feature data into the KMeans algorithm, and the KMeans algorithm divides each standardized feature data into corresponding data points. Each data point includes no less than one standardized feature data, and the clustering number kkk is an integer not less than. The KMeans algorithm includes the following steps:
[0182] 2b3a. Initialization: Randomly select kkk data points as initial cluster centers.
[0183] 2b3b. Assign data points: Assign each data point to the cluster center closest to it, and calculate the distance from each data point to all cluster centers:
[0184]
[0185] In the above formula, x and y are two data points, n is the dimension of the feature, that is, the number of feature data for each data point, and i represents the dimension of the i-th feature.
[0186] 2b3c. Update cluster centers: Recalculate the center of each cluster using the formula:
[0187]
[0188] In the above formula, C j is the new center of the jth cluster, S j is the set of data points belonging to the jth cluster, x S,j For the data, |S j |For S j The number of data points.
[0189] Repeat 2b2 and 2b3 until each cluster center no longer changes significantly (i.e., the position change of each cluster center does not exceed the preset position threshold) or the preset maximum number of iterations is reached. The cluster center no longer changes significantly means that, with reference to the following formula (2), during iteration, the distance between the position of the cluster center calculated at the current time and the position calculated at the last time is less than a preset threshold, such as 0.01.
[0190] 2b4: Calculate the distance from each data point to the center of the cluster to which it belongs:
[0191]
[0192] Where x is the data point, C j is the center of the cluster to which it belongs.
[0193] 2b5: Calculate the mean and standard deviation of the distances from all data points to their cluster centers using the following formula:
[0194]
[0195] In the above two formulas, μ d is the mean of the distances, σ d is the standard deviation of the distance, N is the cluster center C j The data point x in the cluster i1 The total number of .
[0196] 2b6: Set the threshold H of outliers:
[0197] H=μd +k·σ d ,
[0198] Here, k is a multiple, which can be 2 or 3.
[0199] 2b7: Filter out abnormal data:
[0200] According to the set threshold H, the data points x that exceed the threshold are marked as abnormal data. The threshold H satisfies:
[0201] d(x i ,C j )>H.
[0202] 2b8: Manual confirmation and processing:
[0203] The screened abnormal data are manually confirmed, the k value is selected for the abnormal data (the abnormal data ratio is controlled to be 5% or less of the total data), and the abnormal data are deleted or corrected to ensure accuracy.
[0204] Through the above steps, the KMeans algorithm can be used to effectively filter out outliers in driving behavior data and environmental data, and perform corresponding processing to improve data quality and the accuracy of subsequent model predictions.
[0205] Step 3: Divide the standardized training data into training set and test set.
[0206] The standardized training data obtained after processing in step 2 is divided into a training set and a test set. Usually, 70% of the standardized training data is used as the training set, and 30% of the standardized training data is used as the test set.
[0207] Step 4: Feature segmentation.
[0208] Use feature scaling methods to scale the data in the training set and test set to ensure that all features have relatively equal weights in model training, that is, to ensure that all features have consistent scales in subsequent model training and prediction processes, thereby improving the model's training efficiency and prediction performance. Common methods include standardization and normalization. This step includes the following steps:
[0209] 4a. The following characteristic data obtained in step 1: average vehicle speed, maximum vehicle speed, average speed, maximum speed, average pedal position, total mileage, average single mileage, number of accelerations, average number of accelerations per kilometer, number of brakes, average number of brakes per kilometer, idling time, average idling time each time, number of speeding, average number of speeding per kilometer, instantaneous fuel consumption, average fuel consumption per 100 kilometers are normalized to the maximum and minimum using the following formula.
[0210]
[0211] In the above formula, x1 is the normalized feature data value, x0 is the original feature data value, and x min and x max are the minimum and maximum values of the feature data respectively.
[0212] 4b. Feature standardization: Scale the value of the feature data obtained after processing in step 4a to within the range of a mean of 0 and a standard deviation of 1. This method is applicable to the case where the data distribution has no obvious boundaries. The standardization adopts formula (1). The difference is that the processed data is different. In this step, when formula (1) is adopted, the feature data therein is the feature data obtained after processing in step 4a, and in step 2b2, the feature data in formula (1) is the feature data to be processed collected in step 2b1. The standardization can eliminate magnitude differences: the magnitudes of different features may vary greatly. Through standardization, they can be scaled to the same magnitude to prevent certain features from having too much influence on the model; improve model training efficiency: many machine learning algorithms (such as KMeans, KNN, SVM, etc.) are sensitive to the scale of the data. Through standardization, the convergence of the model can be accelerated and the training efficiency can be improved; applicable to most models: the standardized data is more suitable for most machine learning algorithms, especially those that need to calculate distances.
[0213] Step 5: Build a random forest model.
[0214] This step includes splitting feature variables and target variables. The target variable is the value we want to predict, and the feature variable is the data used to predict the target variable. For example, among the features mentioned in step 1: driving behavior data such as vehicle type, average speed, and environment, and operating environment data are the data features used to train the prediction results, i.e., feature variables, while maintenance data such as maintenance time and component replacement time are the target variables we want to predict.
[0215] In this step, a random forest model is constructed to predict the life of vehicle components. The construction of this model includes the following specific steps:
[0216] The Random Forest model is an ensemble learning method that improves the accuracy and robustness of predictions by building multiple decision trees and combining their results.
[0217] The construction of the model includes the following specific steps:
[0218] 5a1. Feature selection and preparation:
[0219] As mentioned above, feature data that has a greater impact on the life of vehicle components is selected, including vehicle data, driving behavior data, and operating environment data;
[0220] Use the standardized and normalized feature data as the input of the model.
[0221] 5a2. Divide the training set and test set:
[0222] As mentioned above, the standardized and normalized dataset is divided into 70% training set and 30% test set.
[0223] 5a2. Model initialization:
[0224] Determine the initial parameters of the random forest model: the number of trees (n_estimators), maximum depth (max_depth), etc. The initial parameters are set to:
[0225] on_estimators = 100,
[0226] omax_depth=None.
[0227] Step 6: Use the training set to train the random forest model. The training in this step is the initial training.
[0228] Step 7: Use the test set to test the random forest model.
[0229] During the test, for the prediction object component A, the difference between the component life in the prediction result obtained from the test and the actual component life is classified and compared according to the three dimensions of vehicle data, driving behavior data, and vehicle operating environment data, that is, classified and compared according to vehicle data, driving behavior data, and vehicle operating environment data, and the error mean and standard deviation of each type of data are calculated. Through the classification comparison, the impact of different features on the prediction error can be comprehensively analyzed to find out the possible improvement direction of the model, such as which features have large errors, which may need further optimization or increase related features for correction. If the mean square error (MSE) of the difference between the prediction result and the actual measurement is large or the R squared (R-squared) is not close to 1, the number of trees n_estimators and the maximum depth max_depth of the model are unreasonable, and the parameters are adjusted and steps 2 to 7 are repeated for retraining.
[0230] Optimize the model while testing:
[0231] By adjusting the parameters through grid search and performing cross-validation, the best parameter combination was found. The parameters adjusted by the inventor are: n_estimators = 200 max_depth = 20. Among them, grid search (Grid Search): select the best parameter combination by exhaustively enumerating all possible parameter combinations;
[0232] Cross-Validation: Split the training data into multiple subsets and perform multiple training-testing cycles to evaluate the performance of the model and select the best parameters.
[0233] Among them, the hyperparameter adjustment of the random forest model is an important step to improve the performance of the model. For the random forest model, the two main hyperparameters are the number of trees n_estimators and the maximum depth max_depth. The number of trees n_estimators determines the number of trees in the forest. More trees usually improve the performance of the model, but also increase the computational cost. The maximum depth max_depth controls the maximum depth of the tree. A larger depth allows the model to learn more complex patterns, but it may also lead to overfitting. Here we use grid search to adjust the hyperparameters. Grid search selects the best parameter combination by exhaustively searching all possible combinations of parameters. By adjusting the number of trees n_estimators and the maximum depth max_depth, we can optimize the performance of the random forest model. Using grid search or random search, you can systematically explore different combinations of these parameters to find the best parameter combination to improve the prediction accuracy and generalization ability of the model.
[0234] The classification comparison comprises the steps of:
[0235] 7a. Data grouping. Group the data according to different classification standards to obtain at least three groups of vehicle data such as vehicle model, driving behavior data such as average vehicle speed, and operating environment data such as ambient temperature: vehicle model data group, average vehicle speed data group, and ambient temperature data group.
[0236] in,
[0237] The data element in the vehicle model data group contains three sub-elements: a vehicle model sub-element storing a vehicle model, a component A life actual value element storing a component A life actual value element storing a component A life predicted value element. The vehicle models include: SUV, compact car, and truck. In the vehicle model data group, the data elements whose vehicle models are all SUVs constitute the SUV group, the data elements whose vehicle models are all compact cars constitute the compact car group, and the data elements whose vehicle models are all trucks constitute the truck group. Thus, the vehicle model data group contains three data groups: SUV group, compact car group, and truck group. Among them, the actual vehicle models include sedans, sports cars, pickup trucks, sport utility vehicles (SUV), multi-purpose vehicles (MPV), and trucks. However, considering that the differences in vehicle model structures are small and the vehicle production and assembly processes are similar, sedans and sports cars are combined into compact car models, pickup trucks and trucks are combined into truck models, and SUVs and MPVs are combined into models.
[0238] The data element in the average vehicle speed data group includes three sub-elements: an average vehicle speed sub-element storing the average vehicle speed, a component A life actual value element storing the actual value of component A life, and a component A life prediction value element storing the predicted value of component A life. The average vehicle speed is divided into three gears: low speed, medium speed, and high speed. Low speed refers to an average vehicle speed of 0 to 40 km / h (kilometers per hour), medium speed refers to an average vehicle speed of more than 40 km / h to 70 km / h, and high speed refers to an average vehicle speed of more than 70 km / h. In the average vehicle speed data group, the data elements whose average vehicle speeds are all low speeds constitute a low speed group, the data elements whose average vehicle speeds are all medium speeds constitute a medium speed group, and the data elements whose average vehicle speeds are all high speeds constitute a high speed group, so that the average vehicle speed data group includes three data groups: a low speed group, a medium speed group, and a high speed group.
[0239] The data element in the large environmental temperature data group includes three sub-elements: an environmental temperature sub-element storing the environmental temperature, a component A life actual value element storing the actual value of component A life, and a component A life prediction value element storing the predicted value of component A life. The environmental temperature is divided into three levels: low temperature, medium temperature, and high temperature. Low temperature refers to an environmental temperature of 0-15°C (degrees Celsius), medium temperature refers to an environmental temperature of 15°C to 30°C, and high temperature refers to an environmental temperature of 30°C or above. In the large environmental temperature data group, the data elements whose environmental temperatures are all low temperatures constitute a low temperature group, the data elements whose environmental temperatures are all medium temperatures constitute a medium temperature group, and the data elements whose environmental temperatures are all high temperatures constitute a high temperature group, so that the large environmental temperature data group includes three data groups: a low temperature group, a medium temperature group, and a high temperature group.
[0240] 7b. Calculate the mean square error (MSE) of each data group.
[0241] For each data group, the mean square error MSE is calculated separately:
[0242]
[0243] Wherein, M is the number of data elements in any data group, which is an integer not less than 1, and i2 represents the i2th data element in any data group.
[0244] 7c. Conduct comparative analysis on the data of each major group.
[0245] For large groups of data, such as vehicle type data, average speed data, and ambient temperature data, the MSE of each group of data is compared to find the data group whose MSE exceeds the expected threshold, such as 0.01, and then the corresponding model parameters are optimized. For example, for the vehicle type data group, if the MSE of the SUV group, small car group, and truck group are 0.04, 0.01, and 0.01, respectively, the model parameters corresponding to the SUV group are repeatedly optimized (i.e., adjusting parameters-comparative analysis) until the MSE does not exceed the expected threshold.
[0246] The above provides detailed steps for grouping and comparing the three characteristic data based on vehicle model, average vehicle speed, and ambient temperature. For vehicle data other than vehicle model, driving behavior data other than average vehicle speed, and operating environment data other than ambient temperature, technicians can formulate grouping rules for comparison based on the above grouping and comparison steps and specific operating conditions.
[0247] Step 8: Application of random forest model.
[0248] Use the trained model to make predictions on new feature data.
[0249] Through the above steps, the present invention establishes, trains and verifies a vehicle component life prediction model based on user driving habits, and evaluates and optimizes its performance, so that the model can more accurately predict the life of vehicle components.
[0250] In summary, the present invention proposes a method for predicting the life of automobile parts based on user driving habits. By collecting and analyzing the user's driving behavior data, a dynamic component loss model is constructed, and combined with machine learning and big data analysis technology, it can effectively capture the impact of driving behavior on automobile parts and provide more accurate and practical component life prediction results and maintenance suggestions. This innovative method not only conforms to the trend of intelligent and personalized automobile maintenance, but also helps to improve the reliability, safety and economy of vehicles.
[0251] Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent substitutions for some of the technical features therein; and these modifications or substitutions do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A vehicle component life prediction method based on user driving habits, It is characterized in that Includes steps:
1. Characterization processing of training data; 2. Perform data cleaning on the standardized data obtained through the processing in step 1; 3. Divide the standardized training data obtained after processing in step 2 into a training set and a test set; 4. Feature data division, including maximum and minimum normalization processing and feature standardization of the data in the training set and the test set; 5. Build a random forest model; 6. Using the training set to train the random forest model; 7. Use the test set to test the random forest model; 8. Application of the random forest model.
2. The vehicle component life prediction method based on user driving habits according to claim 1, It is characterized in that The step 1 includes determining the type of the feature data to be extracted, the feature data including vehicle data, driving behavior data, vehicle operating environment data and a combination thereof; The step 1 comprises the steps of: 1a. Standardization of the vehicle data, each piece of the vehicle data includes the frame number, model, year, manufacturer, component name, replacement time, maintenance time, and component material; 1b. Standardization of the driving behavior data, each of which includes the following data: vehicle frame number, user identifier, average vehicle speed, maximum vehicle speed, average engine speed, maximum engine speed, average pedal position, total mileage, average single mileage, number of accelerations, average number of accelerations per kilometer, number of brakes, average number of brakes per kilometer, idling time, average idling time per time, number of speeding, average number of speeding per kilometer, instantaneous fuel consumption, average fuel consumption per 100 kilometers; 1c. Standardization of the vehicle operating environment data, each piece of the vehicle operating environment data includes: frame number, user's city, ambient temperature, ambient humidity, and road conditions.
3. The vehicle component life prediction method based on user driving habits according to claim 2, It is characterized in that The step 2 comprises: 2a. Missing value processing: For any vehicle data A in the vehicle data, if the vehicle frame number is missing, directly delete the vehicle data A; For any driving behavior data B in the driving behavior data, if the driving behavior data lacks the vehicle frame number and the user identifier, directly delete the driving behavior data B; For any piece of vehicle operating environment data C in the vehicle operating environment data, if the vehicle frame number is missing, the vehicle operating environment data C is directly deleted. 2b. Outlier processing: Use KMeans algorithm to filter out abnormal data; 2c. Feature encoding: For non-numeric feature data, one-hot encoding or label encoding is used.
4. The vehicle component life prediction method based on user driving habits according to claim 3, It is characterized in that The step 2b comprises the steps of: 2b1: Feature data preparation: Collect feature data to be processed, including driving behavior data and environmental data, including vehicle frame number, user identifier, average vehicle speed, maximum vehicle speed, average speed, maximum speed, average pedal position, total mileage, average single mileage, number of accelerations, average number of accelerations per kilometer, number of brakes, average number of brakes per kilometer, idling time, average idling time per time, number of speeding, average number of speeding per kilometer, instantaneous fuel consumption, average fuel consumption per 100 kilometers, ambient temperature, ambient humidity, and road conditions; 2b2: Characteristic data standardization: The characteristic data to be processed are standardized so that the values of the characteristic data to be processed are within the same range. The standardization formula used in this step is as follows: In the above formula, z is the value of the feature data to be processed after standardization, x 0 is the original eigenvalue, μ is the mean of the feature data to be processed, and σ is the standard deviation of the feature data to be processed; 2b3: KMeans clustering: Select an appropriate cluster number kkk, and input the standardized feature data to be processed into the KMeans algorithm as the corresponding data point. The KMeans algorithm includes the following steps: 2b3a, Initialization: Randomly select kkk points from the data points as initial cluster centers; 2b3b. Assign the data points: assign each data point to the cluster center closest to it, and calculate the distance from each data point to all cluster centers: In the above formula, x and y are two data points, n is the dimension of the feature, and i represents the dimension of the i-th feature; 2b3c. Update the cluster center: recalculate the center of each cluster, the formula is: In the above formula, C j is the new center of the jth cluster, S j is the set of data points belonging to the jth cluster, x S,j For the data, |S j |For S j The number of data points; Repeat steps 2b2 and 2b3 until the position change of the cluster center does not exceed a preset position threshold or reaches a preset maximum number of iterations; 2b4: Calculate the distance between each data point and the center of the cluster to which it belongs: Among them, C j is the center of the cluster to which the data point x belongs; 2b5: Calculate the mean and standard deviation of the distances from all data points to their cluster centers using the following formula: In the above two formulas, μ d is the mean, σ d is the standard deviation, N is the cluster center C j The data point x in the cluster i1 Total number of; 2b6: Set the threshold H of outliers: H=μ d +k·s d , Where k is a multiple, either 2 or 3: 2b7: Filter abnormal data: Mark the data points exceeding the threshold value H as abnormal data, and the threshold value H satisfies: d(x,C j )>H; 2b8: Manual confirmation and processing: Manually confirm the abnormal data screened out, select a k value, wherein the proportion of the abnormal data is controlled to be 5% or less of the total data, and delete or correct the abnormal data.
5. The vehicle component life prediction method based on user driving habits according to claim 4, It is characterized in that In step 2b2, for non-numerical feature data: vehicle model, year, manufacturer, component name, component material, one-hot encoding or label encoding is used.
6. The vehicle component life prediction method based on user driving habits according to any one of claims 1 to 5, It is characterized in that In step 3, 70% of the standardized training data obtained after processing in step 2 is used as the training set, and 30% of the standardized training data obtained after processing in step 2 is used as the test set.
7. The vehicle component life prediction method based on user driving habits according to claim 6, It is characterized in that In step 4, 4a. The following characteristic data obtained in step 1: average vehicle speed, maximum vehicle speed, average speed, maximum speed, average pedal position, total mileage, average single mileage, number of accelerations, average number of accelerations per kilometer, number of brakes, average number of brakes per kilometer, idling time, average idling time per time, number of speeding, average number of speeding per kilometer, instantaneous fuel consumption, average fuel consumption per 100 kilometers are normalized to the maximum and minimum using the following formula: In the above formula, x 1 is the normalized feature data value, x 0 is the original eigenvalue, x min and x max are respectively the minimum and maximum values of the characteristic data; 4b. Standardization of characteristic data: The characteristic data obtained after processing in step 4a is scaled to a range with a mean of 0 and a standard deviation of 1 using the method in step 2b2.
8. The vehicle component life prediction method based on user driving habits according to claim 7, It is characterized in that In step 5, The initial parameters of the random forest model: the number of trees n_estimators, the maximum depth max_depth: on_estimators = 100, omax_depth=None.
9. The vehicle component life prediction method based on user driving habits according to claim 8, It is characterized in that In step 7, During the test, for the prediction object component A, the difference between the component life in the prediction result obtained by the test and the actual component life is classified and compared according to the three dimensions of vehicle data, driving behavior data, and vehicle operating environment data, and the error mean or standard deviation of each type of data is calculated. If the mean square error MSE of the difference between the prediction result and the actual measurement is large or the R square is not close to 1, the number of trees n_estimators and the maximum depth max_depth of the random forest model are unreasonable, and the parameters are adjusted and steps 2 to 7 are repeated for retraining. The model was optimized while testing: the parameters were adjusted by grid search and cross-validation, and the best parameter combination was found: n_estimators = 200 max_depth = 20, The classification comparison comprises the steps of: 7a. Data grouping: group the data according to different classification standards to obtain at least three groups of data: vehicle type in vehicle data, average vehicle speed in driving behavior data, and ambient temperature in operating environment data: vehicle type data group, average vehicle speed data group, and ambient temperature data group. in, The data element in the vehicle model data group includes three sub-elements: a vehicle model sub-element storing a vehicle model, a component A life actual value element storing an actual value of component A life, and a component A life prediction value element storing a predicted value of component A life. The vehicle models include: SUV, compact car, and truck. The data elements of vehicle models in the vehicle model data group that are all SUVs constitute an SUV group, the data elements of vehicle models that are all compact cars constitute a compact car group, and the data elements of vehicle models that are all trucks constitute a truck group. Thus, the vehicle model data group includes three data groups: an SUV group, a compact car group, and a truck group. The data element in the average vehicle speed data group includes three sub-elements: an average vehicle speed sub-element storing the average vehicle speed, the component A life actual value element storing the actual value of component A life, and the component A life prediction value element storing the predicted value of component A life. The average vehicle speed is divided into three gears: low speed, medium speed, and high speed. Low speed refers to an average vehicle speed of 0 to 40 km / h, medium speed refers to an average vehicle speed of 40 km / h to 70 km / h, and high speed refers to an average vehicle speed of more than 70 km / h. In the average vehicle speed data group, the data elements whose average vehicle speeds are all low speeds constitute a low speed group, the data elements whose average vehicle speeds are all medium speeds constitute a medium speed group, and the data elements whose average vehicle speeds are all high speeds constitute a high speed group. Thus, the average vehicle speed data group includes three data groups: a low speed group, a medium speed group, and a high speed group. The data element in the large environmental temperature data group includes three sub-elements: an environmental temperature sub-element storing the environmental temperature, an actual life value element of component A storing the actual life value of component A, and a predicted life value element of component A storing the predicted life value of component A. The environmental temperature is divided into three levels: low temperature, medium temperature, and high temperature. Low temperature refers to an environmental temperature of 0-15°C, medium temperature refers to an environmental temperature of 15°C to 30°C, and high temperature refers to an environmental temperature of 30°C or above. In the large environmental temperature data group, the data elements whose environmental temperatures are all low temperatures constitute a low temperature group, the data elements whose environmental temperatures are all medium temperatures constitute a medium temperature group, and the data elements whose environmental temperatures are all high temperatures constitute a high temperature group. Thus, the large environmental temperature data group includes three data groups: a low temperature group, a medium temperature group, and a high temperature group. 7b. Calculate the mean square error (MSE) of each data group. For each of the data sets, the mean square error MSE is calculated: Wherein, M is the number of data elements in any of the data groups, which is an integer not less than 1, and i2 represents the i2th data element in any of the data groups; 7c. Compare and analyze the data of each major group. For the large group of vehicle model data, the large group of average vehicle speed data, and the large group of ambient temperature data, the mean square error (MSE) of each of the data groups is compared respectively, and the data group whose mean square error (MSE) exceeds the expected threshold is found, and then the corresponding model parameters are optimized.
10. The vehicle component life prediction method based on user driving habits according to claim 9, It is characterized in that For the vehicle data other than the vehicle model, the driving behavior data other than the average vehicle speed, and the operating environment data other than the ambient temperature, according to the grouping comparison steps 7a to 7c, grouping rules are formulated for comparison based on specific operating conditions.
Citation Information
Cited By
Wireless screen projection remote control method based on random forest optimization algorithm
CN119545075A
Automobile part analysis system and method based on AI visual inspection
CN120852431A
Automobile part analysis system and method based on AI vision detection
CN120852431B
Automobile preventive maintenance intelligent optimization system based on random forest algorithm
CN120875837A
