Heavy-duty truck service life prediction method, device and equipment and storage medium

Abstract

The invention discloses a heavy-duty truck service life prediction method, device and equipment and a storage medium, relates to the technical field of vehicle service life prediction, and discloses a heavy-duty truck service life prediction method, which comprises the following steps: acquiring standard prediction feature information, road condition information, historical track data and historical operation data of a vehicle; determining the vehicle mass according to the standard prediction information and the road condition information; identifying a road working condition proportion according to the historical track data and the historical operation data; and determining a service life prediction result of the vehicle according to the vehicle quality, the road working condition proportion, the standard prediction feature information and a preset random survival forest model. Through obtaining vehicle standard prediction feature information, road condition information, historical track and operation data, calculating vehicle mass, identifying a road working condition proportion, and combining with a preset random survival forest model, accurate prediction of the heavy-duty truck life is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicle life prediction, in particular to a heavy truck life prediction method, device, equipment and storage medium. BACKGROUND

[0002] In recent years, with the rapid development of China's transportation industry, heavy trucks, as the main tool for long-distance transportation, have seen a continuous increase in their number and usage frequency, and the demand for their full life cycle management has also increased. Heavy trucks gradually deteriorate in performance during long-term and high-intensity operation, break down, require maintenance, and even eventually become scrap. The performance deterioration process is influenced by various factors, such as the mechanical performance of the vehicle itself, the use environment, road conditions, maintenance, and driving behavior.

[0003] Traditional technology mainly relies on the driving mileage and service life of the vehicle to make a rough judgment on the life evaluation of the vehicle. Some enterprises may make auxiliary evaluation by combining information such as whether the vehicle is repaired or involved in an accident, but the essence is still the evaluation logic based on static data or human experience, which cannot reflect the real running load and use intensity of the vehicle, and may lead to one-sided or even distorted life evaluation results, lacking scientificity and accuracy, and making it difficult to provide reliable basis for subsequent decision-making.

[0004] The above content is only used to assist in understanding the technical solutions of the present application and does not represent the acknowledgement of the above content as prior art. SUMMARY

[0005] The main purpose of the present application is to provide a heavy truck life prediction method, device, equipment and storage medium, which aims to solve the technical problem of being unable to accurately predict the life of heavy trucks.

[0006] To achieve the above purpose, the present application provides a heavy truck life prediction method, which comprises:

[0007] obtaining standard prediction feature information, road condition information, historical trajectory data and historical operation data of the vehicle;

[0008] determining the vehicle quality according to the standard prediction information and the road condition information;

[0009] identifying the road working condition proportion according to the historical trajectory data and the historical operation data;

[0010] determining the life prediction result of the vehicle according to the vehicle quality, the road working condition proportion, the standard prediction feature information and a preset random survival forest model.

[0011] In an embodiment, the step of determining the life prediction result of the vehicle according to the vehicle mass, the road working condition proportion, the standard prediction characteristic information, and a preset random survival forest model comprises:

[0012] constructing a prediction data set according to the vehicle mass, the road working condition proportion, a historical trip prediction result, and the standard prediction characteristic information, and determining a preset survival split rule according to a preset random survival forest model;

[0013] randomly sampling the prediction data set to generate a prediction data subset;

[0014] performing data processing on the prediction data subset according to the preset survival split rule to obtain a survival tree set;

[0015] determining the cumulative risk of the leaf nodes of the survival trees in the survival tree set, and performing weighted average on the cumulative risks of the survival trees to obtain a target cumulative risk value;

[0016] determining the life prediction result of the vehicle according to the target cumulative risk value.

[0017] In an embodiment, the step of determining the cumulative risk of the leaf nodes of the survival trees in the survival tree set comprises:

[0018] obtaining a prediction time, and determining a leaf node according to a survival tree in the survival tree set, and determining prediction data according to the leaf node;

[0019] determining a first prediction number of vehicles in a discarded state and a second prediction number of vehicles in an effective state in the leaf node according to the prediction data within the prediction time;

[0020] calculating the cumulative risk value of the leaf node according to the first prediction number and the second prediction number.

[0021] In an embodiment, the step of determining the vehicle mass according to the standard prediction information and the road condition information comprises:

[0022] determining an engine output power, a transmission efficiency, an air resistance area, a driving speed, and an acceleration according to the standard prediction information;

[0023] determining a wind resistance coefficient, a rolling resistance coefficient, a slope, and a rotational mass conversion coefficient according to the road condition information;

[0024] determining the vehicle mass according to the engine output power, the transmission efficiency, the air resistance area, the driving speed, the acceleration, the wind resistance coefficient, the rolling resistance coefficient, the slope, and the rotational mass conversion coefficient.

[0025] In one embodiment, the step of identifying the proportion of road conditions based on the number of historical trajectories and the historical operation data includes:

[0026] Based on the historical trajectory data, latitude and longitude information and a preset road network map are determined, and the latitude and longitude information and the preset road network map are matched to obtain a matching result;

[0027] Identify the road type based on the matching results;

[0028] The vehicle's operating time and travel distance on different road types are determined based on the historical operating data.

[0029] The road condition percentage is determined based on the running time, the travel distance, and the road type.

[0030] In one embodiment, the step of acquiring the vehicle's standard predicted feature information, road condition information, historical trajectory data, and historical operation data includes:

[0031] Acquire engine output power, transmission efficiency, drag coefficient, frontal area, driving speed, rolling resistance coefficient, gradient, rotational mass conversion factor, acceleration, and historical trajectory data;

[0032] Initial prediction feature information is determined based on the engine output power, transmission efficiency, frontal area, driving speed, and acceleration.

[0033] Data cleaning is performed based on the initial prediction feature information to obtain standard prediction feature information;

[0034] Road condition information is determined based on the drag coefficient, the rolling resistance coefficient, the slope, and the rotational mass conversion factor.

[0035] The corresponding historical operation data is determined based on the historical trajectory data.

[0036] In one embodiment, the step of cleaning the data based on the initial predicted feature information to obtain standard predicted feature information includes:

[0037] The numerical fields in the initial predicted feature information are checked for missing values.

[0038] When missing values ​​are detected in the numerical field, the numerical field is filled in to obtain complete numerical prediction feature information;

[0039] The classification field in the complete numerical prediction feature information is judged for missing values, and the missing categories are filled when there are missing values ​​in the complete numerical prediction feature information to obtain the complete category prediction feature information.

[0040] The complete category prediction characteristic information is subjected to abnormal data identification, and when the abnormal data is identified in the complete category prediction characteristic information, the abnormal data is deleted from the complete category prediction characteristic information or replaced with a reasonable value to obtain standard prediction characteristic information.

[0041] In addition, to achieve the above-mentioned purpose, the application further provides a heavy haul freight car life prediction device, which comprises a data acquisition module configured to acquire standard prediction characteristic information, road condition information, historical trajectory data and historical operation data of a vehicle.

[0042] A quality determination module is configured to determine the vehicle quality according to the standard prediction information and the road condition information.

[0043] A working condition identification module is configured to identify the road working condition proportion according to the historical trajectory data and the historical operation data.

[0044] A life prediction module is configured to determine the life prediction result of the vehicle according to the vehicle quality, the road working condition proportion, the standard prediction characteristic information and a preset random survival forest model.

[0045] In addition, to achieve the above-mentioned purpose, the application further provides a heavy haul freight car life prediction device, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the computer program is configured to implement the steps of the heavy haul freight car life prediction method as described above.

[0046] In addition, to achieve the above-mentioned purpose, the application further provides a storage medium, which is a computer readable storage medium, and the storage medium stores a computer program, and the computer program is executed by a processor to implement the steps of the heavy haul freight car life prediction method as described above.

[0047] In addition, to achieve the above-mentioned purpose, the application further provides a computer program product, which comprises a computer program, and the computer program is executed by a processor to implement the steps of the heavy haul freight car life prediction method as described above.

[0048] The one or more technical solutions provided by the application have at least the following technical effects:

[0049] The technical means of acquiring standard predicted feature information of the vehicle, road condition information, historical trajectory data and historical operation data, calculating the vehicle mass through the vehicle dynamics formula, identifying the road working condition proportion in combination with the national road network map and the vehicle trajectory data, and inputting these data into the preset random survival forest model to output the life prediction result, solves the problem that the existing technology only relies on the vehicle mileage and the use time to evaluate the residual value of the vehicle, which is subjective and not practical. Compared with the prior art, the multi-dimensional feature construction, accurate vehicle mass and road condition calculation, and the random survival forest model which is good at processing nonlinear relationship are combined to realize the accurate prediction of the life of the heavy truck. BRIEF DESCRIPTION OF DRAWINGS

[0050] The accompanying drawings, which are incorporated herein and form part of the specification, illustrate embodiments consistent with the present application and, together with the description, further serve to explain the principles of the application.

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the accompanying drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, those skilled in the art can obtain other drawings according to these drawings without any creative effort.

[0052] Figure 1 The flowchart provided for the heavy truck life prediction method embodiment one of the present application;

[0053] Figure 2 The flowchart provided for the heavy truck life prediction method embodiment two of the present application;

[0054] Figure 3 The module structure diagram of the heavy truck life prediction device in the embodiment of the present application;

[0055] Figure 4 The device structure diagram of the hardware running environment involved in the heavy truck life prediction method in the embodiment of the present application.

[0056] The purpose implementation, functional characteristics and advantages of the present application will be further explained with reference to the accompanying drawings combined with the embodiments. DETAILED DESCRIPTION

[0057] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application, and are not used to limit the present application.

[0058] In order to better understand the technical solutions of the present application, the following will be described in detail combined with the drawings in the specification and the specific embodiments.

[0059] The main solution of the embodiment of the application is: obtaining standard prediction feature information, road condition information, historical trajectory data and historical running data of a vehicle; determining vehicle quality according to the standard prediction information and the road condition information; identifying road working condition proportion according to the historical trajectory data and the historical running data; and determining a life prediction result of the vehicle according to the vehicle quality, the road working condition proportion, the standard prediction feature information and a preset random survival forest model.

[0060] In the embodiment, for convenience of description, the following describes a heavy load truck life prediction device as an execution subject.

[0061] Since the prior art cannot accurately predict the life of a heavy load truck, the application provides a solution. Since the standard prediction feature information, road condition information, historical trajectory data and historical running data of a vehicle are obtained, the vehicle quality is calculated by a vehicle dynamics formula, the road working condition proportion is identified in combination with a national road network map and vehicle trajectory data, and the data is input into a preset random survival forest model to output a life prediction result, the problem that the residual value of a vehicle is evaluated subjectively and unrealistically only by vehicle mileage and service life in the prior art is solved. Compared with the prior art, the life of a heavy load truck is accurately predicted by multi-dimensional feature construction, accurate vehicle quality and road condition calculation, and a random survival forest model that is good at processing nonlinear relationships.

[0062] It should be noted that the execution subject of the embodiment can be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, a heavy load truck life prediction device, etc. The following describes the embodiment and each of the following embodiments by taking the heavy load truck life prediction device as an example.

[0063] Based on this, the embodiment of the application provides a heavy load truck life prediction method, which refers to Figure 1 , Figure 1 The flowchart of the first embodiment of the heavy load truck life prediction method of the application is shown in the figure.

[0064] In the embodiment, the heavy load truck life prediction method includes steps S10-S40:

[0065] Step S10, obtaining standard prediction feature information, road condition information, historical trajectory data and historical running data of a vehicle;

[0066] It should be noted that the standard prediction feature information refers to a plurality of dimensions of feature sets closely related to the vehicle life, which is used to construct an index system for vehicle life prediction. In this embodiment, 13 dimensions including engine type, horsepower, torque, vehicle age, mileage, maintenance condition, number of accidents, vehicle operation time, number of first and second fault alarms, number of third fault alarms, number of overloads, number of repairs, road conditions, etc. These features reflect the performance, use intensity, maintenance status and fault risk of the vehicle from different angles, and are the core input parameters for subsequent life prediction. For example, the engine type can be divided into diesel engine, natural gas engine, etc., and different types of engines have different durability and wear speed; the maintenance condition can be divided into regular comprehensive maintenance, occasional basic maintenance, almost no maintenance, etc., which directly affects the aging speed of vehicle parts.

[0067] In addition, the road condition information refers to the geographical and engineering characteristics of the road where the vehicle travels, mainly including the slope of the road, the terrain type (plain, mountain, etc.) and the attributes related to the road grade. In this embodiment, the road condition information is used to assist in calculating the vehicle load and identifying the road condition, for example, the longitudinal slope of the highway plain hilly section is about 3%, and the longitudinal slope of the mountain heavy hilly section is about 5%, which will affect the power demand and wear degree of the vehicle during driving.

[0068] In addition, the historical trajectory data refers to the path record generated during the past driving of the vehicle, including the position coordinates of the vehicle at different time points, the driving route and other information. These data are usually collected and stored by the vehicle-mounted T-Box, and can reflect the driving range, frequently traveled routes and other characteristics of the vehicle, which are the basis for identifying the road condition of the vehicle driving. For example, the historical trajectory data of a certain vehicle shows that it often goes back and forth between the city center and the suburbs, passing through several trunk roads and expressways.

[0069] In addition, the historical operation data refers to the real-time state parameters generated during the past operation of the vehicle, mainly from the vehicle CAN bus, including engine output power, driving speed, acceleration, fault alarm signal, operation time, etc. These data record the dynamic running state of the vehicle, for example, in the historical operation data of a certain vehicle, there are 5 times of third fault alarms in the past 3 months, and the average driving speed is 60km / h, which can reflect the wear degree and use intensity of the vehicle.

[0070] Step S20, determining the vehicle mass according to the standard prediction information and the road condition information;

[0071] It should be noted that the vehicle mass refers to the total mass of the vehicle, including the sum of the weight of the vehicle itself and the load. In this embodiment, the vehicle mass cannot be directly obtained and needs to be estimated by vehicle dynamics principles, and the accuracy directly affects the judgment of the vehicle overload condition.

[0072] It can be understood that the engine output power, transmission efficiency, wind resistance coefficient, windward area, driving speed, acceleration and other parameters in the standard prediction characteristic information are utilized, combined with the road slope in the road condition information, and the vehicle mass is estimated through the vehicle dynamics principle. Specifically, based on the power balance relationship when the vehicle is driving, i.e. the effective power output by the engine needs to overcome the wind resistance, rolling resistance, slope resistance and acceleration resistance, the total mass of the vehicle is inversely deduced through the correlation between these forces and mass. For example, when the vehicle is driving at a speed of 60 km / h on a mountainous road section with a slope of 5%, combined with the engine power, wind resistance and other data at this time, the current total mass of the vehicle can be calculated, and it is further judged whether the vehicle is overloaded.

[0073] In a feasible implementation, the step S20 can include steps S21-S23:

[0074] In step S21, the engine output power, transmission efficiency, windward area, driving speed and acceleration are determined according to the standard prediction information.

[0075] It should be noted that the engine output power refers to the work done by the engine per unit time, which is an important indicator for measuring the power performance of the engine, and is usually measured in kilowatts (kW). During the operation of the vehicle, the engine output power will change according to the driving demand (such as acceleration, climbing), which directly affects the power supply of the vehicle. In this embodiment, this parameter can be obtained from the real-time data transmitted by the vehicle CAN bus, which reflects the work capacity of the engine under different working conditions, and is one of the key power parameters for calculating the mass of the vehicle.

[0076] In addition, the transmission efficiency refers to the ratio of the output power to the input power of the vehicle transmission system, which is used to measure the effectiveness of the transmission system in transmitting power, and is usually expressed in percentage. The transmission system includes gearbox, transmission shaft and other components, which will cause power loss due to factors such as friction and wear during power transmission. The higher the transmission efficiency, the smaller the power loss.

[0077] In addition, the windward area refers to the area projected in front of the vehicle, i.e. the vertical projection area occupied by the vehicle in the driving direction, which is usually measured in square meters. The size of the windward area directly affects the air resistance of the vehicle during driving, and the larger the area, the greater the air resistance.

[0078] In addition, the driving speed refers to the distance traveled by the vehicle per unit time, which is usually measured in kilometers per hour (km / h) or meters per second (m / s). The driving speed will change with factors such as road conditions and driving operation, reflecting the real-time motion state of the vehicle. In this embodiment, the driving speed can be collected in real time from the vehicle CAN bus, which is a basic parameter for calculating the power balance of the vehicle.

[0079] In addition, acceleration refers to the rate of change of vehicle speed over time, usually in units of meters per second squared (m / s2), positive value indicates acceleration, negative value indicates deceleration. Acceleration reflects the speed of change of vehicle speed, which is closely related to engine output power, vehicle mass and other factors. In this embodiment, acceleration can be calculated by the driving speed at different times collected through the vehicle CAN bus, that is, the difference between the speed at a certain time and the speed at the previous time divided by the time interval, which is an important parameter to measure the dynamic response of the vehicle.

[0080] Step S22, determining the wind resistance coefficient, rolling resistance coefficient, slope, and rotational mass conversion coefficient according to the road condition information;

[0081] It should be noted that the wind resistance coefficient is a dimensionless parameter that measures the air resistance of the vehicle moving in the air, the smaller the value, the smaller the air resistance of the vehicle. The wind resistance coefficient is closely related to the design of the vehicle body, for example, the wind resistance coefficient of a streamlined body is small. In this embodiment, the wind resistance coefficient is a inherent parameter of the vehicle, which does not change with the road condition by default, and can be determined according to the vehicle design standard corresponding to the road condition information, which is a core parameter for calculating air resistance.

[0082] In addition, the rolling resistance coefficient is a dimensionless parameter that measures the rolling resistance between the tire and the road surface when the vehicle is driving, which is affected by the road surface condition (such as asphalt road, cement road, gravel road), the rougher the road surface, the larger the rolling resistance coefficient. In this embodiment, the rolling resistance coefficient needs to be determined according to the road type and road surface condition in the road condition information, for example, the rolling resistance coefficient of asphalt road is usually smaller than that of gravel road, which is a key parameter for calculating rolling resistance.

[0083] In addition, the slope refers to the degree of inclination of the road in the longitudinal direction, usually expressed as a percentage, that is, the ratio of the elevation difference between the two ends of the road section to the horizontal distance. The slope is divided into uphill (positive value) and downhill (negative value), which directly affects the slope resistance when the vehicle is driving. In this embodiment, the slope is determined according to the road grade and terrain type in the road condition information, for example, according to the highway route design specification, the slope of the plain hilly section of the expressway is 3%, and the slope of the mountainous hilly section is 5%, which is an important parameter for calculating the slope resistance.

[0084] In addition, the rotational mass conversion coefficient is a coefficient for converting the inertial mass of the rotating parts of the vehicle (such as wheels, transmission shafts) into translational mass, which is used to comprehensively measure the influence of the rotating parts on the acceleration performance of the vehicle, and its value is slightly larger than 1. In this embodiment, the rotational mass conversion coefficient is a inherent parameter of the vehicle, which does not change with the road condition by default, and can be determined according to the structural design of the vehicle, which is a parameter that needs to be considered when calculating the acceleration resistance.

[0085] Step S23, determining the vehicle mass according to the engine output power, the transmission efficiency, the windward area, the running speed, the acceleration, the wind resistance coefficient, the rolling resistance coefficient, the slope, and the rotational mass conversion coefficient.

[0086] It can be understood that the vehicle mass is calculated based on the power balance principle when the vehicle is running. When the vehicle is running, the effective power output by the engine needs to overcome the air resistance, the rolling resistance, the slope resistance and the acceleration resistance. The air resistance is related to the third power of the wind resistance coefficient, the windward area and the running speed; the rolling resistance is related to the rolling resistance coefficient, the vehicle mass, the gravitational acceleration and the horizontal component of the road inclination; the slope resistance is related to the vehicle mass, the gravitational acceleration and the slope; the acceleration resistance is related to the rotational mass conversion coefficient, the vehicle mass and the acceleration. By establishing the balance relationship between these resistances and the power, the vehicle mass is inversely deduced. The vehicle mass calculation formula is as follows:

[0087]

[0088] In the formula, m is the vehicle mass; P is the engine output power; η is the transmission efficiency; C is the wind resistance coefficient; A is the windward area; v is the running speed; g is the gravitational acceleration; f is the rolling resistance coefficient; i is the slope; δ is the rotational mass conversion coefficient; and dv / dt is the acceleration. e D

[0089] Step S30, identifying the road working condition proportion according to the historical trajectory number and the historical running data;

[0090] It should be noted that the road working condition refers to the road type of the vehicle running, which is mainly divided into four types in the embodiment: national road, urban road, expressway and other roads (including county road, township road and unknown road). Different road working conditions have different degrees of vehicle wear, for example, the vehicle wear is small when running on the expressway; the engine and brake system wear are large when running on the urban road.

[0091] Further, the road working condition proportion refers to the proportion of the mileage (or time) of the vehicle running in each type of road working condition to the total running mileage (or total running time). For example, the total running mileage of a vehicle is 100,000 kilometers, of which 30,000 kilometers are on the expressway, 50,000 kilometers are on the urban road, 15,000 kilometers are on the national road, and 5,000 kilometers are on the other road, then the expressway working condition proportion is 30%, the urban working condition proportion is 50%, the national road working condition proportion is 15%, and the other working condition proportion is 5%.

[0092] ​​It can be understood that the historical trajectory data is processed to extract the path information of the vehicle driving; at the same time, the national road network OSM map data indexed by GeoHash is called from the database, and the vehicle historical trajectory is matched with the road type in the map through an algorithm to preliminarily identify the road working conditions of the vehicle driving (such as matching to a certain section of trajectory corresponding to the highway network, which is marked as a highway working condition). Then, the matching result is verified with the aid of historical operation data. For example, the vehicle driving speed under the highway working condition is usually 60-120km / h, and the start-stop is less; the speed under the urban working condition is usually 0-60km / h, and the start-stop is frequent. Finally, the driving mileage (or time) of the vehicle under various road working conditions is counted, and the proportion of each working condition in the total mileage (or total time) is calculated, that is, the road working condition proportion is obtained.

[0093] In a feasible implementation, step S30 can include steps S31-S34:

[0094] Step S31, determining the latitude and longitude information and the preset road network map according to the historical trajectory data, and matching according to the latitude and longitude information and the preset road network map to obtain a matching result;

[0095] It should be noted that the historical trajectory data is the path record generated by the vehicle during the past driving process, which contains the position information of the vehicle at different time points. These information is usually collected and stored by the vehicle-mounted T-Box. It records in detail the route driven by the vehicle, and is the basic data for tracing the driving path of the vehicle, which can reflect the driving range and the frequently traveled route of the vehicle. For example, the historical trajectory data of a heavy truck within a week will contain the specific driving path from the factory to the logistics center every day, and the position record of each road section along the way.

[0096] Further, the latitude and longitude information is the geographical coordinates used to represent the position of a point on the earth's surface. The longitude is the degree of a place on the earth from the prime meridian in the north-south direction, and the latitude is the degree of the place from the equator in the north-south direction. In this embodiment, the latitude and longitude information is extracted from the historical trajectory data, which is the key to determining the specific position of the vehicle, and can accurately mark the geographical coordinates of each point passed by the vehicle. For example, the latitude and longitude information at a certain time in the historical trajectory data may be 116.4° east longitude and 39.9° north latitude, indicating that the vehicle is located at this geographical coordinate at that time.

[0097] In addition, the preset road network map is road network data collected and stored in advance nationwide. In this embodiment, the road network OSM map nationwide is collected from the Internet, and the map data is stored in the database after GeoHash indexing. The OSM map, i.e., the open street map, contains detailed road types, directions, connection relationships and other information, and can provide an accurate road network framework for matching the vehicle position and the road. For example, the preset road network map clearly marks the specific direction and distribution of the expressway, national road and urban road in a certain area.

[0098] Further, the matching result refers to the result obtained by corresponding and associating the latitude and longitude information in the historical trajectory data with the road in the preset road network map, which reflects which specific road the latitude and longitude point of the vehicle travels on. For example, if a latitude and longitude information is matched to a certain section of the Beijing-Hong Kong-Macao Expressway in the preset road network map, the matching result is that the latitude and longitude point belongs to the Beijing-Hong Kong-Macao Expressway.

[0099] It can be understood that the latitude and longitude information of each time point in the vehicle driving process is extracted from the historical trajectory data, which accurately marks the geographical position of the vehicle at different times. At the same time, the preset road network map, i.e., the OSM map of the national road network after GeoHash indexing, is retrieved from the database. Then, the extracted latitude and longitude information is matched with the road in the preset road network map through a specific algorithm. The algorithm will find the road most suitable for the latitude and longitude point according to the distribution of the latitude and longitude point, the distance and direction of the adjacent points, and the direction and position of the road in the map. For example, if a series of latitude and longitude points show a continuous distribution along the direction of a certain expressway, and are highly consistent with the coordinate range of the expressway in the preset road network map, the matching result is that the road corresponding to these latitude and longitude points is the expressway. Through this process, the historical driving trajectory of the vehicle can be associated with a specific road.

[0100] In step S32, the road type is identified according to the matching result.

[0101] It should be noted that the road type is a classification of the road, which is mainly divided into four categories in this embodiment: national road, urban area, expressway, and other. Among them, the expressway is a trunk road with special technical standards, which is exclusively for car driving and is fully controlled; the national road is the main trunk road with national political and economic significance; the urban road is the road located in the city, which usually includes main road, secondary road, branch road, etc.; the other road includes county road, township road, and unknown road. Different road types have differences in road conditions, driving rules, and impact on vehicle wear, etc. For example, the expressway is smooth and has relatively small wear on the vehicle, while the urban road has more red and green lights and frequent start and stop, which has greater wear on the vehicle engine and brake system.

[0102] It can be understood that based on the matching result, in combination with the road attribute information recorded in the preset road network map, it is identified which type of road the vehicle has traveled on. Each road in the preset road network map is marked with its corresponding type, for example, a certain road is marked as an expressway in the map, and another road is marked as a national road. When the latitude and longitude information matches a certain road, the corresponding road type can be determined according to the type marking of the road in the map. For example, if the matching result shows that the latitude and longitude information of the vehicle corresponds to the road marked as an urban main road in the preset road network map, the type of the road is identified as urban; if it matches the road marked as G107 national road, the type of the road is identified as national road. Through this process, the corresponding road type of the vehicle when driving on different road sections can be determined.

[0103] In step S33, the running time and driving distance of the vehicle on different road types are determined according to the historical running data;

[0104] It should be noted that the running time refers to the time spent by the vehicle on a certain type of road, usually in units of hours or minutes. It is calculated according to the starting time and ending time of the vehicle on the road type, and can reflect the driving time of the vehicle on the road type. For example, the vehicle drives on the expressway from 8 am to 10 am, and the running time on the expressway is 2 hours.

[0105] In addition, the driving distance refers to the length of the road traveled by the vehicle on a certain type of road, usually in units of kilometers. It can be calculated by the driving speed and running time of the vehicle on the road type, or by the change of latitude and longitude information in the historical trajectory data. For example, the vehicle drives on the urban road at a speed of 30 km / h for 1 hour, and the driving distance on the urban road is 30 kilometers.

[0106] It can be understood that the driving information of the vehicle on each type of road is extracted from the historical operation data in combination with the road type, and then the operation time and the driving distance are determined. According to the time record in the historical operation data, the starting time when the vehicle enters a certain type of road and the ending time when the vehicle leaves the road type are determined, and the difference between the two is the operation time on the road type. For example, the vehicle enters the national road at 9 o'clock in the morning and leaves the national road at 11 o'clock in the morning, and the operation time is 2 hours. Secondly, according to the driving speed in the historical operation data, the driving distance is calculated in combination with the operation time on the road type; or the actual driving distance is calculated through the change of the latitude and longitude information corresponding to the road type in the historical trajectory data. For example, the average driving speed of the vehicle on the national road is 60 km / h, and the operation time is 2 hours, so the driving distance is 60*2=120 kilometers. Through this process, the operation time and the driving distance of the vehicle on different road types such as national roads, urban areas, highways, and others can be obtained.

[0107] In step S34, the road working condition proportion is determined according to the operation time, the driving distance, and the road type.

[0108] It can be understood that based on the operation time and the driving distance corresponding to different road types, in combination with the total operation time and the total driving distance of the vehicle, the proportion of each type of road working condition is calculated. First, the total operation time (i.e. the sum of the operation time of all road types) and the total driving distance (i.e. the sum of the driving distance of all road types) of the vehicle are counted. Then, for each type of road, the operation time of the type is divided by the total operation time to obtain the road working condition proportion based on time; or the driving distance of the type is divided by the total driving distance to obtain the road working condition proportion based on distance. For example, the driving distances of the vehicle on the highway, the urban area, the national road, and the other road type are 300 kilometers, 500 kilometers, 150 kilometers, and 50 kilometers respectively, and the total driving distance is 1000 kilometers, so the proportion of the highway working condition is 300 / 1000*100%=30%, the proportion of the urban working condition is 500 / 1000*100%=50%, the proportion of the national road working condition is 15%, and the proportion of the other working condition is 5%. Through this process, the specific proportion of each type of road working condition can be obtained.

[0109] In step S40, the life prediction result of the vehicle is determined according to the vehicle mass, the road working condition proportion, the standard prediction feature information, and a preset random survival forest model.

[0110] It should be noted that the preset random survival forest model refers to a pre-trained random survival forest (RSF) model. The model is improved based on the random forest algorithm and focuses on survival analysis, can process time-related event (such as vehicle scrapping) prediction, and outputs the time of event occurrence and risk probability through ensemble learning of multiple survival trees. In this embodiment, the model has been trained through a large amount of historical data of vehicles and has the ability to predict the life of new input features.

[0111] Further, the life prediction result refers to the survival time of the vehicle and the scrapping risk probability predicted by the model. The survival time is the total driving mileage or service life of the vehicle from the current time to the scrapping time; the scrapping risk probability is the instantaneous probability of the vehicle being scrapped at a certain time. For example, the prediction result can be that the remaining survival time of the vehicle is 80,000 kilometers, and the scrapping risk probability in the next 2 years is 15%.

[0112] It can be understood that the vehicle quality, the road working condition proportion, and the standard prediction feature information are input into the preset random survival forest model as input features. The model first processes these features, obtains the cumulative risk function of each tree through the splitting calculation of multiple survival trees, then averages the cumulative risk functions of all survival trees to obtain the overall risk function and survival function. Based on the survival function, the survival time of the vehicle can be determined, and based on the risk function, the scrapping risk probability at different time points can be determined, and finally the life prediction result of the vehicle is output.

[0113] In a possible implementation, the step S40 can include steps S41-S45:

[0114] In step S41, a prediction data set is constructed according to the vehicle quality, the road working condition proportion, the historical trip prediction result, and the standard prediction feature information, and a preset survival splitting rule is determined according to a preset random survival forest model.

[0115] It should be noted that the prediction data set is a data set for input into the prediction model and is composed of multiple feature variables. In this embodiment, it includes vehicle quality (reflecting the influence of load and overload), road working condition proportion (reflecting the driving proportion of different road conditions), and standard prediction feature information (such as engine type, horsepower, fault alarm number, etc. 13-dimensional features). These features together constitute comprehensive data describing the state and use of the vehicle and are the basis for training and running the random survival forest model. For example, the prediction data set of a heavy-duty truck can include: vehicle quality 25 tons, high-speed working condition proportion 30%, engine type diesel engine, vehicle age 5 years, driving mileage 300,000 kilometers, etc. In addition, the historical trip prediction result represents the prediction result of different trips of the vehicle.

[0116] Furthermore, the preset survival splitting rule refers to the criteria used to partition nodes when constructing the survival tree. In this embodiment, the log-rank survival splitting criterion is used. This criterion compares the differences in the occurrence of events (such as vehicle scrapping) under different feature partitions and selects the feature that can best distinguish the timing of event occurrence as the splitting basis to ensure that the child nodes after splitting have significant differences in event occurrence patterns.

[0117] Understandably, the prediction dataset is formed by integrating all features from vehicle quality, road condition ratios, and standard prediction feature information, combined with historical trip prediction results. Each data sample contains specific values ​​for all the aforementioned features. The dataset representation is shown below:

[0118] D = {(X1,y1),(X2,y2),…,(X...} N ,y N )},X=(x1,x2,…,x M )

[0119] In the formula, D is the dataset, x1, x2, ..., x M Let X be the set of eigenvectors formed by vehicle mass and road condition ratio, where y1, y2, ..., y3 are the features. N These are the historical itinerary prediction results for different trips.

[0120] Based on the characteristics of the pre-defined random survival forest model, a pre-defined survival splitting rule, namely the log-rank survival splitting criterion, is determined for constructing the survival tree. For example, the prediction dataset contains feature data of 1,000 heavy-duty trucks, and the records for each truck include features such as vehicle weight, percentage of high-speed operation, and engine horsepower. At the same time, the log-rank criterion is determined to be used as the rule for subsequent splitting of survival tree nodes to ensure that node splitting can effectively distinguish the differences in vehicle scrapping time.

[0121] Step S42: Randomly sample the predicted dataset to generate a subset of predicted data;

[0122] It should be noted that random sampling refers to the process of drawing a subset of samples from the prediction dataset. In this embodiment, the Bootstrap resampling method is used, which involves drawing samples from the original dataset with replacement, and the number of samples drawn is the same as that of the original dataset. This sampling method results in some samples being drawn multiple times, while some samples may not be drawn at all, thereby generating subsets that are similar to but not exactly the same as the original dataset distribution, which are used to construct different survival trees.

[0123] Further, the prediction data subset is a subset of data obtained from the prediction data set by random sampling, and the number of samples is consistent with the prediction data set, but the specific samples contained may be repeated or different from the original data set. Each prediction data subset is used to independently build a survival tree, and multiple subsets correspond to multiple survival trees, and the final prediction stability is improved by integrating the results of multiple trees. For example, 1000 samples are randomly sampled with replacement from a prediction data set containing 1000 samples to obtain a prediction data subset.

[0124] It can be understood that the Bootstrap resampling method is used to randomly sample the prediction data set: a sample is randomly selected from the prediction data set, the sample is put back into the data set after being recorded, and the process is repeated until the number of samples extracted is the same as the number of samples in the prediction data set, to obtain a prediction data subset. In order to build multiple survival trees, the above sampling process needs to be repeated to generate multiple prediction data subsets, each of which corresponds to the training data of a survival tree. For example, a prediction data set contains 500 samples, and 100 prediction data subsets are generated by sampling with replacement, each of which contains 500 samples.

[0125] Step S43, data processing is performed on the prediction data subset according to the preset survival split rule to obtain a survival tree set;

[0126] It can be understood that the data processing here refers to the process of splitting nodes based on the preset survival split rule (log-rank criterion) on the prediction data subset to build a survival tree. Specifically, it includes randomly selecting part of the features from the features, using the split rule to determine the best split point, dividing the current node into two child nodes, and repeating the above process for each child node until the stop condition (such as the node sample size reaching the minimum value) is met.

[0127] It should be noted that the survival tree is a tree model for survival analysis, which is composed of a root node, internal nodes and leaf nodes. The root node contains all samples, the internal node represents a split condition based on a certain feature, and the leaf node contains samples with similar survival characteristics, and each leaf node corresponds to a cumulative risk function for describing the risk of events (such as vehicle scrapping) of the sample group.

[0128] Further, the survival tree set is a set composed of multiple independently built survival trees, each of which is built based on a different prediction data subset, and the overall prediction performance is improved by integrating the prediction results of multiple trees. For example, a survival tree set composed of 100 survival trees, each of which has a certain difference in predicting the life of a vehicle, and a more reliable result can be obtained by integrating these differences.

[0129] It can be understood that for each prediction data subset, the survival tree is constructed according to the following process: the prediction data subset is input into the root node of the survival tree; a part of features is randomly selected from all features, the splitting effect of each feature is calculated based on a preset log-rank survival splitting rule, the feature and the splitting point that can most distinguish the survival time of the samples are selected, and the root node is divided into two sub-nodes; the above process of randomly selecting features and splitting according to the splitting rule is repeated for each sub-node until the samples in the node cannot be split (for example, the sample size is less than a preset threshold), a leaf node is formed, and the construction of a single survival tree is completed. The above process is repeated to construct a survival tree for each prediction data subset, and all survival trees together form a survival tree set.

[0130] In step S44, the cumulative risk of the leaf node of the survival tree in the survival tree set is determined, and the cumulative risks of the survival trees are weighted and averaged to obtain a target cumulative risk value.

[0131] It should be noted that the leaf node is the final node in the survival tree that cannot be split, and each leaf node contains a group of samples with similar survival characteristics. In this embodiment, the samples in the leaf node have the same cumulative risk function, that is, the risk of occurrence of events (vehicle scrapping) of all samples in the node changes consistently over time. For example, a leaf node contains 10 heavy-duty truck samples, and these samples have similar characteristics such as age and number of failures, and the trend of their scrapping risk over time is the same.

[0132] Further, the cumulative risk refers to the cumulative risk value of the occurrence of an event (such as vehicle scrapping) from the observation start to a time t, which is represented by a cumulative risk function (Cumulative Hazard Function, CHF). For the leaf node of the survival tree, the cumulative risk is calculated by the event occurrence time of the samples in the node: the ratio of the number of samples scrapped at each time point to the number of samples not scrapped at the time point is calculated, and the ratios are accumulated over time to obtain the cumulative risk of the leaf node at time t. For example, 2 trucks in a leaf node are scrapped in the 5th year, and at this time, there are 18 samples that have not been scrapped, so the risk contribution at this time point is 2 / 18. The risk contributions at all time points are accumulated to obtain the cumulative risk of the node.

[0133] In addition, the target cumulative risk value is a comprehensive risk value obtained by weighted averaging the cumulative risks of all survival trees in the survival tree set. In this embodiment, the weighted average uses arithmetic average (i.e., the same weight), and the cumulative risks of each survival tree are added and divided by the number of survival trees to obtain the final target cumulative risk value, which is used to reflect the overall cumulative scrapping risk of the vehicle.

[0134] It can be understood that the cumulative risk of each leaf node in each survival tree is calculated: for each survival tree, all leaf nodes thereof are traversed, for each leaf node, the failure time points of all samples in the node are counted, at each time point t, the ratio of the number of samples that have failed in the node to the number of samples that have not failed at the time point t is calculated, and the ratios from the initial time to t are accumulated to obtain the cumulative risk of the leaf node at time t. Subsequently, for the vehicle sample to be predicted, the leaf node to which it belongs in each survival tree is determined, and the cumulative risk of the leaf node is extracted as the cumulative risk prediction of the sample by the tree. The cumulative risks of all survival trees are arithmetically averaged (i.e., weighted averaging, with weights being 1 / survival tree number), to obtain the target cumulative risk value of the vehicle sample. For example, the cumulative risks of a certain vehicle in 100 survival trees are H1(t), H2(t), …, H100(t), and the target cumulative risk value is (H1(t)+H2(t)+…+H100(t)) / 100.

[0135] In a feasible implementation, step S44 can include steps S441-S443:

[0136] Step S441, obtaining a prediction time, determining a leaf node according to a survival tree in the set of survival trees, and determining prediction data according to the leaf node;

[0137] It should be noted that the prediction time refers to a specific time point or time period at which the vehicle failure risk needs to be evaluated, and is usually expressed in terms of years of use or total mileage. In this embodiment, the prediction time is the time reference for calculating the cumulative risk.

[0138] It can be understood that the prediction data refers to the vehicle sample data contained in the leaf node of the survival tree, which includes the feature information of the sample (such as vehicle mass, road working condition proportion, etc.) and the corresponding survival state (whether to be scrapped at a certain time point), and is the basis for calculating the cumulative risk of the leaf node. For example, the prediction data of a certain leaf node can include samples of 20 heavy-duty trucks, recording the use years of each vehicle, whether it has been scrapped, etc.

[0139] Step S442, determining a first prediction number of vehicles in the leaf node in the elimination state and a second prediction number of vehicles in the leaf node in the effective state according to the prediction data within the prediction time;

[0140] It should be noted that the elimination state refers to a state in which the vehicle has failed within the prediction time, i.e., the vehicle has stopped being used due to reaching the service life, serious damage, etc. In this embodiment, the elimination state is a symbol for judging whether the vehicle has reached the end event, and directly reflects the failure of the vehicle within the prediction time. For example, a certain vehicle has been scrapped within the prediction time (6 years), and the vehicle is in the elimination state.

[0141] In addition, the effective state refers to a state in which the vehicle does not occur to be scrapped within the prediction time, i.e., the vehicle is still in normal use and does not reach the elimination standard. In the embodiment, the effective state is relative to the elimination state, and is used to describe the survival of the vehicle within the prediction time.

[0142] Further, the first prediction number refers to the number of vehicles in the elimination state in the prediction data of the leaf node within the prediction time, i.e., the total number of vehicles scrapped within the prediction time in the prediction data. For example, the prediction data of a certain leaf node contains 20 vehicles, of which 5 are scrapped within 6 years, and the first prediction number is 5.

[0143] Further, the second sample number refers to the number of vehicles in the effective state in the prediction data of the leaf node within the prediction time, i.e., the total number of vehicles not scrapped within the prediction time in the prediction data. For example, of the 20 vehicles of the above-mentioned leaf node, 15 are not scrapped within 6 years, and the second sample number is 15.

[0144] It can be understood that the state of the vehicle in the leaf node is counted based on the determined prediction time and the prediction data of the leaf node. Each vehicle in the prediction data is traversed to determine whether it is scrapped within the prediction time: if the scrapping time (such as the time when the total mileage reaches the scrapping standard) of the vehicle is less than or equal to the prediction time, it is determined to be in the elimination state; if the scrapping time of the vehicle is greater than the prediction time (or not scrapped), it is determined to be in the effective state. The total number of vehicles in the elimination state is counted to obtain the first prediction number, and the total number of vehicles in the effective state is counted to obtain the second sample number. For example, the prediction time is 6 years, the prediction data of a certain leaf node contains 30 vehicles, of which 10 are scrapped within 6 years and 20 are still in use within 6 years, the first prediction number is 10, and the second sample number is 20.

[0145] Step S443, calculating the cumulative risk value of the leaf node according to the first prediction number and the second prediction number.

[0146] The cumulative risk value calculation formula is as follows:

[0147]

[0148] In the formula, d i，h is the number of samples in the node that occur to default at t i,h , i.e., the first prediction number; Y i,h is the number of samples in the node that do not occur to default until t i,h , i.e., the second prediction number; t i，h is the discrete default time point in the leaf node h, i.e., the determined default time point within the prediction time; H h(t) is the cumulative risk value.

[0149] Further, according to the cumulative risk value calculation formula, the samples in the same leaf node have the same CHF, and the CHF of all survival trees is averaged to obtain the CHF value of the entire random survival tree. The target cumulative risk value calculation formula is as follows:

[0150]

[0151] In the formula, H h (t) is the cumulative risk value, and L is the total number of cumulative risk values. Further, in order to quantify the performance of the prediction model, the present application selects two evaluation indexes, namely consistency index (C-index) and Brier score (BS).

[0152] In step S45, the life prediction result of the vehicle is determined according to the target cumulative risk value.

[0153] It can be understood that the life prediction result refers to the vehicle survival time and the scrap risk probability derived by the target cumulative risk value. The survival time refers to the total driving mileage or service years of the vehicle from the current time to the scrapping time; and the scrap risk probability refers to the instantaneous probability of the vehicle being scrapped at a certain time. In the embodiment, the target cumulative risk value is related to the survival function and the risk function: the survival function S(t) (the probability of the vehicle not being scrapped after t time) can be expressed as S(t) = exp(-H(t)) by the target cumulative risk value H(t); and the risk function h(t) (the instantaneous scrap probability at t time) is the derivative of the target cumulative risk value with respect to time. Based on these relationships, the survival time (such as the median survival time corresponding to S(t) = 0.5) and the scrap risk probability at different time points of the vehicle can be determined. For example, the target cumulative risk value shows that S(t) = 0.3 when t = 8 years, which means that the probability of the vehicle not being scrapped after 8 years is 30%, and the instantaneous scrap risk probability at this time can be calculated by the risk function.

[0154] The embodiment provides a heavy truck life prediction method, which acquires standard prediction feature information, road condition information, historical trajectory data and historical operation data of the vehicle, and solves the technical problem of insufficient vehicle life prediction accuracy by combining vehicle dynamics principles and random survival forest model and other technical means, so that the beneficial effects of accurately predicting the vehicle life and early predicting the vehicle scrap risk are achieved.

[0155] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as the above embodiment one can refer to the above introduction, and will not be described hereinafter. On this basis, please refer to Figure 2The step S10 of the heavy truck life prediction method comprises steps S11-S15:

[0156] In step S11, engine output power, transmission efficiency, wind resistance coefficient, windward area, driving speed, rolling resistance coefficient, slope, rotational mass conversion coefficient, acceleration, and historical trajectory data are acquired.

[0157] It can be understood that the above parameters need to be collected from multiple channels. Among them, the engine output power, driving speed and acceleration are acquired in real time through the vehicle CAN bus, which directly reflects the real-time power and motion state of the vehicle; the transmission efficiency, wind resistance coefficient, windward area and rotational mass conversion coefficient are inherent parameters of the vehicle, which are obtained from the vehicle design data or factory parameters; the rolling resistance coefficient and slope are determined according to the road type and terrain information, for example, extracted from the national highway route design specification or elevation information; the historical trajectory data is retrieved from the path record stored in the vehicle T-Box, which contains the latitude and longitude and time information of the past driving of the vehicle. By comprehensively collecting these parameters, raw data is provided for subsequent steps of building prediction features and determining road conditions.

[0158] In step S12, initial prediction feature information is determined based on the engine output power, transmission efficiency, windward area, driving speed and acceleration.

[0159] It should be noted that the initial prediction feature information refers to a feature set related to the vehicle life that is initially constructed based on parameters such as engine output power, transmission efficiency, windward area, driving speed and acceleration. These features have not been cleaned and may contain missing values or outliers, but can initially reflect the key attributes of the vehicle's power performance, motion state and other factors affecting the life.

[0160] It can be understood that the collected parameters such as engine output power, transmission efficiency, windward area, driving speed and acceleration are arranged and derived to form the initial prediction feature information. Specifically, the statistical quantities (such as mean, maximum value, standard deviation) of these parameters are calculated, or new features (such as engine output power change rate per unit time, product of speed and acceleration, etc.) are obtained by combination, which are related to the power loss and use intensity of the vehicle and are the basis for building standard prediction feature information.

[0161] In step S13, data cleaning is performed based on the initial prediction feature information to obtain standard prediction feature information.

[0162] It should be noted that data cleaning refers to processing missing values and abnormal values in initial prediction characteristic information to improve data quality. In the embodiment, data cleaning includes two parts: one is to process missing values, that is, to supplement the unrecorded numerical values in the data; the other is to process abnormal values, that is, to identify and correct or reject the data obviously deviating from the normal range. For example, if there are 10% missing values in a certain characteristic, they need to be filled in by a reasonable method; if the driving speed of a certain sample is 300 km / h (far exceeding the actual speed of a heavy truck), it needs to be determined as an abnormal value and processed.

[0163] It can be understood that the initial prediction characteristic information is processed in two aspects: one is to process missing values, for numerical characteristic, the missing values are filled with the median, mean or mode of the characteristic; for classification characteristics, the missing values are filled or treated as a separate category. The other is to process abnormal values, which are determined by the 3σ criterion, that is, when the absolute value of the difference between a data point and the mean of the characteristic is greater than 3 times the standard deviation, it is determined as an abnormal value, and then processed by correcting or rejecting the abnormal sample. After the above processing, the standard prediction characteristic information without missing values and abnormal values is obtained. For example, there are 5 missing values in the initial prediction characteristic information of engine output power, which are filled with the median 180 kW; 2 abnormal values (500 kW, far exceeding the normal range) are identified and replaced with 300 kW (95% quantile) to form the standard prediction characteristic information.

[0164] In a feasible implementation, step S13 can include steps S131-S134:

[0165] Step S131, detecting whether there are missing values in the numerical field in the initial prediction characteristic information;

[0166] It should be noted that the numerical field refers to the characteristics in the initial prediction characteristic information represented in numerical form, which can be mathematically operated and directly reflects the quantitative properties of the vehicle. In the embodiment, the numerical field includes driving mileage, horsepower, torque, vehicle age, first and second fault alarm times, third fault alarm times, overload times, maintenance times, vehicle operation time, etc. For example, the driving mileage is 300,000 km and the horsepower is 350, which can directly reflect the use intensity and performance parameters of the vehicle.

[0167] In addition, the missing value is the data that is not recorded or not collected in the numerical field, that is, the field has no specific numerical value in some samples. The missing value may be caused by data collection equipment failure, loss in transmission process, etc. For example, the maintenance times field of a certain vehicle does not show the specific value due to the omission of the after-sales service system record, which is a missing value.

[0168] It can be understood that all numerical fields in the initial prediction feature information are traversed to check whether there is no record for each sample on these fields. Specifically, by scanning the data records, entries marked as null or not filled in the numerical fields are identified to determine which numerical fields of which samples have missing values.

[0169] Step S132, when it is detected that the numerical field has missing values, the numerical field is filled to obtain complete numerical prediction feature information;

[0170] It can be understood that filling refers to the process of supplementing missing values in the numerical field to obtain reasonable numerical values. In this embodiment, the filling method is selected according to the characteristics of the numerical field, aiming to make the supplemented numerical value as close to the true value as possible and reduce the impact on the data distribution.

[0171] Further, the complete numerical prediction feature information refers to a feature set in which all numerical fields have no missing values after filling the missing values. At this time, each sample of the numerical field has a specific numerical value, and the data integrity is guaranteed. For example, after filling the originally missing "mileage" field, all samples have corresponding mileage values, forming complete numerical features.

[0172] It can be understood that suitable methods are used to fill the missing values of the numerical field. For different numerical fields, the median, mean or mode can be selected for filling: if the field data distribution is relatively uniform and has no extreme value, the mean can be used for filling, for example, the horsepower field; if the field has extreme values, the median is more suitable for filling, for example, the mileage (there may be a few vehicles with mileage far exceeding most samples); if a certain value in the field appears extremely high frequency, the mode can be used for filling, for example, the monthly average operation time of most vehicles is 200 hours, so 200 hours is used to fill the missing values. After filling, all numerical fields have no missing values, and complete numerical prediction feature information is obtained. For example, for the 5 missing values of the mileage, the median of the mileage of all samples, 280,000 kilometers, is used to fill the missing values, so that the field no longer has missing values.

[0173] Step S133, judging the missing values in the complete numerical prediction feature information, and filling the missing categories when the complete numerical prediction feature information has missing values to obtain complete category prediction feature information;

[0174] It should be noted that the classification field refers to the characteristics in the initial prediction characteristic information for describing the vehicle category attribute. These characteristics are usually non-numeric and represent different categories or states. In this embodiment, the classification field includes engine type (such as diesel engine, natural gas engine), road condition (such as national highway, urban area, highway, and other), etc. There is no size relationship between the values of the classification field, and it is only used to distinguish different categories.

[0175] Further, the missing category filling refers to the way of processing the missing values in the classification field, that is, marking the missing values as an independent category or filling them with the category with the highest frequency (mode). For example, when there are missing values in the engine type field, they can be marked as unknown type as a separate category; or if the diesel engine has the highest frequency in the field, the missing values can be filled with the diesel engine.

[0176] Further, the complete category prediction characteristic information refers to the characteristic set of the classification field after the missing value processing, in which all samples of the classification field have explicit values (including original categories and newly added missing categories). At this time, the classification field is combined with the complete numerical prediction characteristic information to form a comprehensive feature set without missing values. For example, after the missing values in the engine type field are marked as unknown type, the field no longer has missing values and is combined with the complete numerical field to form the complete category prediction characteristic information.

[0177] It can be understood that the classification field included in the complete numerical prediction characteristic information is checked to determine whether there are missing values. If there are missing values, the missing category filling method is used: for classification fields such as engine type, if one category has the highest frequency (such as diesel engine accounting for 70%), the mode is used to fill the missing values; if the categories are evenly distributed, the missing values can be marked as missing as an independent category. After processing, the classification field no longer has missing values and is integrated with the complete numerical prediction characteristic information to obtain the complete category prediction characteristic information.

[0178] In step S134, the complete category prediction characteristic information is subjected to abnormal data identification. When the abnormal data is identified in the complete category prediction characteristic information, the abnormal data is deleted or replaced with a reasonable value from the complete category prediction characteristic information to obtain standard prediction characteristic information.

[0179] It should be noted that the abnormal data refers to the data in the complete category prediction characteristic information that deviates significantly from the normal range. These data do not conform to the actual situation or reasonable distribution of the characteristics of the vehicle. In this embodiment, the abnormal data may appear in the numerical field (such as 100 million kilometers of driving mileage, far exceeding the normal life of a heavy truck), or in the classification field (such as engine type marked as electric, but heavy trucks usually have no electric type).

[0180] Furthermore, a reasonable value refers to data that conforms to the normal range or distribution of the field and is used to replace outlier data. The selection of a reasonable value should refer to the common value range of the field or the typical value of similar samples. For example, the outlier value of 10 million kilometers of driving mileage can be replaced by the 99th percentile of the field (such as 800,000 kilometers).

[0181] Understandably, the 3σ criterion is used to identify outliers in numerical fields of the complete categorical predictive feature information: the mean (μ) and standard deviation (σ) of each numerical field are calculated, and when the absolute value of the difference between a data point and the mean is greater than 3σ, it is determined to be an outlier. For categorical fields, outlier categories are identified through manual verification (such as categories that do not match vehicle types). After identifying outlier data, it is handled according to the situation: if the outlier data is an isolated case and has little impact on the overall data, the sample is directly deleted; if the outlier data can be corrected, it is replaced with a reasonable value (such as replacing numerical outliers with the 99th percentile of the field, and replacing categorical outliers with the mode). After processing, standard predictive feature information without missing values ​​and without outliers is obtained. For example, if a sample's "mileage" is 10 million kilometers (far exceeding the 3σ range), it is replaced with the 99th percentile of the field, 800,000 kilometers, to finally form standard predictive feature information. The formula for calculating outliers is as follows:

[0182] Outlier = {x∥x-μ∣>3σ}

[0183] In the formula, x is a single data point in the sample dataset; μ is the mean of the sample dataset; and σ is the standard deviation of the sample dataset.

[0184] Step S14: Determine road condition information based on the wind resistance coefficient, the rolling resistance coefficient, the slope, and the rotational mass conversion coefficient;

[0185] It should be noted that road condition information is formed by integrating parameters related to road conditions, such as drag coefficient, rolling resistance coefficient, slope, and rotational mass conversion factor. Among them, slope directly reflects the degree of inclination of the road; rolling resistance coefficient reflects the roughness of the road surface (such as the difference between asphalt road and gravel road); and the combination of drag coefficient and rotational mass conversion factor with slope and rolling resistance coefficient can more comprehensively describe the road environment resistance when vehicles are driving.

[0186] Step S15: Determine the corresponding historical running data based on the historical trajectory data.

[0187] It can be understood that the historical trajectory data is associated with the running parameters recorded by the vehicle CAN bus through the timestamp to obtain the corresponding historical running data. Specifically, each position point in the historical trajectory data has time information (such as 2024-05-01 08:30:00, longitude and latitude (x, y)), and according to the time information, the running parameters (such as driving speed, engine output power, acceleration, etc.) at the same time point are extracted from the historical records stored by the vehicle CAN bus to form the historical running data corresponding to the trajectory point.

[0188] The embodiment provides a heavy truck life prediction method, which further refines the data acquisition and processing process, adopts technical means such as obtaining vehicle running parameters from multiple channels, data cleaning and outlier processing, solves the technical problem of low data quality in vehicle life prediction, and achieves the beneficial effects of improving data accuracy and reliability and providing higher quality input data for vehicle life prediction.

[0189] It should be noted that the above examples are only used for understanding the present application and do not constitute a limitation on the heavy truck life prediction method of the present application. Further simple transformations based on this technical concept are within the protection scope of the present application.

[0190] The present application also provides a heavy truck life prediction device, please refer to Figure 3 The heavy truck life prediction device comprises:

[0191] The data acquisition module 10 is configured to acquire standard prediction feature information, road condition information, historical trajectory data and historical running data of the vehicle.

[0192] The quality determination module 20 is configured to determine the vehicle quality according to the standard prediction information and the road condition information.

[0193] The working condition identification module 30 is configured to identify the road working condition proportion according to the historical trajectory data and the historical running data.

[0194] The life prediction module 40 is configured to determine the life prediction result of the vehicle according to the vehicle quality, the road working condition proportion, the standard prediction feature information and a preset random survival forest model.

[0195] The heavy truck life prediction device provided by the present application adopts the heavy truck life prediction method in the above embodiment, and can solve the technical problem that the life of the heavy truck cannot be accurately predicted. Compared with the prior art, the beneficial effects of the heavy truck life prediction device provided by the present application are the same as those of the heavy truck life prediction method provided by the above embodiment, and the other technical features in the heavy truck life prediction device are the same as those disclosed in the above embodiment method, which will not be repeated here.

[0196] In an embodiment, the life prediction module 40 is further configured to construct a prediction data set according to the vehicle mass, the road working condition proportion, a historical trip prediction result, and the standard prediction feature information, determine a preset survival split rule according to a preset random survival forest model, randomly sample the prediction data set to generate a prediction data subset, perform data processing on the prediction data subset according to the preset survival split rule to obtain a survival tree set, determine a cumulative risk of a leaf node of a survival tree in the survival tree set, and perform weighted average on the cumulative risk of the survival tree to obtain a target cumulative risk value, and determine the life prediction result of the vehicle according to the target cumulative risk value.

[0197] In an embodiment, the life prediction module 40 is further configured to obtain a prediction time, determine a leaf node according to a survival tree in the survival tree set, and determine prediction data according to the leaf node, determine a first prediction number of vehicles in a discard state and a second prediction number of vehicles in an effective state in the leaf node according to the prediction data within the prediction time, and calculate a cumulative risk value of the leaf node according to the first prediction number and the second prediction number.

[0198] In an embodiment, the mass determination module 20 is further configured to determine an engine output power, a transmission efficiency, an airfoil area, a driving speed, and an acceleration according to the standard prediction information, determine a wind resistance coefficient, a rolling resistance coefficient, a slope, and a rotational mass conversion coefficient according to the road condition information, and determine a vehicle mass according to the engine output power, the transmission efficiency, the airfoil area, the driving speed, the acceleration, the wind resistance coefficient, the rolling resistance coefficient, the slope, and the rotational mass conversion coefficient.

[0199] In an embodiment, the working condition identification module 30 is further configured to determine latitude and longitude information and a preset road network map according to the historical trajectory data, perform matching according to the latitude and longitude information and the preset road network map to obtain a matching result, identify a road type according to the matching result, determine driving time and driving distance of the vehicle on different road types according to the historical running data, and determine a road working condition proportion according to the driving time, the driving distance, and the road type.

[0200] In an embodiment, the data acquisition module 10 is further configured to acquire engine output power, transmission efficiency, wind resistance coefficient, windward area, driving speed, rolling resistance coefficient, slope, rotational mass conversion coefficient, acceleration, and historical trajectory data; determine initial prediction feature information according to the engine output power, the transmission efficiency, the windward area, the driving speed, and the acceleration; perform data cleaning according to the initial prediction feature information to obtain standard prediction feature information; determine road condition information according to the wind resistance coefficient, the rolling resistance coefficient, the slope, and the rotational mass conversion coefficient; and determine corresponding historical operation data according to the historical trajectory data.

[0201] In an embodiment, the data acquisition module 10 is further configured to detect whether there is a missing value in a numerical field in the initial prediction feature information; fill the numerical field to obtain complete numerical prediction feature information when it is detected that there is a missing value in the numerical field; perform missing value judgment on a classification field in the complete numerical prediction feature information, and perform missing category filling when there is a missing value in the complete numerical prediction feature information to obtain complete category prediction feature information; perform abnormal data identification on the complete category prediction feature information, and delete the abnormal data from the complete category prediction feature information or replace the abnormal data with a reasonable value when the abnormal data is identified in the complete category prediction feature information to obtain standard prediction feature information.

[0202] The present application provides a heavy hauler life prediction device, which comprises at least one processor and a memory in communication connection with the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the heavy hauler life prediction method in the above-mentioned embodiment one.

[0203] Reference will now be made to the following description Figure 4 which shows a structural schematic diagram of a heavy hauler life prediction device suitable for use to implement the embodiments of the present application. The heavy hauler life prediction device in the embodiments of the present application can include, but is not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), and the like, and fixed terminals such as digital TVs, desktop computers, and the like. Figure 4 The heavy hauler life prediction device shown is merely an example, and should not impose any limitation on the functions and use range of the embodiments of the present application.

[0204] As shown in Figure 4 the heavy-duty truck life prediction device can include a processing device 1001 (for example, a central processing unit, a graphics processing unit, etc.) that can perform various appropriate actions and processes according to programs stored in a ROM (Read Only Memory) 1002 or programs loaded from a storage device 1003 into a RAM (Random Access Memory) 1004. In the RAM 1004, various programs and data required for the operation of the heavy-duty truck life prediction device are also stored. The processing device 1001, the ROM 1002, and the RAM 1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. In general, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; the storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the heavy-duty truck life prediction device to communicate with other devices wirelessly or by wire to exchange data. Although the heavy-duty truck life prediction device with various systems is shown in the figure, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or provided instead.

[0205] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product comprising a computer program carried on a computer readable medium, the computer program containing program codes for executing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network through the communication device, or installed from the storage device 1003, or installed from the ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the methods of the embodiments disclosed in the present application are performed.

[0206] The heavy-duty truck life prediction device provided by the present application adopts the heavy-duty truck life prediction method in the above-mentioned embodiments, and can solve the technical problem that the life of the heavy-duty truck cannot be accurately predicted. Compared with the prior art, the heavy-duty truck life prediction device provided by the present application has the same beneficial effects as the heavy-duty truck life prediction method provided by the above-mentioned embodiments, and other technical features in the heavy-duty truck life prediction device are the same as the features disclosed in the previous embodiment method, which will not be repeated here.

[0207] It should be understood that various aspects disclosed herein can be implemented in hardware, software, firmware, or combinations thereof, to achieve the various aspects disclosed herein. In the description above, specific terminology is used to describe particular features, structures, materials or characteristics. It should be understood, however, that the terminology is used only in a descriptive sense and not for purposes of limitation.

[0208] The above description is only specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0209] The present application provides a computer readable storage medium having stored thereon computer readable program instructions (i.e., computer programs) for performing the heavy haul truck life prediction method in the above-described embodiments.

[0210] The computer readable storage medium provided by the present application may, for example, be a U disk, but is not limited to an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, system, or device, or any combination thereof. More specific examples of the computer readable storage medium can include, but are not limited to, an electrical connection having one or more conductive wires, a portable computer diskette, a hard disk, a RAM (Random Access Memory), a ROM (Read Only Memory), an erasable programmable ROM (EPROM) or flash memory, an optical fiber, a CD-ROM (CD-Read Only Memory), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present embodiment, the computer readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer readable storage medium can be transmitted in any suitable medium, including but not limited to electrical wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0211] The above computer readable storage medium can be included in the heavy haul truck life prediction device; or can exist separately without being assembled into the heavy haul truck life prediction device.

[0212] The computer readable storage medium described above carries one or more programs, when the one or more programs are loaded by the heavy-duty truck life prediction device, the heavy-duty truck life prediction device is caused to: acquire standard prediction feature information, road condition information, historical trajectory data and historical operation data of a vehicle; determine vehicle quality according to the standard prediction information and the road condition information; identify road working condition proportion according to the historical trajectory data and the historical operation data; and determine a life prediction result of the vehicle according to the vehicle quality, the road working condition proportion, the standard prediction feature information and a preset random survival forest model.

[0213] Computer program code for carrying out operations of the present application can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0214] The flow diagrams and the block diagrams in the drawings are illustrations of architectures, functionalities, and operations of possible implementations of systems, methods, and computer program products according to various embodiments of present application. In this regard, each block in the flow diagrams or block diagrams can represent a module, a segment, or a portion of code, which comprises one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and / or flow diagrams, and combinations thereof, can be implemented by special purpose hardware-based systems that perform the specified functions or operations, or combinations of special purpose hardware and computer instructions.

[0215] The modules described in the embodiments of the present application can be implemented in the form of software or in the form of hardware. In some cases, the names of the modules do not constitute a limitation on the modules themselves.

[0216] The computer readable storage medium provided in the present application is a computer readable storage medium, which stores computer readable program instructions (i.e., a computer program) for executing the above-mentioned heavy truck life prediction method, and can solve the technical problem that the life of the heavy truck cannot be accurately predicted. Compared with the prior art, the computer readable storage medium provided in the present application has the same beneficial effects as the heavy truck life prediction method provided in the above-mentioned embodiments, and will not be described here.

[0217] The present application also provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the heavy truck life prediction method as described above.

[0218] The computer program product provided in the present application can solve the technical problem that the life of the heavy truck cannot be accurately predicted. Compared with the prior art, the computer program product provided in the present application has the same beneficial effects as the heavy truck life prediction method provided in the above-mentioned embodiments, and will not be described here.

[0219] The above only describes some embodiments of the present application, and does not limit the patent scope of the present application, and any equivalent structural transformation made by using the content of the present application specification and drawings, or direct / indirect application in other related technical fields is included in the patent protection scope of the present application.

Claims

1. A method for predicting the life of a heavy duty truck, characterized by, The method comprises: acquiring standard prediction characteristic information, road condition information, historical trajectory data and historical operation data of a vehicle; determining vehicle quality according to the standard prediction information and the road condition information; identifying road working condition proportion according to the historical trajectory data and the historical operation data; determining a life prediction result of the vehicle according to the vehicle quality, the road working condition proportion, the standard prediction characteristic information and a preset random survival forest model.

2. The method of claim 1, wherein, The step of determining the life prediction result of the vehicle according to the vehicle quality, the road working condition proportion, the standard prediction characteristic information and the preset random survival forest model comprises: constructing a prediction data set according to the vehicle quality, the road working condition proportion, historical trip prediction results and the standard prediction characteristic information, and determining a preset survival split rule according to the preset random survival forest model; randomly sampling the prediction data set to generate a prediction data subset; performing data processing on the prediction data subset according to the preset survival split rule to obtain a survival tree set; determining a cumulative risk of a leaf node of a survival tree in the survival tree set, and performing weighted average on the cumulative risk of the survival tree to obtain a target cumulative risk value; determining the life prediction result of the vehicle according to the target cumulative risk value.

3. The method of claim 2, wherein, The step of determining the cumulative risk of the leaf node of the survival tree in the survival tree set comprises: acquiring a prediction time, determining a leaf node according to a survival tree in the survival tree set, and determining prediction data according to the leaf node; determining a first prediction number of vehicles in a discarded state and a second prediction number of vehicles in an effective state in the leaf node according to the prediction data within the prediction time; calculating the cumulative risk value of the leaf node according to the first prediction number and the second prediction number.

4. The method of claim 1, wherein, The step of determining the vehicle quality according to the standard prediction information and the road condition information comprises: determining engine output power, transmission efficiency, windward area, driving speed and acceleration according to the standard prediction information; determining wind resistance coefficient, rolling resistance coefficient, slope and rotational mass conversion coefficient according to the road condition information; determining vehicle quality according to the engine output power, the transmission efficiency, the windward area, the driving speed, the acceleration, the wind resistance coefficient, the rolling resistance coefficient, the slope and the rotational mass conversion coefficient.

5. The method of claim 1, wherein, The step of identifying road working condition proportion according to the historical trajectory data and the historical operation data comprises: determining latitude and longitude information and a preset road network map according to the historical trajectory data, and matching the latitude and longitude information with the preset road network map to obtain a matching result; identifying road types according to the matching result; determining driving time and driving distance of the vehicle on different road types according to the historical operation data; determining road working condition proportion according to the driving time, the driving distance and the road types.

6. The method of claim 1, wherein, The step of acquiring standard prediction characteristic information, road condition information, historical trajectory data and historical operation data of a vehicle comprises: acquire engine output power, transmission efficiency, wind resistance coefficient, windward area, running speed, rolling resistance coefficient, slope, rotational mass conversion coefficient, acceleration, and historical trajectory data; determine initial prediction feature information according to the engine output power, the transmission efficiency, the windward area, the running speed, and the acceleration; perform data cleaning according to the initial prediction feature information to obtain standard prediction feature information; determine road condition information according to the wind resistance coefficient, the rolling resistance coefficient, the slope, and the rotational mass conversion coefficient; determine corresponding historical operation data according to the historical trajectory data.

7. The method of claim 6, wherein, The step of performing data cleaning according to the initial prediction feature information to obtain standard prediction feature information includes: detecting whether there is a missing value in a numerical field in the initial prediction feature information; when it is detected that there is a missing value in the numerical field, filling the numerical field to obtain complete numerical prediction feature information; performing missing value judgment on a classification field in the complete numerical prediction feature information, and when there is a missing value in the complete numerical prediction feature information, filling a missing category to obtain complete category prediction feature information; performing abnormal data identification on the complete category prediction feature information, and when the abnormal data is identified in the complete category prediction feature information, deleting the abnormal data from the complete category prediction feature information or replacing the abnormal data with a reasonable value to obtain standard prediction feature information.

8. A heavy haul truck life prediction device characterized by, The device includes: a data acquisition module configured to acquire standard prediction feature information, road condition information, historical trajectory data, and historical operation data of a vehicle; a quality determination module configured to determine vehicle quality according to the standard prediction information and the road condition information; a working condition identification module configured to identify a road working condition proportion according to the historical trajectory data and the historical operation data; a life prediction module configured to determine a life prediction result of the vehicle according to the vehicle quality, the road working condition proportion, the standard prediction feature information, and a preset random survival forest model.

9. A heavy-haul truck life prediction apparatus characterized by comprising: The device includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and the computer program is configured to implement the steps of the heavy truck life prediction method according to any one of claims 1 to 7.

10. A storage medium, characterized by The storage medium is a computer readable storage medium, and the storage medium stores a computer program, and the computer program is executed by a processor to implement the steps of the heavy truck life prediction method according to any one of claims 1 to 7.

Images