2.1

Parameter requirements of performance analysis model of mass production vehicle

The application of reliability technology in domestic automobile product development is still lacking. In the process of automobile design, it still relies on experience design, the reliability design system is not perfect, the relevant reliability management tools are not used, and the popularity of reliability knowledge is not enough; The distribution law of automobile faults can be approximately described by normal distribution, lognormal distribution, exponential distribution and Weibull distribution. It is generally believed that the first failure mileage of complex repairable systems follows Weibull distribution ⁹. In the performance evaluation, it also needs to be carried out according to the requirements of test procedures and test specifications. After the test vehicle reliability driving project is completed, it is necessary to disassemble and inspect the internal structure of each assembly for other abnormal phenomena. In the performance evaluation of production vehicles, it is necessary to record and count the test conditions, faults and other test data for the performance evaluation and evaluation of production vehicles.

The reliability and durability of automobiles depend greatly on the reliability and durability of engines. The overhaul of engine is the main sign that determines the overhaul of automobile. There are many factors that affect the durability of the engine, and there are also many failure modes. The mileage of the car driven by the engine during the stable wear period. Generally, it is the starting mileage of deteriorating wear period minus the ending mileage of early wear period. The pattern of performance evaluation system of mass production vehicle in automobile design and development is shown in Figure 1.

Figure 1.

Performance evaluation system of mass production vehicle

As an efficient means of transportation, the transportation efficiency of automobile depends largely on the dynamic performance of automobile, so the dynamic performance index is the most basic and important index among various performance indexes of automobile, and the quality of dynamic performance directly affects the use of automobile on the road. The efficiency of automobile transportation can be expressed by the driving force that can be distributed per unit mass of automobile to remove wind resistance.

F_t is the ramp resistance, F_w is the clothing rolling resistance, and m is the mass of the vehicle under the calculated load condition, in kg; g is the acceleration of gravity in m/s ².

Fuel economy is one of the basic properties of automobiles. Oil is the main energy source of modern industry, especially transportation. Good fuel economy can reduce the use cost of cars, save energy and protect the environment. In the actual calculation, the engine’s external characteristic power is the actual calculated value obtained from the resistance power and the transmission efficiency, and then divided by 90% to obtain the engine’s external characteristic power ¹⁰.

The ordinate of the engine universal characteristic curve is usually the engine torque, and the engine torque and engine power are converted by the following formula:

Where: T is the torque of the engine, and the unit is N·m. P_e is the engine power in kW. n is the engine speed, r/min.

The quality evaluation model is based on the equivalent volume and kerb quality of production vehicles, and with the help of the basic big data theory, the kerb quality is targeted in the early development of automobiles. Through a large number of data measurement of competing cars, it is found that there are great differences in λ for different models of SUVs and cars. In general, when the λ of SUVs is 0.662 and the λ of cars is 0.5, it is more in line with the actual situation and the error of calculation results is the smallest ¹¹. Equivalent volume formula:

Where: a, b, c is vehicle length, vehicle width and vehicle height, mm; d is the wheelbase, mm; v is equivalent volume, m³.

2.2

Construction of performance evaluation model for mass production vehicles

When developing a new component or vehicle model, designers should incorporate the concept of reliability design and embody it in the design process of the product. While pursuing high performance and structural advancement, they must base themselves on the reliability of the product and adopt new technologies, materials and processes to reassure users of the reliability, durability and safety of the product, thus improving the quality and market competitiveness of the product. Data mining technology can play an active role in evaluating the performance of mass production vehicles. In the development of enterprises, it is an important measure to choose a scientific and efficient way to improve efficiency. Through the application of data mining technology, it is convenient for enterprises to understand and analyze relevant data about automobile manufacturing. In data analysis, you can clearly see the whole database data.

Data mining draws lessons from the knowledge of statistics, machine learning, databases and other fields in methods. Data mining was first carried out by experts in the field of databases, and getting up early in the database is a very difficult professional technology ¹². With the development and commercial popularization of database technology, its operation is becoming simpler and simpler, and the reduction of learning cost makes more and more researchers in other fields begin to use and master database technology. Since then, data mining has gradually become a hot topic in other fields. Functions such as reliability prediction can be realized. In engineering practice, they are often combined and applied, which is convenient to give play to their respective advantages and complement each other, and can receive remarkable benefits with twice the result with half the effort.

Based on the big data theory, this paper establishes a performance target evaluation model, and uses this model to set the performance target of mass production vehicles to improve their reliability. In this way, in the early development stage of the mass production vehicle, the performance target of the mass production vehicle can be formulated under the consideration of the market, which can not only reflect the market competitiveness of the quality, but also improve the reliability of the performance target.

Artificial neural network uses its nonlinear mapping idea and parallel processing method, and can express the related knowledge of input and output with its own structure. It is a multi-input and single-output nonlinear element, and its input-output relationship can be described as:

Where j = 1,2,⋯, n is the input signal from other cells, indicating the connection weight from cell j to cell i, and n is the number of input signals. Y is the neuron output. f (·) is called the transfer function.

Data mining has a wide range of applications, involving different fields, such as supermarket customer resource analysis, student data analysis and so on. In addition, there are different models and algorithms for data mining, such as DT(Decision tree), RF(Random forest), SVM(Support vector machine), KNN (k-nearest neighbor) and so on. Different models and algorithms are suitable for different scenarios. In this paper, KNN, DT and RF models are mainly used to analyze the quality of vehicles and predict whether the quality of vehicles is qualified or not through different characteristics of vehicles.

KNN is a stable and effective classification algorithm based on statistics. KNN algorithm has outstanding advantages: clear concept, simple implementation, good robustness and good classification effect. For a given unknown sample, calculate its similarity with each training sample and sort it, take the first k training samples, and judge the category of the unknown sample according to the category of the k samples, that is, the category of the k samples with the largest number of samples is the unknown sample.

Many DT implementations are based on the principle of information theory. In the process of DT formation, the most important part is the choice of split attributes. A common method is to calculate the information gain. The principle of information gain comes from information theory, which is the expected entropy reduction caused by using an attribute to divide the training set. The formation of DT is to recursively split each node in the data set until all categories of nodes belong to the same class or there are no redundant attributes to divide the training sample set.

Let X₁, X₂,⋯,X_n be the signal sent by the source. Before receiving X_i, the receiver’s uncertainty about the signal sent by the source is defined as the self-information I(X_i) of the information symbol, that is, I(X_i) = –log₂ P(X_i), where P(X_i) is the probability that the source sends X_i.

Self-information can only reflect the uncertainty of symbols, and information entropy can be used to measure the uncertainty of the whole source X, which is defined as follows:

For imbalanced classification, the simplest and most effective method may be the data level method, which directly preprocesses the data samples in the early stage to balance the data samples. For example, random downsampling and random oversampling methods reduce the number of negative samples by randomly removing negative samples and increase the number of positive samples by randomly copying positive samples to achieve the purpose of balancing samples.

RF is a classifier with multiple DTs as the basis, and the final prediction result depends on the voting of each DT’s prediction result, with the class with more votes as the final prediction result. The advantage of combined classifier is that its results are more accurate than each individual classifier. In addition, when RF is divided each time, the accuracy of selecting a single or a small number of features is better than that of multiple features. Therefore, RF is also very effective on large data sets. RF also potentially gives an assessment of the importance of features.