Abrasion state evaluation method for transmission case

Abstract

A transmission case wear state evaluation method belongs to the technical field of wear state evaluation, and comprises the following steps: step S01, collecting wear particle data in transmission case oil; s02, establishing evaluation indexes for the wear particle data in the step S01; s03, optimizing a penalty function in the projection pursuit model PPC by adopting an improved parrot optimization algorithm IPO to obtain an optimal objective function; and S04, constructing a wear state evaluation model IPO-PPC and the like by adopting the optimal objective function and the evaluation indexes. According to the invention, by improving the parrot optimization algorithm, the convergence rate, the accuracy and other aspects of the parrot optimization algorithm are greatly improved, the penalty function in the projection pursuit method is optimized by the improved parrot optimization algorithm, the wear state evaluation model is established, and six evaluation indexes of the oil abrasive particle data are combined, so that the accuracy of the oil abrasive particle data is improved. The abrasion state of the transmission case can be quickly and accurately identified.

Description

Technical Field

[0001] The invention belongs to the technical field of transmission case state assessment, and in particular relates to a transmission case wear state assessment method. Background Art

[0002] In the complex and ever-changing operating environment of equipment vehicles, the transmission box is subjected to high load and high-intensity working conditions, and its internal mechanical parts are prone to wear, fatigue and even failure. Real-time evaluation of the wear status of the transmission box can not only effectively predict the service life of key components and avoid mission interruption or combat damage risks caused by sudden failures, but also provide a scientific basis for the optimization of equipment maintenance strategies, thereby improving the combat availability and full life cycle management level of the vehicle.

[0003] Traditional wear status assessment methods usually rely on vibration signal analysis or regular disassembly inspection, but these methods have problems such as high data acquisition cost, poor real-time performance or insufficient diagnostic accuracy. In order to solve the above problems, wear status assessment can adopt an assessment method based on oil wear particle information, which can more accurately reflect the internal wear status of the transmission case and has the advantages of non-invasiveness and real-time detection.

[0004] Among many data analysis algorithms, the projection pursuit algorithm PPC has shown significant advantages due to its unique dimensionality reduction capability and nonlinear feature extraction performance. Compared with traditional principal component analysis or linear regression algorithms, the projection pursuit algorithm PPC can better capture the potential nonlinear relationships and complex patterns in oil wear data, improving the accuracy and robustness of wear assessment. In addition, the projection pursuit algorithm PPC shows strong stability when processing high-dimensional and noisy data sets, making it particularly suitable for wear status assessment under complex working conditions of vehicle transmissions. However, the projection pursuit algorithm PPC has a strong dependence on initial parameters, such as penalty functions, and is prone to falling into local optimal solutions, and its performance needs to be improved urgently. Summary of the invention

[0005] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a method for evaluating the wear status of a transmission case. By improving the Parrot optimization algorithm, its convergence speed and accuracy are greatly improved. The penalty function in the projection pursuit method is optimized by the improved Parrot optimization algorithm, and a wear status evaluation model is established. Combined with six evaluation indicators of oil abrasive data, it is possible to quickly and accurately identify the wear status of the transmission case.

[0006] In order to achieve the above object, the main technical solutions adopted by the present invention include: A method for evaluating the wear state of a transmission case comprises the following steps: Step S01, collecting wear particle data in transmission box oil; Step S02, establishing an evaluation index for the wear particle data of step S01, and dividing the wear particle data into a training data set and a test data set; Step S03, using the improved parrot optimization algorithm IPO to optimize the penalty function in the projection pursuit model PPC to obtain the best objective function; the improved parrot optimization algorithm includes introducing an adaptive inertia weight in the foraging behavior and introducing a t-distribution disturbance in the communication behavior; Step S04, using the optimal objective function in step S03 and the evaluation index in step S02 to introduce the projection pursuit model PPC, and construct a wear state evaluation model IPO-PPC; Step S05, using the training data set in step S01 to train the wear state assessment model IPO-PPC constructed in step S04; Step S06, using the test data set in step S01 to test the wear state assessment model IPO-PPC constructed in step S05; Step S07: Use the wear state assessment model IPO-PPC that has passed the test in step S06 to assess the wear state of the transmission case and output the assessment result.

[0007] Furthermore, in step S03, the adaptive inertia weight gradually decreases in a nonlinear manner according to the increasing number of iterations. The improved formula of the parrot optimization algorithm IPO after introducing the adaptive inertia weight in the foraging behavior is as follows: ; Where t is the current iteration number; T is the maximum iteration number; is the position of the i-th individual in the t-th iteration; is the position of the ith individual at iteration t+1; X best is the best position from initialization to the current search; is the average position in the current population; Levy(dim) is the Levy distribution; dim is the number of problem dimensions; η is the cheap position coefficient that changes with the number of iterations; δ is the first weight coefficient, β is the second weight coefficient; ξ is the adaptive inertia weight.

[0008] Furthermore, the improved formula of the improved Parrot optimization algorithm for introducing t-distribution disturbance into the communication behavior is as follows: ; Among them, rand(0,1) is a random number uniformly distributed between [0,1], t(θ) is a random perturbation value that obeys a t-distribution with degrees of freedom θ, and C Iter is the number of iterations, and P is a random number between [0,1].

[0009] Furthermore, in the step S02, six evaluation indices are established according to the size information in the wear particle data, including 0-100 μm, 100 μm-200 μm, 200 μm-300 μm, 300 μm-450 μm, 450 μm-600 μm and above 600 μm.

[0010] The beneficial effects of the present invention are: 1. Use the projection pursuit algorithm to analyze the collected data, reduce the dimensionality of high-dimensional data, reduce data redundancy, and improve data quality. This algorithm is efficient and practical, and is suitable for wear status monitoring scenarios; 2. The foraging and communication behaviors of the parrot optimization algorithm are improved by introducing adaptive inertia weights and t-disturbance distribution, solving the problem of weak convergence, increasing the convergence speed and expanding the search range; Introducing adaptive inertia weights into the foraging behavior of the parrot optimization algorithm gives the algorithm stronger global search capabilities and a wider search range, helping the algorithm to break free from the constraints of local optimal solutions; Introducing t-distribution perturbations in the communication behavior of the Parrot Optimization Algorithm. T-distribution combines the advantages of Gaussian distribution and Cauchy distribution, that is, Gaussian mutation can ensure the convergence speed of the algorithm in the later stage, while Cauchy mutation maintains the diversity of the population. In the communication stage, the information of other individuals in the population is introduced into the global search update mechanism, which enhances the diversity of the search space. 3. By optimizing the penalty function in the projection pursuit algorithm using the improved Parrot optimization algorithm, the optimal projection vector can be determined more accurately and quickly, the generalization performance can be enhanced, and the wear status of the transmission case can be identified quickly and accurately, while providing a basis for future offline evaluation of equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 The present invention is a flow chart of a method for evaluating the wear state of a transmission case. DETAILED DESCRIPTION

[0012] In order to better explain the present invention and facilitate understanding, the present invention is described in detail below through specific implementation modes in conjunction with the accompanying drawings.

[0013] like Figure 1 As shown, an embodiment of the present invention provides a method for evaluating the wear state of a transmission case, comprising the following steps: Step S01, collecting wear particle data in the transmission box oil, specifically by using an oil sensor.

[0014] Step S02: establishing evaluation indicators for the wear particle data of step S01, and dividing the wear particle data into a training data set and a test data set.

[0015] Specifically, six evaluation indicators are established based on the size information in the wear particle data, including 0-100μm, 100μm-200μm, 200μm-300μm, 300μm-450μm, 450μm-600μm and above 600μm. The wear particle data is divided into 70% as the training data set and 30% as the test data set.

[0016] Step S03, using the improved parrot optimization algorithm IPO to optimize the penalty function in the projection pursuit model PPC to obtain the best objective function; the improved parrot optimization algorithm includes introducing an adaptive inertia weight in the foraging behavior and introducing a t-distribution disturbance in the communication behavior.

[0017] The specific algorithm and improvements of Parrot Optimization Algorithm IPO are as follows: (1) Population initialization, the initialization formula is:

[0018] Where ub and lb are the lower and upper bounds of the space search; is the position of the i-th parrot at the initial stage; rand(0,1) is a random number uniformly distributed between [0,1].

[0019] (2) Foraging behavior When looking for food, they mainly estimate the approximate location of the food by observing the location of the food or considering the location of the owner. The mathematical formula is: ; Where t is the current iteration number; T is the maximum iteration number; is the position of the i-th individual in the t-th iteration; is the position of the ith individual at iteration t+1; X best is the best position from initialization to the current search; is the average position in the current population; Levy(dim) represents Levy distribution; dim is the number of problem dimensions; η represents the cheap position coefficient that changes with the number of iterations; N is the number of populations; μ, ν are positive integers in (0, dim); Γ is the Gamma function; γ is assigned a value of 1.5; σ is the probability density of Levy flight distribution.

[0020] Adaptive inertia weights are introduced into the foraging behavior. The weights gradually decrease in a nonlinear manner as the number of iterations increases. In the early stages of iteration, larger weight values ​​give the algorithm stronger global search capabilities and a wider search range, which helps the algorithm break free from the constraints of local optimal solutions. As the number of iterations increases, the weight value will gradually decrease, allowing the algorithm to explore the area near the optimal solution in more detail, thereby improving the algorithm's optimization accuracy and accelerating convergence. The improved formula is:

[0021] Among them, δ is the first weight coefficient, β is the second weight coefficient, and ξ is the adaptive inertia weight.

[0022] (3) Staying behavior The parrot's stopping behavior mainly consists of suddenly flying to any part of the owner's body and staying there for a period of time. The mathematical expression of this process is:

[0023] Among them, ones(1,dim) is a vector of all 1s of dimension dim.

[0024] (4) Communication behavior Parrots will communicate closely within the group. This communication behavior includes flying to the flock and not flying to the flock. Assume that the probability of these two behaviors is equal, and use the average position of the current population to symbolize the center of the group. The mathematical expression of this process is:

[0025] The t-distribution perturbation is introduced in the communication behavior. The t-distribution combines the advantages of Gaussian distribution and Cauchy distribution. That is, Gaussian mutation can ensure the convergence speed of the algorithm in the later stage, while Cauchy mutation maintains the diversity of the population. In the communication stage, the information of other individuals in the population is introduced into the global search update mechanism, which enhances the diversity of the search space. The improved new formula is:

[0026] Where t(θ) is a random disturbance value that follows a t-distribution with degrees of freedom θ; C Iter is the number of iterations; P is a random number between [0,1].

[0027] (5) Fear of strangers Generally speaking, birds show a natural fear of strangers. They keep a distance from unfamiliar individuals and seek a safe environment with their owners. The following formula can be used to express this behavior:

[0028] Here, the second term represents the process of redirecting to fly toward the owner, and the last term represents the process of moving away from the stranger.

[0029] The specific steps of the projection pursuit algorithm PPC are as follows: (1) Normalization of sample evaluation index data: In order to eliminate the influence of different abrasive particle concentration dimensions, the wear particle data collected in step S01 is divided into multiple groups of wear particle data according to the evaluation index in step S02, and the wear particle data is normalized. The formula is as follows:

[0030] In the formula, x * i is the value of the ith data indicator; x imax， x imin are the maximum and minimum values ​​of the i-th data index respectively; x i is the eigenvalue normalized sequence.

[0031] (2) Constructing a projection pursuit model Construct the objective function J(a) in the form of:

[0032] Among them, Q(a) is the main function used to reflect the distribution characteristics of the projection data; y i is the projection value of the i-th data point; is the mean of the projection values; Var(y) is the projection variance used to maximize the projection information of the data; P(a) is the penalty function used to constrain the projection direction; λ is the penalty function coefficient; d is the dimension of the original high-dimensional space; a is the projection vector; ‖a‖ is the bi-norm of the projection vector; a j is the jth component of the projection vector a; a T is the transpose of the projection vector.

[0033] (3) Determine the projection vector Through the objective function J(a), the optimal projection vector a* is obtained:

[0034] (4) Projection to low-dimensional space The wear particle data is projected into the low-dimensional space through the optimal projection vector a*:

[0035] (5) Analyze the projection boundary Analyze the distribution range of projection values ​​through historical data to determine the operating projection range (such as boundary value). Compare the projection values ​​of real-time data to see if they fall within the normal range to determine the wear status.

[0036] The improved Parrot optimization algorithm is used to optimize the penalty function in the projection pursuit model PPC to obtain the optimal objective function. The specific execution process of the optimization can be obtained by those skilled in the art based on the content of the present invention and the prior art, and will not be repeated here.

[0037] Step S04, using the optimal objective function in step S03 and the evaluation index in step S02 to introduce the projection pursuit model PPC, and construct a wear state evaluation model IPO-PPC; Step S05, using the training data set in step S01 to train the wear state assessment model IPO-PPC constructed in step S04; Step S06, using the test data set in step S01 to test the wear state assessment model IPO-PPC constructed in step S05; Step S07: Use the wear state assessment model IPO-PPC that has passed the test in step S06 to assess the wear state of the transmission case and output the assessment result.

[0038] The specific steps of evaluating the transmission case using the wear status evaluation model IPO-PPC are as follows: (1) Establishing evaluation indicators based on the size information of wear particle data of the transmission case; (2) Normalize the wear particle data according to the evaluation index to generate the initial population; (3) The improved Parrot optimization algorithm IPO is used to optimize the penalty function of the projection pursuit model PPC and obtain the optimal objective function; (4) Constructing the wear status assessment model IPO-PPC; (5) Set the limit value through historical data, compare the projection vector in the wear state evaluation model IPO-PPC with the limit value, and determine the optimal projection vector; (6) Determine the optimal projection value and evaluate the wear status.

[0039] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. Alterations, modifications, substitutions and variations of the above embodiments by a person skilled in the art are all within the scope of the present invention.

Claims

1. A method for evaluating the wear state of a transmission case, characterized in that: The steps include: Step S01, collecting wear particle data in transmission box oil; Step S02, establishing an evaluation index for the wear particle data of step S01, and dividing the wear particle data into a training data set and a test data set; Step S03, using the improved parrot optimization algorithm IPO to optimize the penalty function in the projection pursuit model PPC to obtain the best objective function; the improved parrot optimization algorithm includes introducing an adaptive inertia weight in the foraging behavior and introducing a t-distribution disturbance in the communication behavior; Step S04, using the optimal objective function in step S03 and the evaluation index in step S02 to introduce the projection pursuit model PPC, and construct a wear state evaluation model IPO-PPC; Step S05, using the training data set in step S01 to train the wear state assessment model IPO-PPC constructed in step S04; Step S06, using the test data set in step S01 to test the wear state assessment model IPO-PPC constructed in step S05; Step S07: Use the wear state assessment model IPO-PPC that has passed the test in step S06 to assess the wear state of the transmission case and output the assessment result.

2. The method for evaluating the wear state of a transmission case according to claim 1, characterized in that: In step S03, the adaptive inertia weight gradually decreases in a nonlinear manner according to the increasing number of iterations. The improved formula of the Parrot Optimization Algorithm IPO after introducing the adaptive inertia weight in the foraging behavior is as follows: ； Where t is the current iteration number; T is the maximum iteration number; is the position of the i-th individual in the t-th iteration; is the position of the ith individual at iteration t+1; X best is the best position from initialization to the current search; is the average position in the current population; Levy(dim) is the Levy distribution; dim is the number of problem dimensions; η is the cheap position coefficient that changes with the number of iterations; δ is the first weight coefficient, β is the second weight coefficient; ξ is the adaptive inertia weight.

3. The wear condition evaluation method of a transmission case according to claim 2, characterized in that: The improved formula of the improved Parrot optimization algorithm for introducing t-distribution disturbance into the communication behavior is as follows: ； Among them, rand(0,1) is a random number uniformly distributed between [0,1], t(θ) is a random perturbation value that obeys a t-distribution with degrees of freedom θ, and C Iter is the number of iterations, and P is a random number between [0,1].

4. The method for evaluating the wear state of a transmission case according to claim 1, characterized in that: In the step S02, six evaluation indices are established according to the size information in the wear particle data, including 0-100 μm, 100 μm-200 μm, 200 μm-300 μm, 300 μm-450 μm, 450 μm-600 μm and above 600 μm.

Images