Multi-objective optimization method for working parameters of sprocket type residual film recycling machine

Abstract

According to the multi-objective optimization method for the working parameters of the sprocket type residual film recycling machine, the sprocket type residual film recycling machine is used as a research object, and the angular velocity of a picking device, the angular velocity of a film and impurity separating device and the working angle of the film and impurity separating device are set as key variables to be optimized. In order to obtain the optimal residual film impurity rate and picking-up rate, discrete element simulation is performed through an EDEM, a back propagation neural network is trained by using simulation data, and a prediction model of the residual film impurity rate and picking-up rate is established. A prediction result of the neural network is optimized by adopting a non-dominated multi-objective genetic algorithm, and optimal working parameters are given in a Pareto frontier form. When the angular speed of the picking device is 58.80 rad / s, the angular speed of the film and impurity separating device is 40 rad / s, and the angle of the film and impurity separating device is 35.51, the impurity content of the residual film is 24.26%, and the picking rate of the residual film is 83.67%. And working parameters are selected in the Pareto front, so that the impurity rate of the residual film can be effectively reduced on the basis of keeping a relatively high residual film picking-up rate.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of optimizing residual film recycling machines, in particular to a multi-objective optimization method for the working parameters of a sprocket type residual film recycling machine. BACKGROUND

[0002] Mulching with plastic film has many significant advantages, not only can greatly improve the yield and quality of crops, but also can effectively reduce the cost of agricultural production, thereby significantly increasing the economic income of farmers. However, this technology also faces some challenges in the application process. Plastic film mulching can reduce soil aeration, negatively affect soil microbial activity and plant root respiration, and ultimately threaten soil health and ecological balance. To solve this problem, the use of residual film recycling machines is vigorously promoted in the field of agriculture, which greatly reduces the residual plastic film in the soil. However, at present, the impurity content of the plastic film recycled by the residual film recycling machine is still at a high level, the main reason is that the plastic film is entangled with soil, straw and other impurities. The high impurity content greatly hinders the recycling and reuse of residual film.

[0003] Therefore, it is crucial to optimize the working parameters of the residue film recycling machine. The adjustment of these parameters will directly affect the recycling efficiency of the machine, including the purity and efficiency of the recycled film. By scientifically and reasonably setting and adjusting these parameters, the performance of the residue film recycling machine can be improved, and the impurity rate of the recycled film can be reduced, providing a solid guarantee for the sustainable development of agricultural production. The selection of working parameters of the residue film recycling machine usually relies on experimental design methods such as orthogonal experiment and response surface. These methods aim to obtain the main effect of the target variable related to the residue film pickup rate and the impurity rate of the recycled film, and determine its optimal level, and then select the best value of each parameter. For example, Zhang Xuejun et al. used the orthogonal experiment method to deeply explore the specific influence of the angle and angular velocity of the film-soil separation device and the angular velocity of the pickup device on the residue film recycling efficiency. Based on the research data obtained, the key parameter combination for achieving the best residue film recycling effect was determined through response surface analysis. Guo Wensong et al. used the discrete element software EDEM to carry out residue film recycling experiments and numerical simulation, explored the relationship between the machine forward speed, pickup device angular velocity and residue film pickup rate, and obtained the optimal parameter combination through response surface analysis. Liu Jiaqiang et al. studied the influence of the rotation speed and angle of the film-soil separation device on the film-soil separation effect, designed a three-factor four-level orthogonal experiment, and determined the primary and secondary relationships of the three factors. In the past research, intelligent optimization algorithms have been widely and deeply explored. Although the application of intelligent optimization algorithms in the field of residue film recycling machines is still in the development stage and has not yet matured, the successful practice in other fields provides a wealth of theoretical and experimental basis for the optimization of working parameters of residue film recycling machines. For example, Feng Yi et al. used neural networks to realize real-time prediction of mechanical drilling speed. Jia Dewen et al. achieved remarkable results in the optimization of a straight-line four-cylinder diesel engine. They combined genetic algorithms with neural networks to optimize the key parameters of the diesel particulate filter (DPF), such as length-to-diameter ratio, wall thickness and pore density, and successfully obtained the Pareto front that fully reflects the performance of the DPF.

[0004] In the field of residue film recycling, many researchers have devoted a lot of effort to exploring the optimization method of parameters, and the core goal is to find the most suitable parameter configuration for a specific working condition. However, in the actual residue film recycling operation, the diversity of working conditions and the difference of technical requirements make the optimization parameters have different goals. How to flexibly select the optimization goal in different working stages and adapt the appropriate optimization method for different goals has become a key problem that needs to be solved in this field. In view of the fact that the characteristics of the film and impurity mixture are complex in the residue film recycling process, and there is a mutual influence between the working parameters of the residue film recycling machine, the traditional single-objective optimization method often cannot meet the actual needs. In this case, the multi-objective optimization method highlights its crucial value. This method allows researchers to consider multiple objective functions simultaneously and effectively balance the contradictions between different objectives by seeking a set of optimal solutions, thereby achieving global optimization. In the residue film recycling process, there are multiple optimization targets, such as reducing the residue film impurity rate and improving the residue film pickup rate.

[0005] To further analyze the matching mechanism of the working parameters of the residue film recycling machine during residue film recycling operation, the neural network is used to approximate the internal relationship between the design variables and the optimization targets. On this basis, the multi-objective genetic algorithm (NSGA-II) is used to construct a multi-objective optimization strategy for the angular velocity of the pickup device, the angular velocity and angle of the film and impurity separation device. This strategy successfully solves the two major problems of high residue film impurity rate and low residue film pickup rate in the residue film recycling process, significantly improving the overall efficiency and quality of residue film recycling. SUMMARY

[0006] To solve the problems existing in the prior art, the purpose of the present application is to provide a multi-objective optimization method for the working parameters of a chain tooth type residue film recycling machine.

[0007] To achieve the above purpose, the technical scheme of the present application is as follows:

[0008] The multi-objective optimization method for the working parameters of the chain tooth type residue film recycling machine takes the chain tooth type residue film recycling machine as the research object, and sets the angular velocity of the pickup device, the angular velocity of the film and impurity separation device, and the working angle of the film and impurity separation device as the key variables to be optimized. In order to obtain the optimal residue film impurity rate and pickup rate, discrete element simulation is performed by EDEM, and a prediction model of the residue film impurity rate and pickup rate is established by training the back propagation neural network with simulation data. The prediction results of the neural network are optimized by using the non-dominated multi-objective genetic algorithm, and the optimal working parameters are given in the form of Pareto frontier. When the angular velocity of the pickup device is 58.80 rad / s, the angular velocity of the film and impurity separation device is 40 rad / s, and the angle of the film and impurity separation device is 35.51º, the residue film impurity rate is 24.26%, and the residue film pickup rate is 83.67%.

[0009] Further, the multi-objective optimization method of the sprocket type residual film recycling machine working parameters is composed of the following steps:

[0010] Step one, parameter optimization:

[0011] The mathematical model for optimizing the parameters of the residual film recycling process is:

[0012]

[0013] In the formula, is the solution to the problem, is the objective function, is the equation constraint, is the inequality constraint;

[0014] The solution of the residual film recycling optimization problem involves accurate determination of the pickup device angular velocity, film and impurity separation device angular velocity, and film and impurity separation device angle. This process can be specifically represented as:

[0015]

[0016] In the formula, is the number of parameter combinations, is the pickup device angular velocity , is the film and impurity separation device angular velocity, is the film and impurity separation device angle ;

[0017] 1.1 Optimization objective

[0018] The core of optimizing residual film recycling is to improve the pickup rate and reduce the impurity rate. However, these two factors are usually contradictory in actual operation, so it is necessary to reasonably and accurately adjust the parameters of the recycling machine to achieve an optimal balance between the two;

[0019] The angular velocity of the pickup device, the angular velocity of the film and impurity separation device, and the angle of the film and impurity separation device are interrelated dependent variables. These three factors are the core control factors affecting the residual film pickup rate and the impurity rate. However, there is a contradiction between the two performance indicators of pickup rate and impurity rate. Pursuing a lower residual film impurity rate may be accompanied by a decrease in pickup rate, and vice versa.

[0020] 1.1.1 Residual film impurity rate

[0021] The mathematical expression of the residual film impurity rate is as follows:

[0022]

[0023] In the formula, represents the pickup device angular velocity , and the film debris separation device angular velocity is , and the film debris separation device angular velocity is , and the film debris separation device angular velocity is

[0024]

[0025] 1.1.2 Residual film pickup rate

[0026] The residual film pickup rate is defined as the ratio of the residual film area recovered by the residual film recovery machine to the total residual film area originally remaining in the farmland. This key indicator is influenced by the multiple synergies of the pickup device angular velocity, the film debris separation device angular velocity, and the angles of the two devices. The mathematical expression is as follows:

[0027]

[0028] In the formula, represents the corresponding residual film pickup rate under the working condition combination of the pickup device angular velocity , the film debris separation device angular velocity , and the film debris separation device angular velocity . In order to maximize the residual film pickup rate, the target function should be set to maximize the residual film pickup rate, and the target function is constructed as follows:

[0029]

[0030] 1.2 Constraints

[0031] During the operation of the residual film recovery machine, the key constraints are closely related to the angular velocity of the pickup device, the angular velocity of the film debris separation device, and the angles of the two devices. From the perspective of ensuring the stable and efficient performance of the residual film recovery machine, the above parameters must be accurately and strictly controlled within the appropriate value range.

[0032]

[0033] 1.3 Discrete element simulation and results

[0034] The farmland soil is composed of granular soil particles, and the residual film is made of high molecular polyethylene material and has strong plasticity, so that the finite element grid division is needed for subsequent analysis; the soil has cohesiveness and elastoplasticity, and the JKR model is selected to simulate the contact between the residual film and the soil; the mulch thickness t is between 0.01mm and 0.03mm, and this thickness range has an influence on the cohesive properties of the residual film and the soil; since the molecular particle size of the residual film is extremely small, it is difficult to directly carry out mechanical analysis, therefore, the enlarged molecule method is needed to solve the problem; in the application, the Hertz-Mindlin with Bond model is used to bond small molecules into large particles, so as to carry out quantitative analysis, and the EDEM software is used to construct the working model of the residual film recycling machine by combining the Hertz-Mindlin with JKR model.

[0035] In the application, the discrete element simulation technology is used to comprehensively simulate the residual film recycling process; the simulation experiment is carried out in the parameter range of the pickup device angular velocity of 40-60 rad / s, the film and impurity separation device angular velocity of 40-70 rad / s and the specific film and impurity separation device angle; after the rigorous experiment process, 363 groups of experimental data are obtained.

[0036] In order to verify the accuracy and reliability of the above-mentioned discrete element simulation model, the application starts from the two key indicators of residual film impurity content and pickup rate, and carries out comprehensive model verification; specifically, the simulation results of the application are compared with the experimental results of the residual film recycling machine with the same geometric structure in the "soil discrete element parameter calibration based on unconfined compressive strength test".

[0037] In terms of residual film impurity content, the average relative error is 5%; in terms of residual film pickup rate, the average relative error is 3%; in comprehensive consideration, the error level is within the acceptable range, which indicates that the discrete element simulation model constructed in the application has high credibility and effectiveness; optimization strategy

[0038] 1.4 Prediction model

[0039] ​When facing the nonlinear prediction task of three-input and two-output, neural network has the ability to directly construct a multiple-input and multiple-output model; compared with the traditional regression analysis method, neural network (NNs) has significant advantages; it does not need to follow the linear assumption, has greater freedom in model construction and parameter adjustment, and can more efficiently process complex nonlinear functional forms; especially in the specific application scenario of residual film recycling machine performance optimization, neural network prediction technology shows high potential and precision; it can accurately fit the complex relationship between residual film recycling machine performance and various influencing factors by deep learning on limited residual film recycling data, and then make accurate prediction on machine performance, and provide strong support for the optimization of residual film recycling machine performance;

[0040] The back propagation neural network is used for predicting the pickup rate and impurity rate of the residual film recycling, and the activation function of the hidden layer is a ReLU function:

[0041]

[0042] In the formula, y is the output of the neuron; x is the input variable;

[0043] The value range of the input value and the output value should be between 0 and 1; the normalization processing is carried out by using formula (9):

[0044] (9)

[0045] In the formula, is the original variable; is the minimum data value; is the maximum data value; is the maximum value of the standardized data, which is 1; is the minimum normalized data value, which is 0;

[0046] In the present application, the topology structure of the constructed neural network (NNs) is in the form of 3*15*15*2, and the specific architecture can be directly referred to Figure 5 ​;In the model training link, mean square error (MSE) is selected as the loss function of the neural network to quantify the deviation between the predicted value and the true value of the model; in order to optimize the parameters of the neural network, the stochastic gradient descent (SGD) optimizer is adopted, and the momentum factor is carefully set to 0.85, which can accelerate the convergence speed of the model and reduce the oscillation, and the learning rate is set to 0.05, which controls the step length of each parameter update; through simulation experiment, 605 groups of data are successfully obtained; in order to realize effective training and evaluation of the neural network, the data are divided according to the proportion of 70% and 30%, wherein 70% of the data constitute the training set of the neural network, which is used to guide the model to learn the data characteristics and rules in the supervised learning process; and the other 30% of the data are used as the test set, which is used to evaluate the generalization ability and prediction accuracy of the trained model on the unseen data;

[0047] 1.5 Optimization algorithm

[0048] In view of the difficulty of directly obtaining gradient information of the neural network prediction model, the genetic algorithm (GA) is adopted to explore the global optimal solution; during the execution of the genetic algorithm, the optimal individual in the population continuously evolves with the iteration of the population evolution, and then guides the algorithm to approach the global optimal solution;

[0049] As a heuristic optimization algorithm, the genetic algorithm has strong search ability and adaptability to complex problems, and is widely used in the solution of various optimization problems; in the field of multi-objective genetic algorithm, the NSGA-II algorithm is a major improvement of the original non-dominated sorting genetic algorithm (NSGA), and is especially suitable for solving complex nonlinear and large-scale multi-objective optimization problems; the NSGA-II algorithm reduces the computational complexity, effectively guarantees the population diversity, and significantly improves the overall quality of the population, thereby greatly improving the precision of the optimization result, enabling it to more efficiently explore the Pareto frontier in the multi-objective optimization scenario, and thus providing a better solution to the multi-objective optimization problem; based on the above advantages, the NSGA-II algorithm is selected to solve the multi-objective optimization problem in the residual film recycling process;

[0050] The optimization algorithm is programmed by using Python 3.11; during the implementation of the algorithm, a series of key parameters are carefully set: binary tournament selection strategy is adopted to select excellent individuals, and the population size is set to 150 to ensure sufficient coverage of the search space; uniform crossover is selected as the crossover operator, and the crossover probability is set to 0.5 to balance the exploration and development ability of the algorithm; at the same time, the Gaussian mutation operator is introduced to enhance the diversity of the population and avoid the algorithm falling into local optimal solution; the optimization algorithm process based on neural network (NNs) and genetic algorithm (GA);

[0051] First, population initialization is performed, and an initial population containing N individuals is randomly generated or constructed based on prior knowledge, each individual representing a potential solution, and a binary coding method is used; second, fitness evaluation and multi-objective processing are performed, the value of each individual on all objective functions is calculated to form a target vector, and the population is stratified according to the Pareto dominance relationship, if individual A is not inferior to individual B on all objectives and at least one objective is better, then A dominates B, and the non-dominated individual is located in the first layer (Pareto front), and the same is true for the subsequent layers; the distribution density of individuals in the same layer is measured by the crowding distance to avoid excessive aggregation of the solution set; then, high-quality individuals are retained and diversity is maintained to provide parents for crossover and mutation; several individuals are randomly selected from the population, and the non-dominated level and crowding distance are compared to determine the winner, which is used as the parent, and the current Pareto front individual is directly retained to prevent the loss of excellent solutions; then, new solutions are generated by simulating biological genetic operations through crossover and mutation operations to explore the solution space; then, the new generation population is selected from the parents and offspring to ensure convergence and diversity; the parents and offspring are combined and sorted by non-dominated level, and the new population is filled layer by layer; if a layer cannot be completely added, individuals with wider distribution are selected according to the crowding distance; finally, the termination condition is judged according to the number of iterations to provide the final approximate Pareto optimal solution set for the decision maker to choose;

[0052] Step two, results and analysis

[0053] To comprehensively and accurately evaluate the performance of the prediction model based on neural network, the invention carries out rigorous test work; through in-depth analysis and processing of experimental data, the prediction value and the actual value show a high degree of fit; specifically, for the residual film impurity rate index G and the residual film pickup rate index H, the average absolute percentage error of both is controlled within 10%; this result fully shows that the prediction model has excellent accuracy and reliability in the prediction task of residual film recycling related indicators, and can provide solid and reliable data support and theoretical basis for subsequent research and analysis and practical application;

[0054] The NSGA-II algorithm is used to solve the neural network (NNs) model of the residual film recycling machine, and the Pareto optimal front can be obtained, as shown in Figure 8 The vertical axis represents the residual film impurity rate, and the horizontal axis represents the residual film pickup rate.

[0055] The Pareto solution set can clearly show the conflict relationship between the residual film impurity rate G and the residual film pickup rate H, all points in the solution set are optimal solutions with conflict objective functions; under different initial conditions, these solutions provide diversified choices for performance optimization of the residual film recycling machine; it is worth noting that in the residual film recycling process, G and H show a positive correlation, that is, the increase of the residual film pickup rate of the residual film recycling machine will cause the corresponding increase of the residual film impurity rate;

[0056] From the mathematical aspect, each point on the curve represents a Pareto optimal solution; but when reaching the inflection point of the curve, with the further improvement of the residual film pickup rate, the residual film impurity rate will show a more significant upward trend; based on this, when manually screening the Pareto solution, the importance of each optimization target must be comprehensively, deeply and prudently weighed, so as to find the most suitable balance point between improving the residual film pickup rate and controlling the residual film impurity rate, and realize the performance optimization of the residual film recycling machine;

[0057] The Pareto front after 150 iterations is taken as the final result; six points are selected in the Pareto front solution set; in order to ensure the accuracy and reliability of the proposed strategy, four solution subsets are selected from the Pareto solution set, and the strategy is fully verified.

[0058] Compared with the prior art, the beneficial effects of the present application are:

[0059] The multi-objective optimization method for the sprocket type residual film recycling machine proposed in the application has achieved remarkable results. Compared with traditional optimization methods, the method closely meets the actual operation requirements. By using EDEM simulation technology, the residual film recycling process is simulated in depth to obtain a large amount of accurate data; by means of neural network modeling, the complex relationship between various factors is accurately analyzed, and an efficient prediction model of residual film impurity rate and net pickup rate is constructed; and then the NSGA-II algorithm is used to optimize the model in depth, and the Pareto frontier solution set is successfully obtained. This series of operations successfully frees us from excessive dependence on experience, and can provide highly adaptive parameter schemes for different operation conditions. After strict testing, the prediction model based on neural network has very low error, providing solid and reliable data support for optimization work. In the whole optimization process, EDEM simulation, neural network (NNs) and genetic algorithm (GA) are mutually coordinated and complementary, and together generate a large number of high-quality and diversified design solutions. This enables us to flexibly select the most suitable scheme according to different operation scenes and requirements in practical application. The sprocket type residual film recycling machine after optimization shows excellent performance under certain parameter combination. The residual film impurity rate is reduced by 28.16%, while the net pickup rate remains stable. This not only significantly improves the efficiency and quality of residual film recycling, but also effectively alleviates the problem of soil pollution caused by residual film, and injects strong impetus for the sustainable development of agriculture. BRIEF DESCRIPTION OF DRAWINGS

[0060] Figure 1 A structure diagram of residual film recycling machine, wherein: 1. soil shovel 2. film lifting device 3. film gathering device 4. soil roller 5. shell 6. film-soil separation device 7. film collection box 8. walking wheel 9. camshaft 10. belt 11. transmission chain;

[0061] Figure 2 Relationship diagram of various factors of residual film recycling;

[0062] Figure 3 Discrete element simulation results of residual film recycling machine, wherein Figure 3 A is the residual film impurity rate, Figure 3 B is the residual film net pickup rate;

[0063] Figure 4 Comparison of experimental value and simulation value, wherein Figure 4 A. Comparison of residual film impurity rate; Figure 4 B. Comparison of residual film net pickup rate;

[0064] Figure 5 BP neural network structure;

[0065] Figure 6 Optimization algorithm flowchart;

[0066] Figure 7 Error between EDEM simulation results and NNs prediction results, where, Figure 7 A is the residual film impurity rate error, Figure 7 B is the residual film pickup rate error;

[0067] Figure 8 Pareto frontier solution set of G and H;

[0068] Figure 9 Pareto frontier. DETAILED DESCRIPTION

[0069] The technical solutions of the present application will be further described in detail below in combination with the drawings and specific embodiments:

[0070] As Figures 1-9 shown,

[0071] Problem description

[0072] The core components of the residual film recycling machine include a traction frame, a machine frame, an earth-penetrating shovel, a transmission device, a residual film pickup device, a film impurity separation device, and a film collection box, as shown in FIG. 1. Among these components, the residual film pickup device and the film impurity separation device play a particularly crucial role. The residual film pickup device is composed of an earth-penetrating shovel, a spring tooth bar, a film gathering device, a soil crushing device, and a conveyor belt. When the residual film recycling machine is working, the earth-penetrating shovel penetrates about 200 mm into the ground, and the mixture of mulch film, soil, crop root stubble, and the like within this depth is scooped up together. The spring tooth bar lifts these mixtures to the conveyor belt, and the mulch film and impurities move upward synchronously with the conveyor belt. During the conveying process, the film gathering device gathers the mulch film to the middle position of the conveyor belt, and the soil crushing device crushes the soil to facilitate the separation of soil particles from the mulch film. At the same time, the conveyor belt continues to vibrate under the action of the camshaft, further shaking off small impurities, achieving preliminary film-impurity separation. Subsequently, the materials are transported to the film-impurity separation device for efficient separation by the friction belt, achieving the effect of secondary film-impurity separation. It is worth noting that in actual application, in order to reduce the impurity rate of residual film, a series of problems such as how to ensure a high residual film pickup rate often arise.

[0073] To effectively solve this problem, it is of great significance to analyze in-depth the influence of the device on the film-impurity recovery effect under various working conditions. The relationship between the residual film pickup rate, the residual film impurity rate, and the working parameters of the residual film recycling machine is shown in FIG. 2. In this process, the working parameters of the machine, the residual film pickup rate, and the residual film impurity rate are not independent of each other, but are dependent variables associated with each other. This characteristic makes it difficult for traditional single optimization methods to meet the demand for comprehensive optimization of parameters under multiple working conditions.

[0074] In actual operation scenarios, the angular velocity of the pickup device, the angular velocity of the film and impurity separation device, and the angle of the film and impurity separation device become the core control elements affecting the residual film pickup rate and the impurity rate. However, there is an inherent contradiction between the two key performance indicators: if more attention is paid to lower residual film impurity rate, it may lead to a decrease in residual film pickup rate; on the contrary, if too much attention is paid to the improvement of residual film pickup rate, it may lead to an increase in impurity rate. In order to resolve this contradiction, it is necessary to optimize the working parameters with the help of scientific methods to achieve all-round improvement of the residual film recovery effect. In view of this, the present application innovatively proposes a multi-objective control optimization strategy, which considers the residual film recovery process as an organic whole for global overall optimization. By using this strategy, the relationship between the residual film pickup rate and the impurity rate can be more effectively coordinated, and a more ideal residual film recovery effect can be achieved to meet the requirements of actual production operation for the quality and efficiency of residual film recovery.

[0075] Mathematical model of parameter optimization

[0076] The purpose of optimizing the parameters of the residual film recovery process is to maintain a low impurity rate while ensuring a high residual film pickup rate. Therefore, the parameter optimization problem of the residual film recovery process belongs to the category of constrained optimization problems, and its mathematical model can be expressed as:

[0077]

[0078] In the formula, is the solution of the problem, is the objective function, is the equation constraint, is the inequality constraint.

[0079] The solution of the residual film recovery optimization problem involves accurate determination of the angular velocity of the pickup device, the angular velocity of the film and impurity separation device, and the angle of the film and impurity separation device, which can be specifically represented as:

[0080]

[0081] In the formula, is the number of parameter combinations, is the angular velocity of the pickup device , is the angular velocity of the film and impurity separation device, is the angle of the film and impurity separation device .

[0082] 0.1.1 Optimization objective

[0083] The core of optimizing the residual film recycling lies in effectively improving the pickup rate and actually reducing the impurity rate. However, the two are usually contradictory in actual operation, so the parameters of the recycling machine must be reasonably and accurately adjusted to achieve the optimal balance between the two. As Figure 2 The relationship between the factors is close and complex. Based on this, the optimization goal is to significantly improve the residual film pickup rate and effectively reduce the residual film impurity rate to improve the overall effect of residual film recycling and meet the high standard requirements of actual production operations for the quality and efficiency of residual film recycling.

[0084] The angular velocity of the pickup device, the angular velocity of the film-impurity separation device, and the angle of the film-impurity separation device are interrelated dependent variables. These three become the core control factors that affect the residual film pickup rate and the impurity rate. However, there is a contradiction between the two performance indicators of the pickup rate and the impurity rate. Pursuing a lower residual film impurity rate may be accompanied by a decrease in the pickup rate, and vice versa.

[0085] 0.1.1.1 Residual film impurity rate

[0086] The residual film impurity rate refers to the proportion of soil, straw, and other impurities in the total residual film collected by the recycling machine. This rate is affected by multiple factors such as the angular velocity of the pickup device, the angular velocity of the film-impurity separation device, and the angle of the film-impurity separation device. The residual film impurity rate is directly related to the subsequent reuse value of the residual film. Its mathematical expression is as follows:

[0087]

[0088] In the formula, represents the corresponding residual film impurity rate when the angular velocity of the pickup device is , the angular velocity of the film-impurity separation device is , and the angle of the film-impurity separation device is To minimize the residual film impurity rate, the minimum residual film impurity rate should be selected as the objective function, as follows:

[0089]

[0090] 0.1.1.2 Residual film pickup rate

[0091] The residual film pickup rate is defined as the ratio of the area of the residual film collected by the recycling machine to the total area of the residual film originally left in the farmland. This key indicator is influenced by multiple factors such as the angular velocity of the pickup device, the angular velocity of the film-impurity separation device, and the angle of each device. Its mathematical expression is as follows:

[0092]

[0093] In the formula, The representative residual film pickup rate under the working condition combination of the pickup device angular velocity , the film and impurity separation device angular velocity , and the film and impurity separation device angle . In order to maximize the residual film pickup rate, the target function should be set to maximize the residual film pickup rate, that is, to maximize the residual film pickup rate:

[0094]

[0095] 0.1.2 Constraints

[0096] During the operation of the residual film recycling machine, the key constraints are closely related to the angular velocity of the pickup device, the angular velocity of the film and impurity separation device, and the angles of the two devices. From the perspective of ensuring the stable and efficient performance of the residual film recycling machine, the above parameters must be accurately and strictly controlled within the appropriate value range.

[0097]

[0098] 0.1.3 Discrete element simulation and results

[0099] The farmland soil is composed of granular soil particles, and the residual film is made of high molecular polyethylene material, which has strong plasticity. In order to facilitate subsequent analysis, it needs to be divided into finite element grids. The soil shows cohesive and elastic-plastic properties, and the JKR model is selected to simulate the contact between the residual film and the soil. According to the GB 1375-2017 standard, the thickness of the mulch film t is between 0.01mm and 0.03mm, which affects the cohesive properties of the residual film and the soil. Since the molecular particle size of the residual film is extremely small, direct mechanical analysis is difficult, so the method of expanding molecules is used to solve it. In this invention, the Hertz-Mindlin with Bond model is used to bond small molecules into large particles for quantitative analysis, and the EDEM software is used to build a working model of the residual film recycling machine by combining the Hertz-Mindlin with JKR model. Table 1 presents the basic parameters of the residual film and soil particles.

[0100] Table Basic parameters of residual film and soil particles

[0101]

[0102] In the present application, the discrete element simulation technology is used to simulate the whole process of residue film recycling. The simulation experiments are systematically carried out in the parameter range of the angle speed of the pickup device being 40-60 rad / s, the angle speed of the film and impurity separation device being 40-70 rad / s, and the specific angle of the film and impurity separation device. After a rigorous experimental process, a total of 363 experimental data are obtained. The detailed results of the simulation can be intuitively presented in Figure 3 .

[0103] In order to verify the accuracy and reliability of the above discrete element simulation model, the present application starts from the two key indicators of residue film impurity rate and pickup rate, and carries out comprehensive model verification. Specifically, the simulation results of the present application are compared with the experimental results of the residue film recycling machine with the same geometric structure in "Discrete element parameter calibration of soil based on unconfined compressive strength test", and the comparison is intuitively presented in FIG. 4.

[0104] As can be seen from the comparison results shown in FIG. 4, in terms of residue film impurity rate, the average relative error is 5%; in terms of residue film pickup rate, the average relative error is 3%. In general, such error level is within the acceptable range, indicating that the discrete element simulation model constructed by the present application has high credibility and effectiveness.

[0105] 0.1.4 Prediction model

[0106] When facing the nonlinear prediction task of three inputs and two outputs, neural networks have the ability to directly construct a multiple-input multiple-output model. Compared with traditional regression analysis methods, neural networks (NNs) have significant advantages. It does not need to follow the linear assumption, has more freedom in model construction and parameter adjustment, and can more efficiently handle complex nonlinear functional forms. Especially in the specific application scenario of residue film recycling machine performance optimization, neural network prediction technology shows high potential and precision. It can accurately fit the complex relationship between the performance of the residue film recycling machine and various influencing factors by deep learning on limited residue film recycling data, and then make accurate prediction on the machine performance, providing strong support for the optimization of the performance of the residue film recycling machine.

[0107] The back propagation neural network is used to predict the pickup rate and impurity rate of residue film recycling. The activation function of the hidden layer is the ReLU function:

[0108]

[0109] In the formula: is the output of the neuron; x is the input variable.

[0110] The value range of the input value and the output value should be between 0 and 1. The normalization processing is performed by using formula (9):

[0111] (9)

[0112] In the formula, is the original variable; is the minimum data value; is the maximum data value; is the maximum value of the standardized data, and the value is 1; is the minimum normalized data value, and the value is 0.

[0113] In the present application, the topology structure of the constructed neural network (NNs) is in the form of 3x15x15x2, and the specific architecture can be directly referred to Figure 5 . In the model training process, the mean square error (MSE) is selected as the loss function of the neural network, so as to quantify the deviation degree between the predicted value and the true value of the model. In order to optimize the parameters of the neural network, the stochastic gradient descent (SGD) optimizer is adopted, and the momentum factor is carefully set to 0.85, which can accelerate the convergence speed of the model and reduce the oscillation, and the learning rate is set to 0.05, which controls the step length of each parameter update. Through simulation experiment, 605 groups of data are successfully obtained. In order to realize the effective training and evaluation of the neural network, these data are divided according to the proportion of 70% and 30%, wherein 70% of the data constitutes the training set of the neural network, which is used to guide the model to learn the data characteristics and rules in the supervised learning process; and the other 30% of the data is used as the test set, which is specially used to evaluate the generalization ability and prediction accuracy of the trained model on the unseen data.

[0114] 0.1.5 Optimization algorithm

[0115] Since the neural network prediction model is difficult to directly obtain gradient information, the genetic algorithm (GA) is adopted to explore the global optimal solution. In the process of executing the genetic algorithm, with the continuous evolution and iteration of the population, the optimal individual in the population continuously evolves, and then guides the algorithm to approach the global optimal solution.

[0116] As a heuristic optimization algorithm, genetic algorithm is widely used in solving various optimization problems due to its strong search ability and adaptability to complex problems. In the field of multi-objective genetic algorithm, NSGA-II algorithm is a major improvement over the original non-dominated sorting genetic algorithm (NSGA), especially suitable for handling complex nonlinear, large-scale multi-objective optimization problems. NSGA-II algorithm significantly improves the accuracy of optimization results by skillfully reducing the computational complexity, effectively ensuring the diversity of the population and significantly improving the overall quality of the population, so that it can more efficiently explore the Pareto frontier in the multi-objective optimization scenario, thereby providing a better solution to the multi-objective optimization problem. Based on the above advantages, the present application selects NSGA-II algorithm to solve the multi-objective optimization problem in the residual film recycling process.

[0117] The present application uses Python 3.11 to program the optimization algorithm. During the implementation of the algorithm, a series of key parameters are carefully set: a binary tournament selection strategy is used to select excellent individuals, and the population size is set to 150 to ensure sufficient coverage of the search space; uniform crossover is selected as the crossover operator, and the crossover probability is set to 0.5 to balance the exploration and development ability of the algorithm; at the same time, a Gaussian mutation operator is introduced to enhance the diversity of the population and avoid the algorithm falling into a local optimal solution. Based on the optimization algorithm process of neural networks (NNs) and genetic algorithms (GA), see Figure 6 .

[0118] First, the population is initialized, and an initial population containing N individuals is randomly generated or constructed based on prior knowledge, with each individual representing a potential solution in binary encoding. Second, fitness evaluation and multi-objective processing are performed, the value of each individual on all objective functions is calculated to form a target vector, and the population is layered according to the Pareto dominance relationship. If individual A is not inferior to individual B in all objectives and at least one objective is better, then A dominates B, and non-dominated individuals are in the first layer (Pareto frontier), and the same applies to the rest. The distribution density of individuals in the same layer is measured by the crowding distance to avoid excessive aggregation of the solution set. Then, high-quality individuals are retained and diversity is maintained to provide parents for crossover and mutation. Several individuals are randomly selected from the population, and the non-dominated level and crowding distance are compared to determine the winner, which is used as the parent. The individuals in the current Pareto frontier are directly retained to prevent the loss of excellent solutions. Then, new solutions are generated by crossover and mutation operations to simulate biological genetic operations and explore the solution space. Then, the new generation of population is selected from the parents and children to ensure convergence and diversity. The parents and children are combined and sorted by non-dominated level, and the new population is filled layer by layer. If a layer cannot be completely added, individuals with wider distribution are selected according to the crowding distance. Finally, the termination condition is judged according to the number of iterations, and the final approximate Pareto optimal solution set is provided for the decision maker to choose.

[0119] 1 Results and Analysis

[0120] To comprehensively and accurately evaluate the performance of the prediction model based on neural networks, the invention carried out rigorous testing work. Through in-depth analysis and processing of experimental data, the error distribution trend between the simulation experiment results and the neural network prediction values is presented in the form of Figure 7. From the results shown in Figure 7, it can be clearly seen that the prediction value and the actual value show a high degree of fit. Specifically, detailed analysis is carried out for the residual film impurity rate index G and the residual film pickup rate index H. The average absolute percentage error of both is controlled within 10%. This result fully shows that the prediction model has excellent accuracy and reliability in the prediction task of residual film recycling related indicators, and can provide solid and reliable data support and theoretical basis for subsequent research and analysis and practical application.

[0121] The NSGA-II algorithm is used to solve the neural network (NNs) model of the residual film recycling machine, and the Pareto optimal frontier can be obtained, as shown in Figure 8 The vertical axis in the figure represents the residual film impurity rate, and the horizontal axis represents the residual film pickup rate.

[0122] The Pareto solution set can clearly show the conflict relationship between the residual film impurity rate G and the residual film pickup rate H. All points in the solution set are optimal solutions with conflicting objective functions. Under different initial conditions, these solutions provide a variety of choices for the performance optimization of the residual film recycling machine. It is worth noting that in the residual film recycling process, G and H show a positive correlation, that is, the increase of the residual film pickup rate of the residual film recycling machine will cause the corresponding increase of the residual film impurity rate.

[0123] From a mathematical perspective, each point on the curve represents a Pareto optimal solution. However, when reaching the inflection point of the curve, with the further improvement of the residual film pickup rate, the residual film impurity rate will show a more significant upward trend. Based on this, when manually selecting the Pareto solution, the importance of each optimization objective must be fully, deeply and prudently weighed, so as to find the balance point that best meets the actual needs between improving the residual film pickup rate and controlling the residual film impurity rate, and realize the optimization of the performance of the residual film recycling machine.

[0124] The Pareto frontier after 150 iterations is taken as the final result. Six points are selected from the Pareto frontier solution set, as shown in To ensure the accuracy and reliability of the proposed strategy, four subsets of solutions are selected from the Pareto solution set, and the strategy is fully verified. As shown in Table 2.

[0125] Table 2 shows the comparison of the predicted values and the simulation values of the angular velocity of the pickup device, the angular velocity of the film impurity separation device and the angle of the film impurity separation device under four different optimal parameters. It can be clearly seen that the prediction errors of the residual film impurity content under three different parameters are 5.91%, 6.06%, 2.97% and 0.36%, all of which are lower than 7%. The prediction errors of the residual film pickup rate are 1.64%, 5.21%, 2.06% and 5.29%, all of which are lower than 6%. Table Performance comparison of residual film recycling machine

[0126]

[0127] Through inspection, the prediction accuracy of the residual film impurity content and the pickup rate is in a reasonable range. At the same time, the change trend of the residual film impurity content and the pickup rate under four different parameters is analyzed, and the results show that the change is highly consistent with the prediction trend of the model. This finding further proves that the optimization strategy adopted by the present application has high reliability and effectiveness, and can provide scientific and accurate guidance for residual film recycling work.

[0128] In the field of parameter optimization of residual film recycling machines, the multi-objective optimization strategy of the chain tooth type residual film recycling machine proposed by the present application innovatively combines neural networks and genetic algorithms, maximizing the advantages of both. Neural networks, with their powerful nonlinear processing capabilities, have shown excellent performance in the residual film recycling process. Through deep learning of massive simulation data, it can accurately capture the complex nonlinear relationship between working parameters and residual film impurity content and pickup rate, and then build a high-precision mapping model. It provides solid data support and theoretical basis for accurately predicting and evaluating the residual film recycling effect under different parameter combinations, so that we can have a rough idea of various possible recycling situations before actual operation. The NSGA-II algorithm in genetic algorithm solves the problem of multi-objective optimization and can search for Pareto front solution set in a huge and complex solution space with high efficiency. It provides decision makers with a variety of scientific and reasonable parameter selection schemes. No matter how the actual demand changes, decision makers can find the optimal solution that fits their own situation, greatly enhancing the adaptability and flexibility of the optimization strategy to different application scenarios.

[0129] From the point of view of practical application, the multi-objective optimization strategy proposed in the application shows very high feasibility and reliability. After a series of strict verification processes, we found that both the residual film impurity rate and the pickup rate showed a satisfactory high consistency between the predicted value and the simulation experimental value, and the error was strictly controlled within a lower range. This excellent performance fully proves that the strategy can provide scientific, reliable and accurate guidance for the design and development of residual film recycling machines and actual operation. Agricultural producers can flexibly select the most suitable working parameters according to the actual situation of residual film pollution in their own farmland and subsequent treatment plans.

[0130] The multi-objective optimization method for sprocket type residual film recycling machine proposed in the application has achieved remarkable results. Compared with traditional optimization methods, this method closely meets the actual operation requirements. By using EDEM simulation technology, the residual film recycling process is simulated in depth to obtain a large amount of accurate data; by using neural network modeling, the complex relationship between various factors is accurately analyzed to build an efficient prediction model of residual film impurity rate and pickup rate; and by using NSGA-II algorithm, the model is deeply optimized to successfully obtain the Pareto frontier solution set. This series of operations successfully frees us from excessive dependence on experience and provides highly adaptive parameter schemes for different operation conditions. After strict testing, the prediction model based on neural network has very low error, providing solid and reliable data support for optimization work. In the entire optimization process, EDEM simulation, neural network (NNs) and genetic algorithm (GA) cooperate with each other and complement each other's advantages to generate a large number of high-quality and diversified design solutions. This enables us to flexibly select the most suitable solution according to different operation scenarios and requirements in practical applications. The optimized sprocket type residual film recycling machine exhibits excellent performance under certain parameter combinations. The residual film impurity rate is reduced by 28.16%, while the pickup rate remains stable. This not only significantly improves the efficiency and quality of residual film recycling, but also effectively alleviates the problem of residual film pollution in the soil, injecting strong impetus for the sustainable development of agriculture. The application provides key technical support and valuable practical reference for the innovation of residual film recycling technology, and is expected to promote further development and breakthroughs in this field.

[0131] The above is only a specific implementation of the application, but the protection scope of the application is not limited thereto, any changes or substitutions without creative labor should be covered within the protection scope of the application. Therefore, the protection scope of the application should be limited by the protection scope defined in the claims.

Claims

1. A multi-objective optimization method for the operating parameters of a chain-tooth residual film recycling machine, characterized in that, Taking a chain-tooth type residual film recovery machine as the research object, the angular velocity of the pickup device, the angular velocity of the membrane impurity separation device, and its working angle were set as key variables to be optimized. In order to obtain the optimal residual film impurity content and pickup rate, discrete element simulation was performed using EDEM, and a backpropagation neural network was trained using simulated data to establish a prediction model for residual film impurity content and pickup rate. The prediction results of the neural network were optimized using a non-dominated multi-objective genetic algorithm, and the optimal working parameters were given in the form of Pareto front. When the angular velocity of the pickup device was 58.80 rad / s, the angular velocity of the membrane impurity separation device was 40 rad / s, and the angle of the membrane impurity separation device was 35.51º, the residual film impurity content was 24.26%, and the residual film pickup rate was 83.67%.

2. The method according to claim 1, characterized in that: The multi-objective optimization method for the operating parameters of a chain-type residual film recycling machine consists of the following steps: Step 1: Parameter Optimization The mathematical model for optimizing the parameters of the residual film recycling process is as follows: ； In the formula, It is the solution to the problem. It is the objective function. It is an equation constraint. It is an inequality constraint; Solving the optimization problem of residual membrane recycling involves the precise determination of the angular velocity of the pickup device, the angular velocity of the membrane impurity separation device, and the angle of the membrane impurity separation device. This process can be specifically expressed as follows: ； In the formula, For the number of parameter combinations, Angular velocity of the pickup device , For the membrane impurity separation device, angular velocity Angle of membrane impurity separation device ; Optimization Objective The key to optimizing residual film recycling lies in improving the collection rate and reducing the impurity rate; however, these two are often contradictory in actual operation. Therefore, it is necessary to make reasonable and precise adjustments to various parameters of the recycling machine to achieve the optimal balance between the two. The angular velocity of the pickup device, the angular velocity of the membrane impurity separation device, and the angle of the membrane impurity separation device are interrelated dependent variables. These three factors are the core control factors affecting the residual membrane pickup rate and impurity content. However, there is a contradiction between the two performance indicators of pickup rate and impurity content. Pursuing a lower residual membrane impurity content may be accompanied by a decrease in pickup rate, and vice versa. Impurity content of residual film The mathematical expression for the impurity content of the residual film is as follows: ； In the formula, This indicates that when the angular velocity of the pickup device is The angular velocity of the membrane impurity separation device is And the angle of the membrane impurity separation device is The impurity content of the residual film at that time; to minimize the impurity content of the residual film, the objective function should be to minimize the impurity content of the residual film, as follows: ； Residual film collection rate The residual film collection rate is defined as the ratio of the area of ​​residual film recovered by the residual film recycling machine to the total area of ​​residual film originally remaining in the farmland. This key indicator is affected by multiple factors, including the angular velocity of the pickup device, the angular velocity of the film-impurity separation device, and the angles of both devices. Its mathematical expression is as follows: ； In the formula, This represents the angular velocity of the pickup device. The angular velocity of the membrane impurity separation device is And the angle of the membrane impurity separation device is The residual film collection rate under the given operating conditions; to maximize the residual film collection rate, the objective function should be constructed by setting a target that maximizes the residual film collection rate. ； constraint During the operation of the residual film recycling machine, the key constraints are closely related to the angular velocity of the pickup device, the angular velocity of the membrane impurity separation device, and the angles of these two devices; From the perspective of ensuring the stable and efficient performance of the residual film recycling machine, the above parameters must be precisely and strictly controlled within an appropriate range of values. ； Discrete Element Simulation and Results Farmland soil is composed of granular soil particles, while the residual film is made of high-molecular-weight polyethylene material, which has strong plasticity. For ease of subsequent analysis, finite element meshing is required. Soil exhibits both cohesive and elastoplastic properties; therefore, the JKR model is used to simulate the contact between the residual film and the soil. The film thickness t is between 0.01 mm and 0.03 mm, and this thickness range affects the adhesion characteristics between the residual film and the soil. Due to the extremely small particle size of the residual film molecules, direct mechanical analysis is difficult; therefore, an enlarged molecular method is needed. In this invention, the Hertz-Mindlin with Bond model is used to bond small molecules into large particles for quantitative analysis. Using EDEM software and the Hertz-Mindlin with JKR model, a working model of the residual film recycling machine is constructed. In this invention, a comprehensive simulation of the residual membrane recycling process was carried out using discrete element method (DEM) simulation technology. Simulation experiments were systematically conducted within the parameter ranges of 40-60 rad / s for the pickup device angular velocity, 40-70 rad / s for the membrane impurity separation device angular velocity, and a specific angle for the membrane impurity separation device. A total of 363 sets of experimental data were obtained through a rigorous experimental procedure. To verify the accuracy and reliability of the aforementioned discrete element simulation model, this invention conducts a comprehensive model verification process, focusing on two key indicators: residual film impurity content and cleanliness. Specifically, the simulation results of this invention are used to verify the model. A detailed and in-depth comparative analysis was conducted on the experimental results of a residual film recovery machine with the same geometry as that used in "Soil Discrete Element Parameter Calibration Based on Unconfined Compressive Strength Test". Regarding the impurity content of the residual film, the average relative error is 5%; regarding the residual film cleanliness rate, the average relative error is 3%. Taking all factors into consideration, such error levels are within an acceptable range, indicating that the discrete element simulation model constructed in this invention has high credibility and effectiveness. Optimize the prediction model When faced with nonlinear prediction tasks with three inputs and two outputs, neural networks have the ability to directly construct multi-input multi-output models. Compared with traditional regression analysis methods, neural networks (NNs) have significant advantages. They do not need to follow linear assumptions, have greater freedom in model construction and parameter tuning, and can handle complex nonlinear functional forms more efficiently. Especially in the specific application scenario of performance optimization of residual film recycling machines, neural network prediction technology has shown extremely high potential and accuracy. It can accurately fit the complex relationship between the performance of residual film recycling machines and various influencing factors by performing deep learning on limited residual film recycling data, thereby making accurate predictions on machine performance and providing strong support for the optimization of residual film recycling machine performance. A backpropagation neural network is used to predict the recovery rate and impurity content of the residual film; the activation function of the hidden layer is the ReLU function. ； In the formula: The output of the neuron is x; the input variable is x. The input and output values ​​should be between 0 and 1; normalization is performed using equation (9): (9)； In the formula, These are the original variables; The minimum data value; The maximum data value; The maximum value of the standardized data is 1. The minimum normalized data value is 0; In this invention, the constructed neural network (NNs) has a topology of 3×15×15×2. During model training, mean squared error (MSE) is used as the loss function to quantify the deviation between the model's predicted values ​​and the actual values. To optimize the neural network parameters, a stochastic gradient descent (SGD) optimizer is employed, with the momentum factor carefully set to 0.85 to accelerate convergence and reduce oscillations. The learning rate is set to 0.05, controlling the step size for each parameter update. Through simulation experiments, 605 sets of data were successfully obtained. To achieve effective training and evaluation of the neural network, these data were divided into a 70% and 30% set. 70% of the data constitutes the training set, used to guide the model in learning data features and patterns during supervised learning. The remaining 30% serves as the test set, specifically used to evaluate the generalization ability and prediction accuracy of the trained model on unseen data. Optimization Algorithm Given that neural network prediction models have difficulty directly obtaining gradient information, this invention uses a genetic algorithm (GA) to explore its global optimum. During the execution of the genetic algorithm, as the population continuously evolves and iterates, the best individual in the population continues to evolve, thereby guiding the algorithm to approach the global optimum. Genetic algorithms, as a heuristic optimization algorithm, are widely used in solving various optimization problems due to their powerful search capabilities and adaptability to complex problems. In the field of multi-objective genetic algorithms, the NSGA-II algorithm is a significant improvement over the original non-dominated sorting genetic algorithm (NSGA), and is particularly suitable for handling complex nonlinear, large-scale multi-objective optimization problems. The NSGA-II algorithm significantly improves the accuracy of optimization results by cleverly reducing computational complexity, effectively ensuring population diversity, and significantly improving the overall quality of the population. This allows it to explore the Pareto front more efficiently in multi-objective optimization scenarios, thus providing better solutions for multi-objective optimization problems. Based on the above advantages, this invention selects the NSGA-II algorithm to solve the multi-objective optimization problem in the residual film recycling process. This invention utilizes Python 3.11 to program the optimization algorithm. During the algorithm implementation, a series of key parameters were carefully set: a binary tournament selection strategy was employed to screen superior individuals, and the population size was set to 150 to ensure sufficient coverage of the search space; uniform crossover was selected as the crossover operator, and the crossover probability was set to 0.5 to balance the algorithm's exploration and development capabilities; simultaneously, a Gaussian mutation operator was introduced to enhance population diversity and prevent the algorithm from getting trapped in local optima; the optimization algorithm flow is based on neural networks (NNs) and genetic algorithms (GA). First, population initialization is performed. An initial population of N individuals is randomly generated or constructed based on prior knowledge, with each individual representing a potential solution using binary encoding. Second, fitness evaluation and multi-objective processing are conducted. The values ​​of each individual on all objective functions are calculated to form an objective vector. The population is then stratified according to Pareto dominance. If individual A is not inferior to individual B on all objectives and is superior to individual B on at least one objective, then A dominates B. Non-dominated individuals are located in the first stratum (Pareto front), and so on. Crowding distance is then used to determine the optimal population size. The Pareto distance is used to measure the distribution density of individuals within the same layer, avoiding excessive clustering of solutions. Then, high-quality individuals are retained to maintain diversity, providing parents for crossover and mutation. Several individuals are randomly selected from the population, and their quality is compared based on non-dominated level and crowding distance; the winners become parents, and individuals at the current Pareto front are directly retained to prevent the loss of excellent solutions. Next, crossover and mutation operations simulate biological genetic operations to generate new solutions and explore the solution space. Then, a new generation of population is selected from the parents and offspring to ensure convergence and diversity. Parents and offspring are merged, sorted by non-dominated level, and the new population is filled layer by layer. If a layer cannot be fully added, individuals with a wider distribution are selected based on crowding distance. Finally, a termination condition is determined based on the number of iterations, providing a final approximate Pareto optimal solution set for decision-makers to choose from. Step 2: Results and Analysis To comprehensively and accurately evaluate the performance of the prediction model built on neural networks, this invention conducted rigorous testing. Through in-depth analysis and processing of experimental data, the predicted values ​​and actual values ​​showed a high degree of agreement. Specifically, a detailed analysis was conducted on the residual film impurity rate index G and the residual film cleanliness index H, and the average absolute percentage error of both was controlled within 10%. This result fully demonstrates that the prediction model has excellent accuracy and reliability in predicting residual film recycling-related indicators, and can provide solid and reliable data support and theoretical basis for subsequent research, analysis, and practical applications. The Pareto optimal front can be obtained by solving the neural network (NNs) model of the residual film recycling machine using the NSGA-II algorithm. The Pareto solution set can clearly show the conflicting relationship between the impurity content of the residual film G and the residual film collection rate H. All points in this solution set are the optimal solutions with conflicting objective functions. Under different initial conditions, these solutions provide diverse options for optimizing the performance of the residual film recycling machine. It is worth noting that during the residual film recycling process, G and H are positively correlated, that is, the improvement of the residual film collection rate of the residual film recycling machine will lead to a corresponding increase in the impurity content of the residual film. From a mathematical perspective, each point on the curve represents a Pareto optimal solution. However, after reaching the inflection point of the curve, as the residual film collection rate further increases, the residual film impurity rate will show a more significant upward trend. Based on this, when manually screening for Pareto solutions, it is necessary to comprehensively, deeply, and carefully weigh the importance of each optimization objective in order to find the balance point that best meets the actual needs between improving the residual film collection rate and controlling the residual film impurity rate, thereby optimizing the performance of the residual film recycling machine. The Pareto front after 150 iterations is taken as the final result; six points are selected from the Pareto front solution set; to ensure the accuracy and reliability of the proposed strategy, four solution subsets are selected from the Pareto solution set, and the strategy is fully verified.

Images