A multi-objective prediction and optimization method for rotary abrasive waterjet polishing

Abstract

The application discloses a kind of multi-objective prediction and optimization method for rotary abrasive water jet polishing, S1: obtaining processing parameter and processing result, obtains experimental data set;S2: the experimental data set is mixed and is divided according to proportion training set and test set;S3: the hyperparameter of machine learning integrated model is optimized using Bayesian optimization algorithm;S4: the experimental data set after mixing and division is input into optimized machine learning integrated model to train and predict;S5: according to R2 and RMSE evaluation index, select the optimal regression prediction model;S6: the optimal regression prediction model is used as the objective function of multi-objective optimization algorithm to carry out multi-objective optimization;S7: the optimal multi-objective optimization algorithm is selected by hyper volume evaluation index;S8: according to entropy weight method, select the optimal solution in the pareto solution set obtained by optimal multi-objective optimization algorithm.The application realizes the ultra-precision polishing of the inner surface of ultra-precision part, and greatly improves the processing efficiency.

Description

Technical Field

[0001] This invention relates to the field of abrasive waterjet polishing technology, specifically an integrated multi-objective prediction and optimization method for machining parameters of rotating abrasive waterjet polishing. Background Technology

[0002] With the increasing demand for ultra-precision components, the requirements for surface quality and shape accuracy of parts are also becoming increasingly stringent. Various surface polishing technologies have been developed to address this issue, such as magnetorheological polishing, electrochemical polishing, and abrasive waterjet polishing. Among these, abrasive waterjet polishing, being a cold-working technology, offers advantages such as high flexibility, low cost, environmental friendliness, and no tool wear. Traditional abrasive waterjet polishing uses a fixed-angle jet polishing method, which cannot handle the needs of polishing jets at different angles. Furthermore, the processing parameters of abrasive waterjet polishing have a significant impact on the polishing effect and processing efficiency. Traditional abrasive waterjet polishing processing parameters are often derived from operator experience and trial and error, making it difficult to find suitable parameter combinations based on specific requirements. Extensive trial and error often leads to wasted resources and increased time costs. While some studies have established mathematical models for abrasive waterjet polishing, many complex phenomena during the jet polishing process are difficult to accurately describe and explain using mathematical models. In addition, most empirical models contain many unknown parameters, making them difficult to use in production practice. Further improvement of these models is essential.

[0003] Therefore, there is an urgent need in this field to study a rotating abrasive waterjet polishing device and a method for predicting and optimizing processing parameters, so as to improve the processing effect and efficiency of abrasive waterjet polishing and thus meet the surface quality requirements of ultra-precision components. Summary of the Invention

[0004] The purpose of this invention is to provide a multi-objective prediction and optimization method for rotating abrasive waterjet polishing to solve the above-mentioned existing technical problems, realize ultra-precision polishing of the inner surface of ultra-precision parts, greatly improve processing efficiency, reduce time, manpower and economic costs, greatly improve the surface quality of the processed parts, and provide a method reference for subsequent surface polishing methods.

[0005] In view of the above-mentioned defects and improvement needs of the existing technology, the technical solution of the present invention is as follows:

[0006] A multi-objective prediction and optimization method for rotating abrasive waterjet polishing includes the following steps:

[0007] S1: The processing parameters and results of the polishing process of the rotating abrasive waterjet polishing device were obtained using the Box-Behnken response surface methodology, and the experimental dataset was obtained.

[0008] S2: Shuffle the experimental dataset and divide it into training and test sets according to the ratio;

[0009] S3: Use the Bayesian optimization algorithm to optimize the hyperparameters of the machine learning ensemble model;

[0010] S4: Input the shredded and partitioned experimental datasets into the optimized machine learning ensemble model for training and prediction, and construct a regression prediction model for processing parameters and processing results;

[0011] S5: Select the optimal regression prediction model based on the evaluation indicators of R2, RMSE and MAE;

[0012] S6: The optimal regression prediction model is used as the objective function of the multi-objective optimization algorithm to perform multi-objective optimization and obtain the Pareto solution set of surface roughness and material removal rate;

[0013] S7: Select the optimal multi-objective optimization algorithm through hypervolume evaluation index;

[0014] S8: Select the optimal solution from the Pareto solution set obtained by the optimal multi-objective optimization algorithm according to the entropy weight method.

[0015] The processing parameters of the experimental dataset in S1 include jet pressure, workpiece rotation speed, abrasive particle size, abrasive concentration, and polishing time; the processing results of the experimental dataset include surface roughness and material removal rate.

[0016] The machine learning ensemble model in S3 includes three machine learning models: CatBoost, XGBoost, and LightGBM.

[0017] The R2 and RMSE evaluation indicators in S5 are as follows:

[0018]

[0019]

[0020]

[0021] Where N represents the number of samples, y i Represents the actual value. R² represents the predicted value, and R² represents the sample mean. When evaluating machine learning performance, R² measures how well the model fits the data. R² ≤ 1. The closer R² is to 1, the better the machine learning model's predictive performance. RMSE measures the deviation between the predicted value and the true value. The smaller the deviation, the closer the predicted value is to the true value. MAE measures the absolute error between the predicted value and the true value. The smaller the error, the closer the predicted value is to the true value.

[0022] The objective function of the multi-objective optimization algorithm in S6 is:

[0023] min F(X)=[f Ra (X), -f MRR (X)] T

[0024]

[0025] X = (A, B, C, D, E) T

[0026]

[0027] 30≤A≤90, 5≤B≤15, 0.6≤C≤1.8

[0028] 1000≤D≤3000, 1≤E≤20

[0029] Where f Ra (X) represents the surface roughness regression prediction model from the optimal machine learning model, f MRR (X) represents the material removal rate regression prediction model from the optimal machine learning model. X is a feature vector composed of 5 variables x in the feasible space. The constraint range of the 5 variables was determined through pilot-scale experiments as A∈[30, 90], B∈[5, 15], C∈[0.6, 1.8], D∈[1000, 3000], and E∈[1, 20].

[0030] The formula for evaluating the super-volume in S7 is as follows:

[0031]

[0032] Where P is the final solution set; r∈R M It is a reference point in the target space; It is the solution set that dominates the reference point r; vol(s, r) represents the volume between s and r. The higher the HV value, the better the convergence and diversity of the algorithm.

[0033] The rotating abrasive waterjet polishing device in S1 includes a metering pump, a damper, a back pressure valve, a pressure gauge, a pressure regulating valve, a polishing moving system, a polishing tool head, a polishing rotating system, a mixing system, and a CNC control system. The damper is connected to the metering pump, and the back pressure valve is connected to the damper; the damper and back pressure valve are used to stabilize the outlet pressure of the metering pump. The pressure gauge is connected to the back pressure valve and is used to observe and record the output pressure of the polishing system. The pressure regulating valve is connected to the pressure gauge and is used to regulate the output pressure of the polishing system. The pressure regulating valve is connected to the polishing moving system via a hose and is used to drive the polishing tool head to move along the z-axis. The polishing rotating system is used to realize the rotational movement of the workpiece. The mixing system is used to fully mix the solid abrasive particles and the liquid aqueous solution in the polishing fluid to achieve a uniform distribution of abrasive particles in the polishing fluid. The CNC control system is used to control the power execution and control of the polishing moving system, the polishing rotating system, and the mixing system.

[0034] Furthermore, the polishing moving system includes a servo motor, a linear slide, and a polishing tool head. The polishing tool head is fixed on the linear slide block, and the servo motor is used to drive the linear slide to move the polishing tool head linearly along the z-axis in both positive and negative directions.

[0035] The polishing rotation system includes a polishing rotation system servo motor and a workpiece fixing seat. The polishing rotation system servo motor is connected to the workpiece fixing seat, and the workpiece is fixed on the workpiece fixing seat. The polishing rotation system servo motor drives the workpiece fixing seat to rotate the workpiece.

[0036] The mixing system includes a pneumatic motor, an air pump, a propeller, and a polishing slurry supply tank. The air pump provides power to the pneumatic motor, which is connected to the propeller to drive the propeller to rotate, thereby stirring the polishing slurry in the polishing slurry supply tank to ensure that the abrasive particles in the polishing slurry are evenly distributed and to prevent abrasive particle deposition.

[0037] Furthermore, the polishing tool head includes a polishing fluid linear jet orifice and a coaxial positioning head. The polishing tool is driven by the polishing moving system and holds the workpiece fixing seat in the center positioning groove of the polishing rotating system to ensure that the workpiece and the polishing tool head remain coaxial when rotating.

[0038] Furthermore, the polishing process of the rotary abrasive waterjet polishing device is as follows:

[0039] 1) First, fix the workpiece on the workpiece holder. Then, control the servo motor through the control system to drive the linear slide to move the polishing tool head down along the z-axis and press against the positioning groove of the workpiece holder, so that the polishing tool head moves into place.

[0040] 2) Prepare the polishing liquid in the polishing liquid supply tank according to the required ratio. Use the pneumatic motor of the control system to drive the stirring propeller to start mixing and stirring the polishing liquid to ensure the uniformity of the polishing liquid.

[0041] 3) The metering pump is started by the control system to drive the polishing fluid from the polishing fluid supply tank through the damper and back pressure valve for pressure stabilization, and finally enter the polishing tool head. It is sprayed from the linear jet nozzle of the polishing tool head onto the inner wall of the workpiece. The polishing fluid then flows back to the polishing fluid supply tank through the return column to enter the next cycle.

[0042] 4) The control system sets the speed of the servo motor in the polishing rotation system, and starts the servo motor in the polishing rotation system to drive the workpiece holder to rotate the workpiece;

[0043] 5) The angle between the polishing liquid spray path and the inner wall of the workpiece can be controlled by controlling the rotation speed of the workpiece, thereby achieving stepless adjustment of the jet angle.

[0044] The beneficial effects of this invention are as follows: This invention achieves ultra-precision polishing of the inner surface of ultra-precision parts, which greatly improves processing efficiency, reduces time, manpower and economic costs, greatly improves the surface quality of the processed parts, and provides an important reference for subsequent surface polishing methods. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the rotating abrasive waterjet polishing device of the present invention;

[0046] Figure 2 This is a flowchart of the method of the present invention;

[0047] Figure 3 A cross-sectional view of the polishing tool head, workpiece, and workpiece holder in the working position;

[0048] In the diagram: 1. Metering pump; 2. Damper; 3. Back pressure valve; 4. Pressure gauge; 5. Pressure regulating valve; 6. Polishing tool head; 601. Workpiece; 602. Polishing fluid linear jet nozzle; 603. Coaxial positioning head; 7. Linear slide; 8. Servo motor; 9. Linear slide; 10. Workpiece holder; 11. Servo motor for polishing rotation system; 12. Control cabinet; 13. Pneumatic motor; 14. Propeller; 15. Polishing fluid supply tank; 16. Air pump. Detailed Implementation

[0049] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0050] This invention designs a rotating abrasive waterjet polishing device, including a metering pump 1, a damper 2, a back pressure valve 3, a pressure gauge 4, a pressure regulating valve 5, a polishing moving system, a polishing tool head 6, a polishing rotating system, a mixing system, and a control system.

[0051] The metering pump is a single-port diaphragm metering pump with a maximum flow rate of 3380 L / h and a standard operating pressure of 200 bar, used to drive the polishing slurry. The damper connected to the metering pump is an RXMZ-L6.0 / 2.5 diaphragm pulse damper; the back pressure valve connected to the damper is an RXBF-L / 1.0 back pressure valve; the damper and back pressure valve are used to stabilize the outlet pressure of the metering pump; the pressure gauge connected to the back pressure valve is used to observe and record the output pressure of the polishing system; the pressure regulating valve connected to the pressure gauge is used to regulate the output pressure of the polishing system; the pressure regulating valve is connected to the polishing moving system through a hose, used to drive the polishing tool head to move along the z-axis; the polishing rotation system is used to realize the rotational movement of the workpiece; the mixing system is used to fully mix the solid abrasive particles and liquid aqueous solution in the polishing slurry to achieve uniform distribution of abrasive particles in the polishing slurry; the CNC control system is used to control the power execution and control of the polishing moving system, the polishing rotation system, and the mixing system.

[0052] The polishing moving system includes a servo motor A8, a linear slide 9, and a polishing tool head 6. The polishing tool head 6 is fixed on the linear slide 7. The servo motor A8 drives the linear slide 7 to move the polishing tool head 6 linearly along the z-axis in both directions. Preferably, the polishing tool head 6 includes a polishing fluid linear jet port 602 and a coaxial positioning head 603. Specifically, the polishing tool head has polishing fluid linear jet ports at 120° intervals on its side. The coaxial positioning head is coaxially connected to the polishing tool head via a threaded connection. The workpiece holder is rigidly connected to the stepper motor shaft in the polishing rotation system. The workpiece is rigidly connected to the workpiece holder via a positioning groove and a positioning pin. The polishing tool head 6 is driven by the polishing moving system to press against the center positioning groove of the workpiece holder in the polishing rotation system to ensure that the workpiece 601 and the polishing tool head remain coaxial during rotation.

[0053] The polishing rotation system includes a polishing rotation system servo motor 11 and a workpiece fixing seat 10. The polishing rotation system servo motor 11 is connected to the workpiece fixing seat 10, and the workpiece is fixed on the workpiece fixing seat. The servo motor B11 drives the workpiece fixing seat to rotate the workpiece.

[0054] The mixing system includes a pneumatic motor 13, an air pump 16, a propeller 14, and a polishing slurry supply tank 15. The air pump provides power to the pneumatic motor, which is connected to the stirring propeller to drive the stirring propeller to rotate, thereby stirring the polishing slurry in the polishing slurry supply tank, so that the abrasive particles in the polishing slurry are evenly distributed and abrasive particles are avoided from depositing.

[0055] The present invention also provides a rotating abrasive waterjet polishing method, applied to the above-mentioned rotating abrasive waterjet polishing apparatus, comprising the following steps:

[0056] First, fix the workpiece on the workpiece holder. Then, control the servo motor through the CNC control system to drive the linear slide to move the polishing tool head down along the z-axis and press against the positioning groove of the workpiece holder, so that the polishing tool head moves into place.

[0057] Prepare the polishing liquid in the polishing liquid supply tank according to the required ratio. The pneumatic motor of the control system drives the propeller to start mixing and stirring the polishing liquid to ensure the uniformity of the polishing liquid.

[0058] The control system starts the metering pump to drive the polishing fluid from the polishing fluid supply tank. The fluid is stabilized by the damper and the back pressure valve and finally enters the polishing tool head. The fluid is sprayed from the linear jet nozzle of the polishing tool head onto the inner wall of the workpiece. The polishing fluid then flows back to the polishing fluid supply tank through the return column to enter the next cycle.

[0059] The control system sets the speed of the servo motor in the polishing rotation system, and starts the servo motor in the polishing rotation system to drive the workpiece holder to rotate the workpiece;

[0060] The angle between the polishing slurry spray path and the inner wall of the workpiece can be controlled by controlling the rotation speed of the workpiece, thereby achieving stepless adjustment of the jet angle.

[0061] This invention mainly involves using a polishing slurry formed by mixing abrasive and water, which is then output under high pressure and impacted onto the surface of a workpiece to perform micro-cutting, thereby achieving a polishing effect. At the same time, by controlling the rotation speed of the workpiece, the jet angle of the polishing slurry can be steplessly adjusted to meet different polishing angle requirements.

[0062] In addition to the rotary abrasive waterjet polishing method of the present invention, a method for predicting and optimizing rotary abrasive waterjet polishing processing parameters and multi-objective optimization is also provided. The aim is to propose a method for predicting and optimizing processing parameters to improve the accuracy and efficiency of rotary abrasive waterjet polishing, so as to meet the surface quality requirements of ultra-precision component processing.

[0063] To achieve the above objectives, according to one aspect of the present invention, an integrated multi-objective prediction and optimization method for parameters of rotating abrasive waterjet polishing is provided, the method comprising:

[0064] S1: The experimental dataset was obtained using the Box-Behnken response surface methodology. The processing parameters included jet pressure, workpiece rotation speed, abrasive particle size, abrasive particle concentration, and polishing time. The processing results included surface roughness and material removal rate.

[0065] S2: Shuffle the dataset and divide it into training and test sets in an 8:2 ratio;

[0066] S3: Use Bayesian optimization algorithm to optimize the hyperparameters of machine learning models; machine learning ensemble models include three machine learning models: CatBoost, XGBoost, and LightGBM.

[0067] S4: Input the shredded and partitioned datasets into the optimized machine learning ensemble model for training and prediction, and build a regression prediction model for processing parameters and processing results;

[0068] S5: Select the better regression prediction model based on the R2 and RMSE evaluation indicators;

[0069]

[0070]

[0071]

[0072] Where N represents the number of samples, y i Represents the actual value. R² represents the predicted value, and Rmean represents the sample mean. In evaluating machine learning performance, R² measures how well the model fits the data; R² ≤ 1, and the closer R² is to 1, the better the machine learning model's predictive performance. RMSE measures the deviation between the predicted and actual values; the smaller the deviation, the closer the predicted value is to the actual value. MAE measures the absolute error between the predicted and actual values; the smaller the error, the closer the predicted value is to the actual value.

[0073] S6: The better regression prediction model is used as the objective function of the multi-objective optimization algorithm to perform multi-objective optimization and obtain the Pareto solution set of surface roughness and material removal rate;

[0074] The objective function obtained by the multi-objective optimization algorithm is:

[0075] min F(X)=[f Ra (X), -f MRR (X)] T

[0076]

[0077] X = (A, B, C, D, E) T

[0078]

[0079] 30≤A≤90, 5≤B≤15, 0.6≤C≤1.8

[0080] 1000≤D≤3000, 1≤E≤20

[0081] Where fRa (X) represents the surface roughness regression prediction model from a superior machine learning model, f MRR (X) represents the material removal rate regression prediction model from the superior machine learning model. X is a feature vector composed of 5 variables x in the feasible space. The constraint range of the 5 variables was determined through pilot-scale experiments as A∈[30, 90], B∈[5, 15], C∈[0.6, 1.8], D∈[1000, 3000], and E∈[1, 20].

[0082] S7: Select a better multi-objective optimization algorithm by using hypervolume evaluation index;

[0083] Super-volume evaluation formula:

[0084]

[0085] Where P is the final solution set; r∈R M It is a reference point in the target space; It is the solution set that dominates the reference point r; vol(s,r) represents the volume between s and r. The higher the HV value, the better the convergence and versatility of the algorithm.

[0086] S8: Select the optimal solution from the Pareto solution set obtained by the better multi-objective optimization algorithm based on the entropy weight method.

[0087] Example:

[0088] The flowchart of the method for predicting and optimizing the machining parameters of rotating abrasive waterjet polishing of this invention is as follows: Figure 1 As shown, to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described in detail below with reference to the accompanying drawings and embodiments. The specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0089] First, the experimental factors and number of levels were determined. In the rotating abrasive waterjet polishing process of this invention, jet pressure, abrasive particle size, abrasive concentration, workpiece rotation speed, and workpiece rotation time were identified as the main factors affecting the experimental objectives through experiments. To improve the polishing efficiency and quality of rotating abrasive waterjet polishing, surface roughness and material removal rate were set as objectives. Based on the experimental parameters, the range of experimental factors was determined, and an experimental matrix was designed using the Box-Behnken response surface methodology. Experiments were conducted based on the experimental matrix, and the final experimental results table was used as the training and testing dataset for ensemble machine learning.

[0090] Table 1. Ranking of experimental factors and levels

[0091]

[0092] The experimental results were divided in an 8:2 ratio, with the training set comprising 80% and the test set comprising 20%. To avoid excessive numerical differences caused by different units of measurement among the experimental factors, the prediction results were normalized using the Max-Min method. This normalization method, also known as linear normalization, linearly transforms the original data to map it to the range [0,1] or [-1,1]. This method is suitable for data distributed within a finite interval. Its formula is... x is the original data, x min and x max x represents the minimum and maximum values ​​of the original data. scale This is the normalized data. Then, the machine learning ensemble regression prediction model is trained to fit the training set, and then the trained model is used to make predictions on the test set. Based on the prediction evaluation criteria R2, RMSE, and MAE, the model with the highest overall score is selected. The evaluation criteria are then used to determine if the prediction accuracy requirement is met. If it is, the model is moved to the next step; otherwise, the hyperparameters of the model with the highest overall score are optimized using a Bayesian optimization algorithm. Then, the training and prediction are repeated until the optimal machine learning regression prediction model is obtained. The formulas for R2, RMSE, and MAE are as follows:

[0093]

[0094]

[0095]

[0096] Where N represents the number of samples, y i Represents the actual value. R² represents the predicted value, and Rmean represents the sample mean. When evaluating machine learning performance, R² measures how well the model fits the data. R² ≤ 1, and the closer R² is to 1, the better the machine learning model's predictive performance. RMSE measures the deviation between the predicted value and the true value; the smaller the deviation, the closer the predicted value is to the true value.

[0097] The optimal machine learning regression prediction model obtained from the ensemble machine learning prediction part is used as the objective function of the ensemble multi-objective optimization algorithm. Next, the optimization range of the parameters is determined, which is the same as the range of experimental factors set in the experimental preparation part. Then, the parameters of the multi-objective optimization algorithm are set: the initial population `pop_size` is set to 100, the crossover parameter `eta` is set to 20 to control the dispersion of solutions after crossover; the larger `eta` is, the closer the obtained solutions are to their parents; the smaller `eta` is, the more dispersed the obtained solutions are. `prob` is set to 0.95; `prob` is a probability value representing the probability of crossover between two parent solutions. The larger `prob` is, the higher the frequency of crossover; the smaller `prob` is, the lower the frequency of crossover. The maximum number of iterations `n_gen` is set to 200.

[0098] Following this, a multi-objective optimization ensemble algorithm performs multi-objective optimization based on the objective function and parameter settings. After optimization, a new population is generated, and the fitness of this population is calculated. Higher fitness indicates better individuals. The algorithm then checks if a preset convergence threshold has been reached. If so, the hypervolume values ​​of each algorithm in the ensemble are calculated, and the algorithms are ranked based on the hypervolume results to select the best optimization algorithm. If convergence has not been reached, the population undergoes a fast non-dominated sorting to obtain different Pareto front levels. For each level, the crowding distance (the sum of the distances between adjacent individuals in the objective space) is calculated. Crossover and mutation are then performed to generate a new offspring population from the selected individuals. The next generation is selected from the larger population until the preset convergence threshold is met. Finally, the entropy weight method is used to evaluate the optimal solution in the Pareto solution set.

[0099] The present invention has the following innovations compared to the prior art:

[0100] 1) For the prediction of processing parameters and processing targets in the specific working condition of rotating abrasive waterjet polishing, traditional prediction methods often use a single machine learning model, which leads to (1) the prediction accuracy being limited by the complexity of the model and the quality of the training data, and may not be able to fully capture the nonlinear and dynamic characteristics in the processing process. (2) the prediction results may have large uncertainties and fluctuations, making it difficult to guarantee the stability and reliability of the prediction. (3) the prediction method lacks universality and adaptability, and may not be able to adapt to different processing conditions and targets.

[0101] In order to overcome the above shortcomings, the main idea of ​​the method proposed in this invention is: (1) to use a variety of different types of machine learning models, such as CatBoost, XGBoost and LightGBM, to predict the processing parameters and the target respectively, and to make full use of the advantages of each model to complement each other; (2) to adopt a model fusion strategy based on weight allocation, and to dynamically adjust the weight of each model in the fusion according to the prediction performance and credibility of each model, and to combine the prediction results of each model to obtain the final prediction value.

[0102] The effectiveness and advantages of this method are: (1) This method can effectively improve the prediction accuracy of rotating abrasive waterjet polishing parameters and targets; (2) This method can effectively reduce the prediction fluctuation of rotating abrasive waterjet polishing parameters and targets; (3) This method has strong versatility and adaptability, and can adapt to different processing conditions and targets. For different types of workpiece surface morphology, material properties, processing requirements, etc., it can give relatively accurate and reliable prediction results.

[0103] 2) Traditional machine learning hyperparameter optimization methods include: manual tuning, which relies on trial and error and experience to set hyperparameters, is time-consuming and inaccurate. Grid optimization is a method that optimizes model hyperparameters by traversing a given set of parameter combinations. However, grid optimization cannot guarantee finding the global optimum and is prone to selecting local optima. Stochastic optimization is a method that randomly selects parameter combinations in the hyperparameter space for model training. However, stochastic optimization easily wastes computer resources on ineffective or inefficient hyperparameter combinations while ignoring more promising regions, and can only approach the optimal solution with a certain probability.

[0104] This invention uses Bayesian optimization to find the optimal hyperparameter combination, which has the following advantages over traditional methods such as gridded optimization and stochastic optimization: High efficiency: Bayesian optimization can search a larger hyperparameter space in the same amount of time, thus improving search efficiency. Adaptability: Bayesian optimization can adaptively select the next hyperparameter combination for evaluation, thus finding the optimal solution faster. Robustness: Bayesian optimization can maintain good performance even under conditions of insufficient sampling or high noise, thus exhibiting better robustness.

[0105] 3) For multi-objective optimization of processing parameters in abrasive waterjet polishing, a single algorithm is often used, such as NSAGAII or AGEMOEA. These algorithms can achieve good results in some cases. However, the performance of a single algorithm is limited and may not yield the optimal solution in all situations. Furthermore, different algorithms have their own advantages and disadvantages; using only a single algorithm makes it difficult to fully utilize the strengths of different algorithms.

[0106] To better optimize the processing parameters of abrasive waterjet polishing, this invention employs an integrated multi-objective optimization strategy, combining the NSGAII and AGEMOEA algorithms to achieve collaborative optimization among multiple algorithms. This allows for mutual supplementation and optimization between different algorithms, resulting in a more comprehensive and accurate solution. The integrated multi-objective optimization strategy helps engineers better balance the relationships between various objectives, leading to more practical results. Therefore, using an integrated multi-objective optimization algorithm is an efficient and feasible method for optimizing the processing parameters of rotating abrasive waterjet polishing.

[0107] The innovative aspects of this invention:

[0108] 4) Traditional methods and techniques for polishing the inner wall of round tubes include: Manual polishing: Using hand tools such as sandpaper or abrasive cloth, rubbing back and forth along the inner wall of the round tube to remove surface impurities and burrs. Mechanical polishing: Using polishing machinery, placing a grinding head or polishing cloth inside the round tube through high-speed rotation or vibration to achieve a polishing effect. Chemical polishing: Using chemicals applied to the inner wall of the round tube, removing surface impurities and burrs through a chemical reaction. Commonly used chemicals include acids and alkalis. Electrolytic polishing: Using the round tube as the anode or cathode, removing surface impurities and burrs through an electrolytic reaction. This method requires specialized equipment and technology. Manual polishing requires a large amount of manpower and time, resulting in low efficiency; mechanical polishing requires specialized equipment and technology, leading to high costs; chemical polishing may generate harmful waste and pollutants, harming the environment; electrolytic polishing requires specialized equipment and technology, resulting in high costs.

[0109] This invention primarily utilizes a polishing slurry formed by mixing abrasive and water, which is then impacted onto the workpiece surface under high pressure to perform micro-cutting, thereby achieving a polishing effect. Simultaneously, the jet angle of the polishing slurry can be infinitely adjusted by controlling the workpiece's rotation speed, thus adapting to different polishing angle requirements. Addressing the challenge of machining inner pipe walls, a novel machining tool is designed: a nozzle with infinitely adjustable incident angle. The tool's end has three square grooves, from which the polishing slurry is sprayed in a beam from three directions. The workpiece base drives the workpiece to rotate, and the incident angle of the jet is adjusted by controlling the workpiece's rotation speed. This infinitely adjustable jet design allows for adjustment of the incident angle as needed, enabling polishing at different angles and resulting in a more uniform polishing effect. The polishing medium used in this invention is a solid-liquid two-phase flow of water and silicon carbide, which is environmentally friendly and more economical.

Claims

1. An integrated multi-objective prediction and optimization method for machining parameters of rotating abrasive waterjet polishing, characterized in that, Includes the following steps: S1: The processing parameters and results of the polishing process of the rotating abrasive waterjet polishing device were obtained using the Box-Behnken response surface methodology, and the experimental dataset was obtained. The rotating abrasive waterjet polishing device in S1 includes a metering pump, a damper, a back pressure valve, a pressure gauge, a pressure regulating valve, a polishing moving system, a polishing tool head, a polishing rotation system, a mixing system, and a control system. The damper is connected to the metering pump, and the back pressure valve is connected to the damper; the damper and back pressure valve are used to stabilize the outlet pressure of the metering pump. The pressure gauge is connected to the back pressure valve and is used to observe and record the output pressure of the polishing system. The pressure regulating valve is connected to the pressure gauge and is used to regulate the output pressure of the polishing system. The pressure regulating valve is connected to the polishing moving system via a hose and is used to drive the polishing tool head to move along the z-axis. The polishing rotation system is used to realize the rotational movement of the workpiece. The mixing system is used to fully mix the solid abrasive particles and the liquid aqueous solution in the polishing fluid to achieve a uniform distribution of abrasive particles in the polishing fluid. The control system is used to control the power execution and control of the polishing moving system, the polishing rotation system, and the mixing system. The processing parameters of the experimental dataset in S1 include jet pressure, workpiece rotation speed, abrasive particle size, abrasive concentration, and polishing time; the processing results of the experimental dataset include surface roughness and material removal rate. S2: Shuffle the experimental dataset and divide it into training and test sets according to the ratio; S3: Use the Bayesian optimization algorithm to optimize the hyperparameters of the machine learning ensemble model; S4: Input the shredded and partitioned experimental datasets into the optimized machine learning ensemble model for training and prediction, and construct a regression prediction model for processing parameters and processing results; S5: Select the optimal regression prediction model based on the evaluation indicators of R2, RMSE and MAE; S6: Multi-objective optimization is performed by using the optimal regression prediction model as the objective function of the multi-objective optimization algorithm to obtain the Pareto solution set of surface roughness and material removal rate; The objective function of the multi-objective optimization algorithm in S6 is: min F(X)=[f Ra (X),-f MRR (X)] T X=(A,B,C,D,E) T 30≤A≤90, 5≤B≤15, 0.6≤C≤1.8 1000≤D≤3000, 1≤E≤20 Where f Ra (X) represents the surface roughness regression prediction model from the optimal machine learning model, f MRR (X) represents the material removal rate regression prediction model from the optimal machine learning model. X is a feature vector composed of 5 variables x in the feasible space. The constraint range of the 5 variables was determined through pilot-scale experiments as A∈[30, 90], B∈[5, 15], C∈[0.6, 1.8], D∈[1000, 3000], E∈[1, 20]; S7: Select the optimal multi-objective optimization algorithm based on the hypervolume evaluation index; S8: Select the optimal solution from the Pareto solution set obtained by the optimal multi-objective optimization algorithm according to the entropy weight method.

2. The integrated multi-objective prediction and optimization method for machining parameters of rotating abrasive waterjet polishing according to claim 1, characterized in that, The machine learning ensemble model in S3 includes three machine learning models: CatBoost, XGBoost, and LightGBM.

3. The integrated multi-objective prediction and optimization method for machining parameters of rotating abrasive waterjet polishing according to claim 1, characterized in that, The R2 and RMSE evaluation metrics in S5 are as follows: Where N represents the number of samples, y i Represents the true value. R² represents the predicted value, and R² represents the sample mean. When evaluating machine learning performance, R² measures how well the model fits the data. R² ≤ 1. The closer R² is to 1, the better the machine learning model's predictive performance. RMSE measures the deviation between the predicted value and the true value. The smaller the deviation, the closer the predicted value is to the true value. MAE measures the absolute error between the predicted value and the true value. The smaller the error, the closer the predicted value is to the true value.

4. The integrated multi-objective prediction and optimization method for machining parameters of rotating abrasive waterjet polishing according to claim 1, characterized in that, The formula for evaluating the super-volume in S7 is as follows: Where P is the final solution set; r∈R M It is a reference point in the target space; It is the solution set that dominates the reference point r; vol(s,r) represents the volume between s and r. The higher the HV value, the better the convergence and diversity of the algorithm.

5. The integrated multi-objective prediction and optimization method for machining parameters of rotating abrasive waterjet polishing according to claim 1, characterized in that, The polishing moving system includes a servo motor, a linear slide, and a polishing tool head. The polishing tool head is fixed on the slider of the linear slide, and the servo motor is used to drive the linear slide to move the polishing tool head linearly along the z-axis in both positive and negative directions. The polishing rotation system includes a polishing rotation system servo motor and a workpiece fixing seat. The polishing rotation system servo motor is connected to the workpiece fixing seat, and the workpiece is fixed on the workpiece fixing seat. The polishing rotation system servo motor drives the workpiece fixing seat to rotate the workpiece. The mixing system includes a pneumatic motor, an air pump, a propeller, and a polishing slurry supply tank. The air pump provides power to the pneumatic motor, which is connected to the propeller to drive the propeller to rotate, thereby stirring the polishing slurry in the polishing slurry supply tank to ensure that the abrasive particles in the polishing slurry are evenly distributed and to prevent abrasive particle deposition.

6. The integrated multi-objective prediction and optimization method for machining parameters of rotating abrasive waterjet polishing according to claim 5, characterized in that, The polishing tool head includes a polishing fluid linear jet orifice and a coaxial positioning head. The polishing tool head is driven by the polishing moving system and presses against the center positioning groove of the workpiece fixing seat in the polishing rotating system to ensure that the workpiece and the polishing tool head remain coaxial when rotating.

7. The integrated multi-objective prediction and optimization method for machining parameters of rotating abrasive waterjet polishing according to claim 5, characterized in that, The polishing process of the rotary abrasive waterjet polishing device is as follows: 1) First, fix the workpiece on the workpiece holder. Then, control the servo motor through the control system to drive the linear slide to move the polishing tool head down along the z-axis and press against the positioning groove of the workpiece holder, so that the polishing tool head moves into place. 2) Prepare the polishing liquid in the polishing liquid supply tank according to the required ratio. Use the pneumatic motor of the control system to drive the stirring propeller to start mixing and stirring the polishing liquid to ensure the uniformity of the polishing liquid. 3) The metering pump is started by the control system to drive the polishing fluid from the polishing fluid supply tank through the damper and back pressure valve for pressure stabilization, and finally enter the polishing tool head. It is sprayed from the linear jet nozzle of the polishing tool head onto the inner wall of the workpiece. The polishing fluid then flows back to the polishing fluid supply tank through the return column to enter the next cycle. 4) The control system sets the speed of the servo motor in the polishing rotation system, and starts the servo motor in the polishing rotation system to drive the workpiece holder to rotate the workpiece; 5) The angle between the polishing liquid spray path and the inner wall of the workpiece can be controlled by controlling the rotation speed of the workpiece, thereby achieving stepless adjustment of the jet angle.

Images