Research On Multi-Objective Optimization Method Of Process Parameters In Dynamic CNC Machining Scene

CNC machine tool is the "mother machine" of the manufacturing industry and represents the core manufacturing capabilities of a country.With the increasing complexity of manufacturing requirements,more and more CNC machining tasks need to balance the relationship among multiple goals by adjusting process parameters.Therefore,researchers applied multi-objective optimization algorithms into CNC machining to explore the best processing solution.However,it is difficult for most of the existing evaluation functions based on empirical formulas to establish the mapping relationship between process parameters and optimization objectives accurately and quickly.Besides,most solutions cannot automatically make adjustments due to the lack of consideration of changes in constraints caused by the dynamic processing conditions.All those limit the implementation of multi-objective optimization for CNC machining.With the advent of the era of big data,the processing data of CNC has continuously accumulated and improved,and its implicit information has been mined and utilized.Combining historical processing data with an Artificial Neural Network can establish a complex nonlinear mapping relationship between process parameters and multi-objectives.In addition,Context Awareness can provide human with dynamic changing of constraints and corresponding adjustment strategies through the collection,modeling,storage,and mining of task-related data.Therefore,based on the multi-objective algorithms,this paper studied a multi-objective optimization method in the dynamic CNC machining scene with the help of ANN and context awareness to achieve accurate and fast optimization of CNC process parameters.Firstly,the framework of multi-objective optimization for CNC process parameters is designed according to requirements of multi-objective tasks in CNC machining.This framework achieves multi-objective optimization through the collaboration of the context awareness module,the parameters-objectives mapping module,and the multi-objective optimization and decision module.Then,the CNC machining context-aware scheme is designed.Context-aware model is used to analyze machining tasks,express structured data,guide the operation of related algorithms and generate/adjust processing solutions based on dynamic constraints.Data Organization and Data Management schemes are designed based on the guidance of context-aware model.The collaboration of the three parts achieves accurate description and dynamic mapping of the changing constrains of process parameters and provides stable data support for ANN model training.Furthermore,the mapping scheme of CNC machining between process parameters and multi-objectives is designed.According to the characteristics of CNC machining data,BP neural network is selected for mapping realization,and partial gradient descent and adaptive learning rate methods are designed for the shortcomings of BP neural network to overcome the problems of modeling difficulties and poor prediction accuracy of evaluation function;Based on above studies,the multi-objective optimization and decision algorithms for CNC process parameters are designed.The combination of differential evolution algorithm and TOPSIS algorithm realizes the non-dominated solution set acquisition and decision-making.To solve the problem of fixed parameters in differential evolution algorithm,dynamic mutation parameter and adaptive crossover parameter are designed to realize the rapid generation of global non-dominated solution sets.Then,the optimal solution can be obtained with the help of the dimensionality reduction method and the screening method of similar solutions;Finally,the feasibility of the method proposed in this paper is verified through the experiment based on the public data set and the collected CNC machining data.