Data Storage,Sample Extraction And Sample Update Technology For Self-learning Of Response Law Of CNC Machining Process System

The self-learning of the cutting process response law aims to extract and abstract the model of the response signal(such as cutting force,spindle power,etc.)about the working conditions from the data generated by the daily processing of CNC machine tools.The self-learning of response law is of great significance to realize the intelligentization of CNC machine tools.However,the data generated by actual cutting cannot be directly used for the self-learning of the response law.It is necessary to extract the sample data from it,and realize the efficient and reliable storage and update of the sample data.In response to the above problems,this paper proposes a self-learning process of the response law of the CNC machining process system,and proposes data storage,sample extraction and sample update programs for this process.In order to realize the self-learning of the power response law of the rough milling spindle,this paper divides the self-learning process into five stages: data acquisition,sample extraction,sample management,response law learning and response law management.In order to achieve efficient access to sample data and response laws in this process,this paper formulates different storage schemes for structured sample data and semi-structured response laws,and implements three of process system,sample data and response laws Connection.The self-learning of the cutting process response law requires sample data.In order to extract the data used for the self-learning of the response law from the data generated by the daily processing of CNC machine tools,the effective samples are defined in combination with the working conditions and redundancy factors,and the design is effective.Sample pre-positioning method.In order to achieve the matching between the working data of the effective samples and the response data,an algorithm is designed to extract the spindle power of the effective samples.Aiming at the possible abnormal data segment of the effective sample spindle power,a detection algorithm based on dynamic time bending distance and hierarchical clustering algorithm is designed.Taking the experimental data of rough milling as an example,the effective sample extraction process was completed in the order of effective sample pre-positioning,effective sample spindle power extraction,and effective sample spindle power abnormal band detection,and high-quality sample data was obtained.The response characteristics of the process system will change with the extension of service time,which determines that the response law is time-sensitive.In order to avoid the negative impact of the old training samples on the self-learning effect of the response law,the BP neural network model of the spindle power is used as an example to derive the training samples to test The influence of this strategy is used to design a training sample selection strategy based on test errors,and the effectiveness of the strategy is verified through experiments.According to the design of the self-learning process for the power response of rough milling spindle,a self-learning software system was developed.Design the framework of some modules of the software,complete the development of some modules based on C ++ and C #,and realize the functions of data visualization and effective sample extraction.