Research On The Key Intelligent Technologies For Machine Tool Using Instruction-domain Based Approach

After more than sixty years’ development,the CNC machine tool born in 1952 has been a sufficient improvement in auto-execution which can mostly solve the problem of replacing the physical labor,while the CNC machine tool still exists the problems of incomplete data-sensing and inadequate decision-making capacity making it hard to replace the mental labor well.In order to improve the intellectualized level of the CNC machine tool,this dissertation studied the implementation methods and technologies of key technologies such as autonomously sensing,intelligent decision-making and automatically monitoring based on the massive real-time data aquired by CNC system in the machining process.Regarding to the recording massive time-seriese data in the machining process,it is hard to automatically establish the mapping relationship between the data features,obtained by the regular time-domain or frequency-domain analysis methods,and the working tasks causing the human beings’ deep engagement in the machine tool’s learning and decision-making behaviors.Thus,the instruction-domain based data acquisition and analysis methods are proposed.When the CNC system periodically records the position,speed,current,and power,the instruction line number of the NC code in the same cycle is synchronously recorded,thereby establishling the precise mapping relationship between the state data and the working task by using the instruction line number as an index,and automatically labelling the state data with the system response features.Then,as the instruction line number is the basic unit of feature statistic and analysis,and corresponds with the minimum scope of the process parameters,the state data marked with the instruction line number is used to design the instruction-domain analysis method which differs from the regular time-domain and frequence-domian analysis,and lays foundations of data analysis for intelligent applications.It is of great significance to preprocess the data acquired by the CNC system.On the one hand,fully utilizing the precise mapping relationship between the state data and the working tasks,the specific working task related data in a single set of time series data is automatically separated to improve the data analysis pertinence and the same working task related data in two sets of time series data is compared and analyzed to improve the sensitivity of the weak features.On the other hand,the low-frequency nonlinear interference components,caused by the thermal effect,contained in the spindle power and the feed axis current are eliminated without any temperature sensors by the empirical mode decomposition method,and thus the data validity is improved.Improving the machining efficiency by optimizing the machining process parameters is one of the typical intelligent functions that the CNC machine tools should have.A feedrate optimization method based on the instruction-domian analysis is proposed.Firstly,the spindle power is accurately segmented,and the thermal components are eliminated,and finally the cutting force related components are left as the research object.The average value of the spindle power corresponding to each instruction line is calculated and the orthogonal experiments are used to establish the index model for the spindle power and feedrate.Giving the index model,a multi-objective optimization model is proposed to improve the machining efficiency and reduce the spindle power fluctuation,which is heuristicly learned and optimized by the Controlled NSGA-II algorithm.And finally,the feedrate of each instruction line is adjusted based on the measured data,and the adaptability of the feedrate to the processing conditions is improved,thereby optimizing the efficiency.Aiming at the demand of the rough machining efficiency improvement in the 3C field of aluminum alloy mobile phone,the process parameters optimization based on the instruction-domain analysis method was designed,and the feedrate optimization software was developed.After optimization,the machining efficiency can be improved by 5%-20%.Aiming at the demand of automatically tool breakage monitoring for unmanned automation workshops,a monitoring method for end mills based on the instructioin-domain analysis is proposed.Firstly,the spindle power is accurately segemented and comparatively analyzed,and the thermal components are eliminated,and finally the sensitive components related to the cutting force are left as the research object.Aiming at the problems of automatic acquisition of samples,high initial acquisition cost of tool breakage samples,incremental learning from the samples,imbalanced distribution of dataset and high real-time requirement,the ICSSVM based tool breakage monitoring method in end milling is proposed,which has achieved incremental learning of the new samples while improving the efficiency of the calculation,and improved the classification accuracy of the imbalanced samples,and has good industrial application value.Compared with the methods of monitoring by the external sensors or the additional tool detecting devices,the method requires neither an external sensor nor an additional operation for moving the tool to the detecting position in the machining program,resulting in the low cost and high efficiency.Aiming at the demand of end mill monitoring in the intelligent workshops of aluminum alloy mobile phone case processing in 3C field,the tool breakage monitoring software running under the cloud platform was developed and deployed at the industrial site.The instruction-domain analysis method has produced some good results in the applications of process parameters optimization and tool condition monitoring,which has laid a good foundation for its future intelligent applications.