The Research Of Fuzzy Neural Network On Solving Error Reflection Phenomenon In Machining

In machining process, the change of the work system distortion caused by the amount of machining remaining's changing or workpiece's odds produces the machining error is called the error reflection. Based on the error reflection theories, the error reflection coefficient reflects the influence degree of the semi-finished product to the work after machining and it is decided by work system hardness, cutting condition, machining quantity and workpiece's rigidity, etc., which presents complicated non-line relation and is hard to calculate by formula precisely. The usual method to solve the problem is machining for many times with the experience of persons and choosing appropriate amount of remaining every time to reduce the error reflection. But this method has strong subjectivity and can't remove the influence carried by error reflection nicety and efficiently, which brings difficulty to computer assistant machining and hasn't been solved yet. The fuzzy neural network is the outcome that the neural network technique combines the fuzzy logic control techniques, which is a fuzzy control method based on the neural network. This text puts forward the fuzzy neural network to resolve the error reflection phenomenon in machining. Although the fuzzy logic and the neural network have the obvious dissimilarity on the concept and contents, they are all used for solving the difficulty on system control caused by the uncertainty and imprecision. The fuzzy logic imitates the logic thinking of person's brain and used to deal with the imprecision in control. The neural network imitates the function of the brain nerves and can approach any non-line function and reflect the relation of input and output. The combining of two to solve the error reflection problem has certain possibility. This subject attempts to solve the problem of error reflection by fuzzy neural network and discusses the possibility of it firstly, then designs the fuzzy neural network model combining the basic principle and the theories characteristics of the fuzzy control and the neural network control. The method has clear theories and perspicuity structure and its exactness and feasibility has been validated through emluation on the computer to train and test the established network model. The main content of the text includes: 1. The analysis and research of the error reflection phenomenon. Combining the error reflection theories and the machining circumstance analysis the factors influence the error reflection coefficient, then educes the complicated non-line relation of it and the semi-finished product error(Efront), the craft system hardness(K),machining quantity(f),the workpiece rigidity(HBS) and machining times(Z).According to experience, along with the increment of machining times, the error can be minished, but in physically machining, the number of times generally isn't more than 3 times. So we can suppose the machining times is 3,the machining quantity each time is f1, f2, f3.Because f= f1+ f2+ f3, and f has already been known, if we know f1 and f2,we can control the whole machining process. So put the craft system hardness(K),the semi-finished product error(Efront), the error after machining(Eafter), process behind path to the error margin E is behind, enter to be the importation item of the misty nerve network for the quantity f, machining quantity(f)as input, the machining quantity each time(f1, f2)as output to form a four input two output network. Train and amend the network to attain the ideal output; 2.The establishment of network model. Resolving the error reflection phenomenon is a process that approaches a complicated non-line function, but different FNN has its merits and shortcomings. This text compares different type of FNN to choose the I FNN which is applicable for this subject and oppositely easy to realize. Then establish the FNN structure model compared with the error reflection model; 3.Collect the experiment data. Use a easy eccentric clamp to machine on a lather for many times with different hardness material( including iron, aluminum,20# steel,45# steel), record the results( including K,Efront,Eafter,f,f1,f2) as the data for network training and testing; 4.Analyze and study the data. Classify the data based on the machining theory and set up the fuzzy control rules database; 5.The structure optimize of the network model. Because the study arithmetic of I FNN is based on the BP network 's arithmetic and it exists such shortcomings as large expense, slow constringency speed, easy getting into local tittle, etc., usually faces the problem that can't be trained completely, it needs to be bettered. The main resolvents include choosing the appropriate original authority and using less study velocity. Because initial authorities can't be estimated...