When error compensation is performed, the error of the measured point is often superposed on each discrete target point after the error value is inverted; equal to the noise is superimposed instead of homogenized, thus affecting the error compensation accuracy. For this reason, this paper aims to apply the planar orthogonal grating to the error measurement of the three-axis vertical machining center, and establish the spatial error compensation model of the planar orthogonal grating measurement system by neural network technology.
Practice has proved that this method can effectively filter the sample data and weaken the noise interference. The expected data accuracy is within the tolerance range. The neural network error compensation model is shown. Neural network error compensation model In the model, there are 2 input nodes corresponding to the input vector (x, y); one output node is corresponding to the spatial position error $x or $y; the intermediate hidden layer nodes are 11 . The nodes between the three layers are fully connected, that is, each neuron node in the left layer is connected to each neuron node in the right layer. There is no connection between the neuron nodes in the same layer. Wi,j is the weight vector of the input layer to the hidden layer, and Wj,k is the weight vector of the hidden layer to the output layer. The activation function of the output layer node is a linear function and is added as the expected output vector to the output of the neural network error compensation model. The vector added at the input is the target vector corresponding to the error vector. Network training is then performed until the accuracy requirements are met or the number of training sessions is met. The trained neural network model is used to simulate the error of the measured point and the unmeasured point. The error value vector obtained by the simulation is the value to be compensated. Then the error value is inverted and the NC machining program is reconstructed to generate a new NC command code, so that the CNC controller can make corresponding actions to achieve the purpose of error compensation.
For ease of measurement and verification, the predicted point error of the design is measured together (not shown). Designing measurement trajectories In order to ensure the validity and verifiability of the test data, the degree of influence of the machine speed on the spatial position error is also studied. At the same time, in order to eliminate the influence of random errors on the measurement results, for each target point and its vicinity The sample data is average filtered within a certain error range, and the result is used as the error value of the target point. It can be seen from the sum that the position error along the X direction on the curved surface is generally between +110 and 215 Lm, and the negative value of the positive error along the X axis increases. The speed of the machine tool has little effect on the measurement results.
Conclusion (1) Under the no-load condition, whether the actual measurement error value or the neural network error compensation model is used to compensate the machine space position error, the positional accuracy along the X and Y directions is greatly improved after one compensation. (2) Using the neural network technology to establish the error compensation model is completely feasible for the position error compensation of the machine tool. (3) The experimental study in this paper fails to give a better explanation for the origin of the error. (4) The detection and compensation of thermal error in vertical machining centers has yet to be further studied.
(Finish)
Load Skates,Heavy Duty Caterroller Load Skates,Enerpac Tools
HENGTAI INTERNATIONAL CO., LTD. , http://www.hengtai-power.com