Adaptive EDM Simulator Test

Abstract : The EDM machining model has been established using neural network technology. Experiments show that the neural network model fully reflects the machining characteristics of the machine tool. Based on this model, the relationship between the main processing parameters such as peak current, pulse width, and pulse interval, and the processing speed and surface roughness was simulated. The adaptive EDM machine processing process that is in line with the actual production situation was obtained. Laws and use simulation results to reveal the difference between adaptive control machine tools and traditional machine tools.
Keywords: EDM; simulation test; neural network; adaptive control

Since the adaptive control EDM has adaptive control capability, it can adjust the electrical processing parameters in real time and improve the gap state. Therefore, under the condition that the processing parameters are not very reasonable, the discharge can be stably performed without The occurrence of workpiece burns and other phenomena, compared with the traditional electric machine tools, has a great advantage. However, at present, the study of its process rules is still relatively lacking, which leads to blind selection of process parameters in production practice, and it is difficult to obtain ideal productivity. The study of the process control law of EDM under the action of adaptive control system will not only have important practical significance for the selection of process parameters, but also have great reference significance for the design and improvement of the control system. However, because the mechanism of EDM is very complicated and it is a highly nonlinear system, it is very difficult to establish a strict mathematical model. It is not only time-consuming and costly to test completely on EDM machines. It is difficult to systematically study the process rules. In addition, some process rules may not necessarily be studied on the machine tool, such as deep holes or arc discharges that are easily destructive. If an EDM model can be established, not only can the computer be used to systematically study the rules of the process on the basis of the model, but also the cost is low and it is safe and reliable. It can also be used to simulate certain processes that are not easily tested on the machine tool. . Based on a large number of process tests, this paper establishes the EDM neural network model and uses this model to simulate the relationship between the main processing parameters such as peak current, pulse width, and pulse interval, and the processing speed and surface roughness.

1 EDM experimental design

1.1 Determination of input and output variables

In EDM machining, peak current Ip, pulse width ton, pulse interval toff, lifting time tup, and processing time tdn, no-load voltage, servo reference voltage, processing depth, processing area, electrode shape, electrode material, and dielectric Liquid factors are closely related to the speed, surface roughness, and electrode loss of EDM. However, treating all these parameters as independent variables will make the number of experiments too numerous and unrealistic, and the test data of all these parameters cannot be obtained at the same time. Based on the above analysis and discussion with experts in electric machining and actual operators, the input variables are determined as the peak current, pulse width, pulse interval, tool-lifting time and processing time; and the processing speed and surface roughness reflect the main processing of the machine tool. characteristic. EDM processing experimental conditions are as follows:

Machine: Sodick A3C
Working fluid: Domestic electric discharge processing oil Electrode material: Copper Workpiece material: Iron-based alloy Processing polarity: Electrode (+)

1.2 Experimental Methods

In the machine used in the experiment, there are 30 peak currents and 63 pulse widths. The available pulse intervals are 63×9=567. It is extremely difficult to select a reasonable combination of parameters by experience. Brings trouble to the experimental design and it is impossible to conduct full-factor experiments. Therefore, it is necessary to determine an experimental design method that can more fully reflect the machining characteristics of the machine without too many experiments.

The mixed orthogonal design method is currently the most used and most effective experimental design method. As long as a small number of experiments can fully reflect the relationship between each independent variable and the corresponding variable. In the experiment, the peak current, pulse width, and pulse interval were all designed in 9 levels. Since the tool change time and the processing time were relatively small, a 3-level design was used. Therefore, the total number of experiments was 9×9=81 times. Relative to the full factor test 9 × 9 × 9 × 3 × 3 = 6561 times, the number of trials greatly reduced.

After the machining is stable, the Z-axis position of the machine tool is read at regular intervals, and then converted into the workpiece erosion volume during the period of time, the processing speed can be obtained. In order to ensure the reliability of the measured data, each data is generally measured five times, averaged to get the processing speed under the corresponding process conditions. The surface roughness was measured using a Surfcom Model 130A special surface roughness tester manufactured by Japan SEIMITSU Co., Ltd. This instrument can directly output various standard surface roughness values ​​with an accuracy of 0.1 μm.

2 Neural Network Modeling

Artificial neural network has the characteristics of self-organization, self-adaptation, self-learning and non-linear dynamic processing. It can realize the generalization, analogy, and generalization of the human brain to a certain degree, and it does not need to pre-specify the model and can automatically start from a large number of The law of data extraction, through the associative memory and promotion capabilities to obtain the required data, is suitable for solving complex nonlinear problems. EDM is precisely such a system that has a high degree of non-linearity and it is difficult to describe the process rules with specific mathematical expressions. Applying neural networks in the EDM process modeling can precisely exert the superiority of the neural network.

The neural network system is a system with complex structure and perfect performance composed of a large number of simple neurons. First, according to the actual situation to determine the topology of the network, and then use a certain number of samples to train the network to establish the mapping relationship between input and output.

The EDM neural network model can be represented by Figure 1, which uses the most commonly used BP network. The determination of the number of hidden layer units in the neural network model has not yet been clearly defined. It is generally determined through network structure experiments. Using the experimental data as a sample, the network structure was changed from 5-5-2 to 5-28-2, and the 5-14-2 structure was found to be the most suitable. This structure has high learning accuracy and learning speed is not slow. The most reasonable.


Fig. 1 EDM process model

After the establishment of the neural network model, in order to verify the correctness of the network, the experiment shown in the following table was verified. From the results in the table, it can be seen that the predicted value of the neural network is quite close to the experimental value, indicating that the model has been able to reflect the process characteristics of the machine tool, embodies the process law of the machine tool, and can be used to simulate the EDM process test.

Neural network model prediction results


3 Process Simulation Simulation

Through the previously established neural network model, the relationship between the main processing parameters of EDM and processing performance was simulated. The simulation curves obtained are shown in Fig. 2 to Fig. 7, the basic conditions of the simulation test and 1.1 regulations. The same.


Figure 2 Pulse Width and Surface Roughness Figure 3 Pulse Width and Processing Speed

Figure 3 shows that when the peak current is held constant, the processing speed increases with the increase of the pulse width; but when the pulse width increases to a certain extent, the processing speed no longer increases with the increase of the pulse width, and even there is The decline. According to the analysis, for a certain peak current, when the pulse width is increased, the radius and depth of the discharge mark are increased due to the increase of the discharge energy, but when the discharge duration is too long, the discharge mark radius and discharge mark depth are Increased, but due to the thermal conduction of the workpiece itself, the discharge energy is not effectively used, and the processing speed will still decrease due to the increase of the pulse width, while the increase in the depth of the discharge trace will increase the surface roughness, which can be increased from Figure 2 is verified. Therefore, there is an optimal pulse width for a certain peak current, which maximizes the processing speed. Excessive pulse width is harmful. Simulation studies also show that when the pulse width is too small, the discharge energy is very unstable due to the small discharge energy, resulting in very low processing speed.


Figure 4 Peak Current and Surface Roughness Figure 5 Peak Current and Processing Speed

When the pulse width remains constant, the discharge energy will increase with the increase of the peak current, and the erosion of the unit pulse will increase. Therefore, the processing speed and the surface roughness will increase with the increase of the current. When the peak current increases to a certain extent, the gap state gradually deteriorates due to the excessive discharge energy. At this time, the conventional machine tool cannot adjust the machining parameters adaptively, and the machining state gradually becomes unstable, and if the peak current is further increased, the machining Will not continue, but the machine with adaptive control system will automatically increase the pulse gap, so that discharge products can be better discharged, the gap state is improved, but the increase rate of processing speed will be greatly reduced, and Gradually flattened. When the peak current continues to increase, due to severe deterioration of the gap state, processing is extremely unstable, and the adaptive control system will further increase the pulse interval width until the gap state is improved. This obviously reduces the processing efficiency and reduces the processing speed. The increase in pulse energy will lead to an increase in the depth of the discharge trace, which in turn leads to a worsening of the surface roughness. The simulation results in Fig. 4 and Fig. 5 reflect this process rule of EDM.


Fig. 6 Pulse interval and surface roughness Fig. 7 Pulse interval and processing speed

When the peak current and pulse width are kept constant and the pulse interval is changed, the simulation curve shown in FIG. 7 is obtained. In actual processing, when the pulse interval is small, the processing is unstable, and the processing speed is generally very small and difficult to measure. Therefore, it is difficult to use experiments to study the relationship between pulse interval and processing performance in the case of small pulse interval, but the simulation test will not be subject to various limitations during the actual processing, which is conducive to the study of certain extreme phenomena. From the simulation curve, it can be seen that the processing speed does not increase infinitely with the reduction of the pulse interval, but rather it decreases, indicating that there is a critical point in the pulse interval. When the pulse interval is too small, the gap state is deteriorated due to the high discharge frequency, the processing is difficult to stabilize, and even the workpiece burns. Self-adaptive control machine has adaptive performance, can automatically adjust the pulse parameters to avoid the workpiece burn, but because the adaptive control system always has a delay delay process, it may not be completely in place, resulting in a reduction in the processing speed, and the more the deviation The critical point, the faster the decline. The position of the critical point to some extent reflects the control performance of the machine.

It is generally believed that the surface roughness mainly depends on the discharge energy, which is not much related to the pulse interval. However, the simulation curve obtained at the same time as FIG. 7 shows that when the pulse interval is increased to a certain degree, the surface roughness is basically the same. The pulse interval is irrelevant; however, when the pulse interval is relatively small, the surface roughness decreases with the decrease of the pulse interval, and this phenomenon is relatively rare in actual processing. The reason for the analysis is that when the interval between the pulses is too small, discharge concentration tends to occur, and the discharge is concentrated around a certain discharge point for a relatively long time, making the discharge channel thicker, resulting in a larger discharge radius and shallower depth, and the surface roughness. It is reduced accordingly.

4 Conclusion

(1) The EDM technology model was established by using neural network technology. The prediction error of the model was basically controlled within 10%, which reflected the machining process law of the machine tool and provided the possibility of optimizing the process parameters.
(2) Using neural network model to simulate and study the relationship between machining parameters and machining performance of EDM machining on the computer, revealing the process law of the machine tool with adaptive control system.
(3) The simulation results show that the machine tool with adaptive control system can not only cause the workpiece to be burned, but also the traditional method. Machine tools do not have the advantages.
(4) Through computer simulation, some difficult process experiments in actual processing were studied, some of these phenomena were explained theoretically, and the superiority of using simulation technology to study the rules of EDM was fully reflected. That is, the simulation test not only saves time and cost, but also is safe and reliable, and it is also easy to reveal the characteristics of some process rules that are not easily tested on the machine tool.

         At present, the heat metering market commonly used heat table is divided into mechanical heat meter(Mechanical Heat Metering) and ultrasonic heat meter.The mechanical heat meters is consist of temperature sensor, flow sensor and calculator. With the aid of the temperature sensor, The Mechanical Heat Meters measure the inlet water and return water temperature. Flow rate sensor measures the volume rate of the hot water in the tube trough the time difference. The figures recorded will be delivered to integral calculator which calculates the total heat and displays.


       The mechanical Heat Meters sensors measures the flow rate through the circles of the wheels. According to the specification, The DN can be â‰¤40mm and ≥50mm. According to the inner structure, the small size mechanical heat meters can be single jet and multi jet. At present, our mechanical heat meters belong to single jet mechanical heat meters and they have higher request from the water quality. According to the counter of the flow sensor, the mechanical heat meters can be dry and wet type. Our mechanical heat meters are the dry type.Except mechanical heat meters,the Plastic Heat Meter is also a good product.

Mechanical Heat Meters

Mechanical Heat Meters,Mechanical Heat Energy Meter,Mechanical Digital Meters,Mechanical Heat Metering,Plastic Heat Meter

Berggruen Technology Co.,Ltd ,