Due to the randomness and complexity of the molybdenum wire cut EDM process, it is difficult to establish a precise mathematical model between process parameters and process effects. It is very difficult to study the machining process with traditional methods. The research on the Thousand Processes is not deep enough to improve the quality of the processed products and the processing efficiency is low, which seriously impedes the development of wire cutting processing. Therefore, it is imperative to conduct research on the process specification. People have made some efforts in this area, but no breakthroughs have been made.
- Scott et al., University of Nebras, Ka-Lincoln, USA, based on 729 experimental data, establishes between process parameters (discharge current, pulse width and pulse frequency) and process effects (cutting speed and surface roughness with molybdenum cutting wire) using a factorial design method. The relationship, and according to the concept of non-and advantages, choose the best 32 groups of processing parameters. This method requires too many process tests, and the corresponding processing parameters cannot be optimized according to the specific process requirements. Taiwan’s Y. S. Liao et al. performed a variable analysis of the wire cutting process and established a mathematical model of the different processing parameters in the flying process using regression and correlation analysis methods. The Beijing No. 4 Machine Tool Plant in China has also studied this, adopting an orthogonal regression design, and through a full-factor test, established a cutting efficiency model and a surface roughness model, respectively. The above three methods are similar in nature. These methods do not take into account the randomness and complexity of the WEDM process USING molybdenum wire, especially the digitized characteristics of the electrical parameters and control parameters of the CNC electric discharge machine, which are described by an accurate mathematical model. WEDM processing must inevitably make the solution to the results cannot meet the actual processing requirements. Ale kseyer and Korenblum analyzed the mechanism of line cutting in depth and combined with the methods of statistical mechanics, proposed a calculation method for discharge pulse parameters and fluent volume.
Their research has a certain guiding effect on the research of the process, but the proposed calculation method is very different from the actual processing and cannot be used as the basis for the selection of the process parameters in actual production and processing. The Nanyang University of Singapore used an artificial neural network to establish an empirical model for multiple WEDM processes and predicted cutting speed, surface roughness and surface waviness based on Ton, Toff, line tension, and line speed. Their preliminary study provided a new way of thinking and method. Based on a large number of process experiments, teachers and students at Shanghai Jiaotong University learn from previous research results to establish an artificial neural network model for high-speed wire-cut WEDM and use this model to develop high-speed molybdenum wire-cut WEDM (using molybdenum cutting wire or other wire) process effect simulation system. The simulation system can not only predict the process effect and optimize the process parameters, but also can perform some difficult-to-implement machining test simulation tests as required to obtain the desired results。
Want to buy molybdenum wire? And need molybdenum wire price? Please visit www.edm-taguti.com, and http://www.filter-supplier.com/, China’s biggest molybdenum wire china manufacturer and exports self developed and patented “cold-drawing” molybdenum cutting wire. Related knowledge: molybdenum wire uses, molybdenum wire edm, molybdenum wire glass removal tool, molybdenum wire manufacturers in china, molybdenum wire UK
Neural Network Model for WEMD using molybdenum wire or brass wire
Using the neural network’s ability to map input and output, a BP model of the WEDM process can be established to simulate the WEDM process and provide a reliable basis for the research and development of the computer simulation and intelligent system of WEDM.
- Artificial Neural Network (ANN)
Artificial Neural Network (ANN) is based on the basic research results of neuroscience and physiology. It abstracts, simplifies, and reproduces some aspects of the human brain in information storage search and processing from different perspectives to a certain degree. Principles; In-depth study of the possibility of these principles applied to various fields and ways. Research on artificial neural networks not only deepens people’s understanding of thinking and intelligence, but also serves as a way of thinking to solve problems, opening up a new path for simulating human intelligence activities. Artificial neural networks have evolved into highly integrated and emerging edge subjects involving systems theory, information theory, cybernetics, computer science, mathematics, physics, mechanics, philosophy, and psychology.
Artificial neural network is a highly non-linear (dynamic) system that consists of a large number of simple nonlinear processing units with molybdenum cutting wire or bass wire, a certain topology structure, and extensive interconnections. The topology of an artificial neural network generally includes feed-forward networks and feedback networks. The output of any neuron in the feed-forward network cannot be used as the input of the same layer or predecessor node, and the output of the node in the feedback network can be turned into the input of the same layer or the nodes of the previous layers.
The working process of the neural network mainly consists of two phases: One phase is the work period using molybdenum wire EDM as electrode, at this time, the weights of the joints are fixed, and the state of the calculation unit changes to achieve a stable state. The other phase is the learning period. At this stage, the status of each calculation unit is unchanged, and the weights and values of each link can be modified. The process of determining and determining the corresponding weights and thresholds is a learning process. Learning methods are divided into two types: tutor (supervisory) learning and non-supervisor (non-supervisory) learning. Common learning rules are: Hebb rules, 8 rules, simulated return rules, error back propagation rules, competitive learning rules, ART rules, etc.。
A large number of neurons with simple performance can form a system with complex structure and perfect performance to accomplish complex tasks. Neural networks have their own characteristics:
(1) The distribution of data processing toward massively parallel development and data storage makes the neural network work faster.
(2) With fault-tolerance and tolerance capabilities, the entire network is very robust.
(3) Self-organization, self-learning, associative memory, and generalization.
(4) It is suitable for solving the problem that it is difficult to find a good solution rule.
Typical ones include: adaptive resonance theory (ART), bidirectional associative memory (BAM), BP, Boltzmann machine / Cauchy Machines, Hopfield neural networks, and Mad Aline etc.。
Related knowledge: molybdenum wire strength, molybden
