With the increasing demand for miniaturized products and micro profiles on the products, Micro-EDM (µEDM) is attracting the attention of researchers. µEDM can create a complicated micro profile on the products, therefore the process is finding usefulness in the aerospace, electronics, automotive industries, biomedical instrumentation, etc. µEDM works the same way as EDM, but with some differences in the amount of energy, plasma temperature, and electrode size. µEDM needs nano-Joule energy, which can be provided by an RC circuit. In contrast, macro EDM needs a DC source that can deliver pulse energy of 1J. The plasma temperature can be as high as 50,000°C and as low as 4,000°C [1].
To increase the machining effectiveness of EDM techniques, researchers are constantly focusing on a variety of different issues. One area of interest is the machining of conductive materials such as aluminum alloys, stainless steel, titanium and its alloys, tungsten, and Inconel alloys. Investigations are being made to estimate the process parameters that can improve the machining efficiency of these materials and their alloys [2],[3], [4]. Additionally, µEDM is being used to machine insulating materials as well. Machining of non-conducting materials like zirconia, silicon carbide, and many other ceramics is being done by µEDM using an additive electrode technique, making this process a versatile tool for various materials [5], [6]. Researchers continue to explore new materials and techniques to expand the capabilities of micro-EDM and its potential applications in various industries. The shape, size, and material of electrodes are being investigated to enhance the performance of the process. Various electrode shapes are being used to improve the flushing ability of the dielectric fluid. Helical, tubular, cylindrical, orbital, and planetary electrodes have been developed to decrease the tool wear rate (TWR) and to improve the MRR [7]. The electrode material is one of several factors that can impact performance parameters. Electrical resistivity and thermal conductivity can either increase or decrease TWR and MRR[8]. Researchers have investigated dielectric fluids, which help flush away the debris that can affect surface quality and MRR [9]. In a study by Gholipoor et al.[10], three types of dielectric fluids were compared: air, deionized water, and a mixture of air and deionized water. The method, known as dry-EDM, has been found to be very effective for mass manufacturing when air is employed as the dielectric fluid.
In µEDM processes, the gap between the electrodes is very small, typically they are a few microns apart. This can cause debris to get trapped between them and may lead to short circuits. To address this problem, some researchers have suggested adding a mixture of micro and nano-sized powder particles to the dielectric fluid. This technique is called "powder mixed µEDM " [11][12][13][14]. They found that adding particles of a certain size in the right concentration can decrease the TWR, increase the MRR and improve surface quality. Powder Mixed Micro Electro-Discharge Machining (PMMEDM) works similarly to other EDM processes, with the key difference being the formation of a plasma channel. In PMMEDM, particles suspended in the dielectric fluid move and concentrate at the strongest point in the plasma channel. These particles are charged and move in a zig-zag fashion due to the electric field that accelerates them and makes them conductive. The conductive particles lower the breakdown strength of the insulating dielectric, which causes them to cluster together and form a chain known as the "bridging effect." This effect increases the gap between the workpiece and electrode and modifies the plasma channel due to the added powder. At the machining zone, the discharge is evenly distributed among the powder particles, creating craters on the workpiece's surface, and enhancing the quality of the machined surface [15]. With the increase in the demand for micromachining, the research interest in the optimization of parameters and making the system sustainable has also increased. One concept that has been instrumental in optimizing the process is the artificial neural network (ANN). The idea of the ANN was first introduced in 1943 by McCulloch and Pitts, who outlined the first mathematical model of the neuron. Since then, the concept of the ANN has been further developed and refined, with the introduction of the back-propagation algorithm in 1986 [16] This algorithm has proven to be a powerful tool for minimizing error, and has been widely used in many fields, including pattern recognition and industrial processes. In recent years, many authors, including Coit et al.[17], Haykin [18], and Al-Ahmari [19], Sable et al. [20], Akar et al. [21] have chosen to utilize ANN to solve a variety of problems. One reason is that the ANN can learn through examples, making it an effective tool for generalizing the network for unseen data. To achieve the best results, the ANN must be trained with a set of data that is collected through experimental runs.
ANN can learn from experimental data sets to describe a non-linear and interaction effect with great success. It plays an important role in studying linear as well as non-linear problems. The advantage of using ANN for modeling is the construction of the model is very easy. A set of inputs is required, and the network can be trained to produce a targeted set of outputs. The application of neural networks to machining has been explored by many researchers. Researchers have applied ANN based on a backpropagation algorithm to build a relationship between input and output parameters [22][23]. Many have combined ANN with other algorithms like particle swarm optimization, grey regression analysis, and fuzzy logic and have optimized process parameters to enhance the machining efficiency [24][25][26]).
Badar et al. [27] in their adaptive sampling procedure used parameters such as manufacturing surface error patterns and optimization search methods to achieve reasonably high accuracy with a smaller sample size. They used end- and face- milled fabricated plates for modeling and validation. Mounayri et al. [28] developed and used an ANN model to predict cutting force and surface roughness in end milling. Verma et al. [29] used an un-replicated factorial design to optimize zinc coating thickness in hot-dip galvanization process. Jiang et al. [30] conducted experiments to investigate the characteristics of gas film in micro-electrochemical discharge machining. Li et al. [31] studied quantitative evaluation of bubble characteristics in micro-hole drilling using EDM.
It is obvious from the above literature review that several studies have used design of experiments, ANN, and other techniques in their investigations to optimize machining performance or predict characteristics. Micro-EDM has been a focus area for researchers recently because of increasing demand for micro-featured products. Hence, there is a necessity to improve the process capability of µEDM. This work aimed to investigate the impact of process parameters on the micro-drilling of blind holes using µEDM. The process parameters considered for the investigation included voltage, capacitance, and electrode material (electrical conductivity and thermal conductivity). The performance parameters were MRR, and the geometric profile obtained. To improve machining efficiency, an Artificial Neural Network model was developed to estimate process parameters. The paper is structured into several sections. Section 1 introduces the topic at hand, including the background and purpose of the study. Section 2 details the experimental setup, including the materials and methods used for the fabrication of the micro-blind hole. Additionally, this section provides an overview of the theoretical framework for the study. The experiment's findings are presented in Section 3, along with a thorough examination of the data gathered. Section 4 describes the prediction of MRR and SR of blind micro holes using ANN. Finally, Section 5 presents the conclusions drawn from the study, along with suggestions for future research on the topic.