Abstract
In this research, a data-driven approach to metamodeling of manufacturing/machining processes is developed. Instead of the conventionally used second-order polynomial regression metamodels, a non-predefined form-free approach is discussed. The highly adaptive metamodeling strategy, called symbolic regression, is carried out by using genetic programming. A central composite design based experimental dataset on electric discharge machining is used as the training and the testing data. Four different process parameters namely (voltage, pulse on time, pulse off time, and current) are used as the independent parameters to quantify three different responses (material removal rate, electrode wear rate, and surface roughness). The performance of the metamodels are evaluated by using various statistical metrics like R2, MAE, MSE. The performance of the metamodels on the training and testing data is found to be adequate for all the responses.
TopIntroduction
When the conventional machining processes are unable to process the materials, non-traditional machining like electric discharge machining (EDM) are used. EDM produces high accuracy and is applicable for any conductive material. EDM is a non-traditional machining process (NTMP) in which material removal takes place with the help of thermal energy generated spark generated through electricity flow between the electrode tool and workpiece under the envelope of dielectric fluid. Similar to some NTMPs like laser cutting process; it does not require force to remove material. EDM is popular in tool and mold making. EDM can be applied to machine any conductive material, even the hardest of materials such as tungsten carbide. It processes the material without force which makes it suitable for fragile materials too. Turning, cutting, drilling, milling, etc. operations can be performed using EDM. Applied voltage, current, pulse on and off time, tool gap etc. are the controlling parameters of EDM process and material removal rate (MRR), surface roughness, overcut, tool wear rate, electrode wear rate etc. are the machining output characteristics. The process parameters, their interaction and influence on the response are the main concerns of the researchers working in the field of machining of materials using EDM. Some researchers studied the effect of the process controlling parameters of the EDM process and measured the 3D surface topography for the surface quality of machined surface and stated that applied current in the most significant parameter (Ramasawmy & Blunt, 2004). The EDM drilling process parameters of Inconel 718 materials were studied by Kuppan et al. (Kuppan, Rajadurai, & Narayanan, 2008) for selecting the best combination of process parameters aiming maximum MRR and minimum surface roughness. Researchers also studied the effect of change in the electrode conditions in EDM for better outcomes. Pandey et al. (Srivastava & Pandey, 2012) applied the ultrasonic waves to cool the electrode during EDM of high-speed steel and reported that current machining system improved the surface integrity of the workpiece compared to conventional EDM.
During recent developments in the manufacturing industries, the market need is to generate the product with high quality for less time and cost. Achievement of this aim is possible through optimization of the various machining process parameters. Optimization has become a vital engineering tool which gives the best utilization of given resources for maximum possible outcomes. Different techniques used in this context are simulated annealing, particle swarm optimization, ant colony optimization, artificial bee colony etc. It is found in the literature that GA is one of the most preferred optimizers in the optimization of the machining process parameters (Yusup, Zain, & Hashim, 2012). It is based on the survival of fittest principle of genetics, which starts with a basic pool of alternative called population and head towards the best solution. However, to mathematically optimize the process an approximation function of the machining/manufacturing process is needed. This is because in general, no closed-form equations are available for expressing the machining/manufacturing process as a function of the process parameters. In this regard, researchers have relied on metamodeling approaches to form approximate relations between the independent process parameters and the dependent responses. Polynomial regression is perhaps the most widely used metamodeling strategy in machining/manufacturing processes. Metamodels based on evolutionary programming techniques are becoming very popular for optimization of parameters of the machining process. Genetic programming (GP) metamodel can be used to developed relationships between the input process parameters for superior output in case of various engineering applications. This approach significantly reduces computational time and cost (Kalita, Mukhopadhyay, Dey, & Haldar, 2019). Critical engineering problems like turbulent flow in pipe bends can also be solved using the GP metamodel-based evaluation with understanding the influence of different parameters affecting the system (Narayanan, Joshi, Dutta, & Kalita, 2019).
Key Terms in this Chapter
Regression: It is a statistical approach to express the process response as a function of the process parameters.
Process Response: The measured output (i.e., dependent variable) of a process is known as its response. Example—material removal rate, surface roughness, etc.
Predictive Model: Predictive models (also referred to as metamodels) are approximate functions of the actual function. They are inexpensive and can be used as surrogates to the actual functions.
Process Parameter: The various independent variables involved in a physical process or phenomenon may be referred to as process parameters.