Development of energy consumption model of abrasive machining process by a combined evolutionary computing approach
Introduction
Abrasive machining is widely used in aerospace manufacturing industries for achieving the desired surface finish of engineering components. Since power is an essential and expensive component for driving the machining operations, the saving of energy would result in higher environmental performance and productivity. Some studies conducted by peer researchers [1], [2], [3] reveal that more focus has been paid in optimizing the metal cutting operations based on cost and productivity. Recently, few studies are conducted that address the implications of conducting machining operations on its environmental performance [4], [5]. However, to the best of authors’ knowledge, any study that discusses the implications of abrasive machining of components on the environmental aspect is hardly noticed.
The mechanics of abrasive machining process with respect to input process variables has been well investigated in literature studies. Nouhi et al. [6] presented a novel technique using shadow masks which can be moved over the surface of the machined component. This allowed in direct measurement of features from the surface. Çelik et al. [7] investigated the effects of cutting parameters and geometrical features of novel α/β-SiAlON drilling tools on the cutting forces and the peel-up delamination of the machined holes. They observed that the point angle and chisel edge length of the novel SiAlON drilling tools are the main parameters that affect the maximum thrust force and delamination during drilling. Hadavi et al. [8] presented a novel model capable of predicting the instantaneous particle orientation and velocity within and downstream of the nozzle. It was evident from their results that formulation of mathematical models may have important implications for optimizing solid particle erosion tests and in abrasive jet machining applications. Wu et al. [9] proposed a new ultra-precision magnetic abrasive finishing process using low frequency alternating magnetic field. They demonstrated that the surface roughness of SUS304 stainless steel plate was improved from 240.24 nm to 4.38 nm by their novel approach. Saidi et al. determined the factors governing the dust emission during granite polishing in order to reduce the risk of exposition while producing quality parts. They performed experiments on a vertical CNC machining center using an adaptable tool holder which controlled the contact pressure during the tests. The studies established that the polishing tool grit size governs not only the granite surface finish but also the fine particle emission and the chip removal mechanism.
Some finite element modeling (FEM) studies have also been dedicated to understand the mechanics of abrasive machining process. Barletta et al. [10] modeled the material removal during fluidized bed of ductile metals by combining the theory of localization of plastic deformation during abrasive–workpiece impacts with an energy absorption approach. Hence, analytical modeling of abrasive machining process is also gaining popularity due to their cost effectiveness and time savings factor. However, the formulation of these models requires a thorough knowledge on the functionality and the configuration of the abrasive machining system.
In addition to finite element modeling of abrasive machining process, standardized Computational Intelligence (CI) techniques such as Genetic Programming (GP), Multi-Adaptive Regression Splines (MARS), and Artificial Neural Networks (ANN) can be used as an alternative method for modeling complex physical non-linear systems such as abrasive machining process. GP offers the advantage of fast and cost-effective explicit formulation of a mathematical model based on multiple variables with no existing analytical models. However, application of a standardized GP method may not perform satisfactory on the test data. Poor performance of the model on the test data implies its poor generalization ability and results in false information of the process in uncertain input process conditions. In addition, the data samples are costly to obtain and therefore the experts are looking for the high fidelity models that can predict the process behavior in extreme conditions. It is also well learned from the literature [11], [12] that by integrating the characteristics of the two or more methods, the generalization ability of the given method can be improved. This forms the motivation to introduce a new hybrid concept of parallel mechanism based on two potential CI methods of GP and MARS.
In the present work, the combined MARS–GP approach is proposed, which comprises of MARS model and GP error model in parallel. The error model is formulated using the GP approach based on evaluating the errors of the predictions of the MARS model. During the training of GP, the four process parameters, viz. component, machine acceleration, abrasive grade and machining time. Based on these inputs, a computational model for power consumption is formulated which needs to be optimized for improving the environmental sustainability. The MARS–GP model performance is further compared to that of standardized MARS model. The parametric and sensitivity analysis is further conducted to validate the robustness of the proposed approach.
Section snippets
Experimental study of abrasive machining process
The data for model development in our study is obtaining by conducting comprehensive experiments on abrasive machining process. Three types of workpiece components were tested in our study, viz. steel, nickel and titanium. The workpiece component is taken into consideration in the mathematical model by the component density. In addition, the abrasive grade is defined by the abrasive grain size in our study (hereafter referred to as “abrasive”). At the beginning, we loaded the component in the
Combine MARS–GP approach
In order to understand the notion of MARS–GP approach, firstly each of them is discussed as follows.
Genetic Programming (GP) generates the models automatically based on the given data using the Darwinian principle of “Survival of the fittest” [13]. Working principle of GP is same as GA but the only difference between them is that GA evolves solutions represented by strings (binary or real number) of fixed length whereas GP generates solutions represented by tree structures of varying sizes. GP
Evaluating proposed models against the experimental obtained data
The results obtained from the two models are illustrated in Fig. 4 on the training and testing data. Performance of the proposed models is evaluated using the five metrics: the square of the correlation coefficient (R2), the mean absolute percentage error (MAPE), the RMSE, relative error (%) and multi-objective error function (MO) given by:where and
Study of power consumption based on input variables
Sensitivity and parametric analysis about the mean is conducted for the validation of the proposed MARS–GP models. The sensitivity analysis (SA) percentage of the outputs to each input parameter is determined using the following formulas:where and are, respectively, the maximum and minimum of the predicted output over the ith input domain, where the other variables are equal to their mean values.
Fig. 5 shows the sensitivity analysis results
Practical significance of proposed model in abrasive machining process
The parametric analysis provides a measure of the relative importance among the inputs of the mathematical model and illustrates the response of measured outputs with variations in specific input parameters (Fig. 6). On the formulated MARS–GP model, the first input is varied between its mean ± definite number of standard deviations, and the power consumption is computed, while, the other input is fixed at its mean value. This analysis is then repeated for the other input. Fig. 6 display the plots
Conclusions
The present work proposed a combined MARS–GP approach for formulating models for predicting the power consumption in abrasive machining process. The performance of proposed model was found to be in good agreement with the actual data obtained from the experiments. Based on the sensitivity and parametric analysis, the dominant input process parameters and the hidden non-linear relationships are unveiled, which further validate the robustness of the proposed MARS–GP model. The higher
References (22)
- et al.
A review of optimization techniques in metal cutting processes
Comput. Ind. Eng.
(2006) Optimization of cutting parameters for minimizing power consumption and maximizing tool life during machining of Al alloy SiC particle composites
J. Cleaner Prod.
(2013)- et al.
Investigating the minimum quantity lubrication in grinding of Al2O3 engineering ceramic
J. Cleaner Prod.
(2014) - et al.
Optimization of process parameters using a response surface method for minimizing power consumption in the milling of carbon steel
J. Cleaner Prod.
(2014) - et al.
Implementation of a shadow mask for direct writing in abrasive jet micro-machining
J. Mater. Process. Technol.
(2015) - et al.
Investigation on the performance of SiAlON ceramic drills on aerospace grade CFRP composites
J. Mater. Process. Technol.
(2015) - et al.
Measurements and modeling of instantaneous particle orientation within abrasive air jets and implications for particle embedding
Wear
(2015) - et al.
Study on ultra-precision magnetic abrasive finishing process using low frequency alternating magnetic field
J. Magn. Magn. Mater.
(2015) - et al.
An integrated SRM-multi-gene genetic programming approach for prediction of factor of safety of 3-D soil nailed slopes
Eng. Appl. Artif. Intell.
(2014) - et al.
Stepwise approach for the evolution of generalized genetic programming model in prediction of surface finish of the turning process
Adv. Eng. Softw.
(2014)
Application of soft computing techniques in machining performance prediction and optimization: a literature review
Int. J. Adv. Manuf. Technol.
Cited by (6)
Particle swarm optimization tuned cascade control system in an Internet of Things (IoT) environment
2018, Measurement: Journal of the International Measurement ConfederationCitation Excerpt :The successes of these methods are dependent on the analysis of system behavior and its accuracy. Researchers have demonstrated several useful artificial intelligence based computational models to predict the effect of process parameters on surface characteristics of finished materials [27–34]. Many industries utilize the internet work as an effective and organized tool to enable systematic remote maintenance activities.
Recent Advancements and Challenges of Abrasive Jet Machining: A Review
2024, Lecture Notes in Mechanical EngineeringExploration of MFOA in PAC parameters on machining Inconel 718
2022, Materials and Manufacturing ProcessesCutting factors and testing of highly viscoelastic fluid abrasive flow machining
2021, International Journal of Advanced Manufacturing TechnologyPolishing path planning of complex curved surfaces and the realization of automatic control
2016, International Journal of Smart Home