Advances in Production Engineering & Management Volume 11 | Number 4 | December 2016 | pp 311-323 http://dx.doi.Org/10.14743/apem2016.4.229 ISSN 1854-6250 Journal home: apem-journal.org Original scientific paper A case-based reasoning approach for non-traditional machining processes selection Boral, S.a, Chakraborty, S.a* department of Production Engineering, Jadavpur University, Kolkata, West Bengal, India A B S T R A C T A R T I C L E I N F O To sustain in the modern era of rapid manufacturing development, it becomes necessary to generate complex shapes on materials which are highly temperature and corrosion resistant, hard to machine, and have high strength-to-weight ratio. Generation of complex shapes on those materials using conventional machining processes ultimately affects surface finish, material removal rate, accuracy, cost, safety etc. Non-traditional machining (NTM) processes have the capability to machine those advanced engineering materials with satisfactory results. But, selection of the most appropriate NTM process for a particular machining application is often a complicated task. Case-based reasoning (CBR), a domain of artificial intelligence, is a paradigm for reasoning new problems from the past experience. In CBR, a memory model is assumed for representing, indexing and organizing past similar cases, and a process model is supposed for retrieving and modifying the past cases and assimilating the new ones. This paper primarily focuses on the application of CBR approach for NTM process selection. Based on different process characteristics and process parameter values, the past similar cases are retrieved and reused to solve a current NTM process selection problem. For this, a software prototype is developed and three real time examples are cited to illustrate the application potentiality of CBR system. © 2016 PEI, University of Maribor. All rights reserved. Keywords: Non-traditional machining processes Process selection Artificial intelligence Case-based reasoning *Corresponding author: s_chakraborty00@yahoo.co.in (Chakraborty, S.) Article history: Received 7 June 2016 Revised 25 October 2016 Accepted 12 November 2016 1. Introduction With the development of newer materials having improved thermal, mechanical and chemical properties, it has now become quite difficult to machine those materials using conventional machining processes. These processes, generally based on cutting and abrasion mechanism, incur higher machining cost while generating complex shape features on composites, ceramics and other advanced engineering materials. The achieved surface quality and dimensional accuracy of the machined components are also not satisfactory, and often fail to meet the desired target In these machining processes, unwanted material from the parent workpiece is generally removed employing mechanical energy. This energy is supplied by means of a cutting tool kept in contact with the workpiece, causing shear deformation along the shear plane, leading to chip formation. New exotic work materials and complex geometrical shapes on those materials have been putting more pressure on the capabilities of the conventional machining processes. This leads to the development and deployment of a new set of machining processes, popularly known as non-traditional machining (NTM) processes. In these processes, unwanted material is removed from the parent workpiece using various forms of energy, like chemical, thermal, mechanical, electrical or combination of those energies. In an NTM process, there is no direct contact between the 311 Boral, Chakraborty cutting tool and the workpiece. In abrasive jet machining process, excess material is removed by means of microscopic chips and in electrochemical machining process by electrolytic dissolution. In laser beam machining process, there is even no need of any cutting tool. It is also not necessary that the cutting tool should be harder than the workpiece material in an NTM process. Now-a-days, it has become easier to generate complex shapes on materials, like steel, carbide, titanium and its alloys, ceramics, superalloys (Inconel 718, hastelloy) etc. employing NTM processes [1,2]. Till date, there have been approximately 20 NTM processes developed and applied in modern manufacturing industries. Selection of the best suitable NTM process for a particular work material and shape feature combination is generally made by a domain expert on the basis of various factors, such as workpiece material, shape feature to be generated, material removal rate, surface finish, surface damage, corner radii, tolerance, cost, safety, power requirement etc. Thus, an expert in this domain must have a vast and in-depth knowledge about the characteristics and capabilities of different available NTM processes. But, in the present manufacturing scenario, most of the process engineers lack the requisite domain knowledge and availability of experts is also sometimes constrained. Usually, a domain expert acquires knowledge from the past experience as well as from other reliable sources. Taking this concept as a plinth, when an expert attempts to select an NTM process for a given machining application, he/she just recalls the similar past situations and their solutions. Thus, based on the similar past problems and their solutions, new NTM process selection cases are solved. This entire cognitive process of a domain expert's thinking has given birth to a new branch of artificial intelligence (AI) technique, known as case-based reasoning (CBR) approach. This CBR approach is applied here for NTM process selection. In this paper, in order to choose the most suitable NTM process for a specific machining application, an exhaustive case-base containing the machining characteristics of various available NTM processes and their pertinent process parameters is first created. These machining characteristics and process parameter data are later used to select the feasible NTM processes according to the end requirements. The selection procedure is based on retrieval of the best matched case from the case-base using the nearest neighbourhood technique, while calculating the similarity score between two cases. The best matched case, which is retrieved from the case-base according to the values of different process characteristics as set by the process engineer/end user, has the similarity score greater than the other cases. To automate and simplify the application of CBR approach in NTM process selection, a software prototype having a graphical user interface (GUI) is designed and developed in Visual Basic 6.0. The developed system simultaneously considers both the user requirements (product characteristics) and technical requirements (process characteristics) for a given NTM process selection problem. 2. Literature review Using two multi-attribute decision making (MADM) tools, i.e. analytic hierarchy process (AHP) and technique for order preference by similarity to ideal solution (TOPSIS), Yurdakul and Cfogun [3] attempted to simplify the NTM process selection procedure for the manufacturing personnel. A list of feasible NTM processes satisfying the users' requirements was first generated and those processes were then ranked based on their suitability to meet the desired machining operation. An expert system was developed by Chakraborty and Dey [4] for selecting the best NTM process under constrained material and machining conditions. It would rely on the priority values of different criteria and sub-criteria for a specific NTM process selection problem, and the NTM process with the highest acceptability index was finally identified. Chakraborty and Dey [5] developed a quality function deployment (QFD)-based expert system for NTM process selection. An overall score for each of the NTM processes was estimated using the weights extracted from the house of quality matrix for various process characteristics. The overall scores of some of the NTM processes simultaneously satisfying certain critical criteria requirements were again compared and the NTM process having the maximum score was finally selected as the optimal choice. A web-based knowledge base system was proposed by Edison Chandraseelan et al. [6] for identifying the most suitable NTM process to meet some input parametric requirements, 312 Advances in Production Engineering & Management 11(4) 2016 A case-based reasoning approach for non-traditional machining processes selection like type of the work material, shape application, process economy, and other process capabilities, like surface finish, corner radii, width of cut, tolerance etc. Sadhu and Chakraborty [7] applied an input minimized Charnes, Cooper and Rhodes (CCR) model of data envelopment analysis for NTM process selection. Employing weighted-overall efficiency ranking method of MADM theory, the efficient NTM processes were ultimately ranked in descending order of their priorities. Temufin et al. [8] designed a fuzzy decision support model for NTM process selection while assessing the potentials of some distinct NTM processes. Chatterjee and Chakraborty [9] proved the application potentiality of evaluation of mixed data (EVAMIX) method for solving NTM process selection problems using three demonstrative examples. Roy et al. [10] integrated fuzzy AHP and QFD techniques for selection of NTM processes based on some predefined customers' perspectives. Temufin et al. [11] solved the NTM process selection problem under fuzzy and crisp environment, and proposed a decision support model to guide the process engineers to explore the potentials of some distinct NTM processes. The applicability of the proposed model was also validated. Khandekar and Chakraborty [12] applied fuzzy axiomatic design principles for selection of NTM processes. Madic et al. [13] demonstrated the applicability, suitability and computational procedure of operational competitiveness ratings analysis (OCRA) method for solving NTM process selection problems. Nowadays, CBR as a part of cognitive science, has been emerged out as an interesting research topic. Amen and Vomacka [14] employed CBR approach as a tool for selection of material and heat treatment process from an exhaustive database to simplify the task of a designer. Khe-mani et al. [15] applied CBR approach in fused cast refractory manufacturing industry. Fang and Wong [16] applied a hybrid CBR approach in agent-based negotiation for effective supply chain management Armaghan and Renaud [17] adopted CBR approach to prove the complementary nature of multi-criteria decisions and CBR approach. Although the past researchers applied numerous MADM methods and developed different distinct decision aids for selection of NTM processes for varying machining applications, but till date, no attempt has been put forward on selection of NTM processes using CBR approach. This paper thus proposes development of a decision making model based on CBR approach for selecting the best suited NTM process for a given machining application. It is observed that CBR is the correct and simplest approach in this domain where availability of experts is sometimes constrained. In CBR approach, a set of feasible NTM processes is first retrieved from the case-base satisfying the work material and shape feature combination. Based on the user and technical requirements, it then identifies the best matched NTM process from the stored similar cases. The past cases are just reused here for NTM process selection for providing the optimal solution. 3. CBR approach Intelligence, being a part of cognitive science, can be defined as the process involving rational and abstract thinking. It is often goal oriented and purposeful. It consists of knowledge and feats, both conscious and unconscious, which are acquired through continuous study and experience. The AI is actually the intelligence in machines. Intelligent system is the basement of knowledge engineering. It involves several tasks, like knowledge acquisition, creation of a knowledge base, knowledge representation and use of the acquired knowledge. The represented knowledge is basically used for reasoning or inference. In AI, knowledge is represented using symbols along with heuristics or rules of thumb. While using these heuristics, one should not have to rethink when a similar problem is encountered. The expert system can be defined as an intelligent computer program that uses knowledge and inference procedures to solve problems that are difficult enough requiring significant human expertise for their solution. Basically in expert system, knowledge is represented using 'if-then' rules. The CBR approach is a part of AI technique that utilizes information stored in the knowledge base, when similar past problems are encountered again. It provides solution to the present problem that is almost similar to the past. In CBR approach, a problem is represented as an input in the present situation. It just retrieves the most similar case to the new one from its case-base while calculating the similarity score over the defined parameters. It first searches the case history Advances in Production Engineering & Management 11(4) 2016 313 Boral, Chakraborty and chooses that case having the closest similarity to the current problem. In CBR system, the case-base is well structured and documented. The case representation may be flat, where all cases are represented at the same level, or it can be hierarchical, expressing relationship between cases and sub-cases. There are four major steps that constitute a CBR system, i.e. retrieve, reuse, revise and retain. Thus, it is also called as 4-R cycle or CBR cycle, as shown in Fig. 1. When a problem occurs in the current situation, similar past situations are retrieved from the case-base. Reusing the past cases, a predictable solution to the current problem is thus provided. If there is a need of any revision, the retrieved data are revised and retained as a new case in the case-base for future use [18-21]. New case -------------- Confirmed [Retain / ) Retrieve Most solution " * Case-Base -► similar cases Fig. 1 A CBR cycle or 4-R cycle Retrieving the most similar case along with the solution is based on some logical expressions. The similarity between two cases is usually measured with respect to each parameter. It also depends on the type of parameter (beneficial or non-beneficial) being used. The followings are the most common methods for calculating similarity between two cases: a) Numeric: Sim (a,b) = |a - bl/Range where Range is the difference between the upper and lower boundaries of a set. b) Symbolic: Sim(a,b) = 1 if a = b = 0 if a * b (1) c) Multi-valued: Sim(a, b) = Card (a) n Card (b) Card (a) U Card (b) where Card is the cardinality (size) of a set. d) Taxonomy: (2) (3) Sim(a, b) = h (common node (a, b)) (4) min(h(a),h(b)) where h is the height (number of levels) of the specified taxonomy tree. The procedural steps of a CBR approach are presented as below: a) A solution is first defined using several parameters. One of the parameters should be chosen carefully so that it would remain unique throughout the documentation procedure, e.g. case number. b) A huge set of known solutions is put into the case-base of CBR system. An existing database can also be used for this purpose. c) The CBR system generally reads the database and organizes a copy of its own. 314 Advances in Production Engineering & Management 11(4) 2016 A case-based reasoning approach for non-traditional machining processes selection d) The user generally formulates a query according to the end requirements. All the available variables are first displayed. The user has the option to choose all or few variables based on the problem statement. The query includes those variables as set by the user. The user also has the option to allocate different priority weights to the considered variables. e) As a result of the user-defined query, CBR system may display a number of cases or the best matched case. It may also be possible that none of the cases would match the query exactly. Favouring CBR technique as the most efficient tool for NTM process selection is a challenging task, as several other approaches have already been available for the same purpose. It is observed from the available literature that none of the MADM methods, like AHP, EVAMIX, TOPSIS etc. can provide complete solution when the domain is ill-structured and murky. The working principle of CBR is based on some available specific experiences instead of abstracted rules. It is considered as a useful tool if the utilization of prior experience is more vital than to produce a thoroughly optimized solution according to the specifications. The CBR approach has no optimizing potentiality, but it can be used for searching, not for calculations. Its efficiency is determined by fast retrieval of the most similar cases from the case-base. The principle of CBR also states that it can find the similarities between cases but not reasons. So, it is unable to judge how important the encountered departures are that can be determined only by an experienced user. A comparison between the existing search techniques and the adopted CBR approach is elucidated in Table 1. Table 1 Comparison between different search techniques and CBR approach Method Flexibility Operational approach Computational time Programming complexity Decision maker's involvement Type of data Genetic algorithm Medium (lack of learning ability) Artificial neural High network Simulated annealing Expert system Medium Medium CBR High Population based probabilistic search and optimization technique High (based on the desired accuracy and termination criterion) Mimics the working principle of biological High neurons Cooling process of molten metal is modeled artificially to construct an optimization algorithm Exact matching of input and stored data producing several Medium 'if-then' rules for inference Notion of similarity between present and prior stored cases Low High Medium (based on the desired accuracy and termination criterion) Medium High Medium Low High High High Medium Numerical Medium Numerical Numerical Both numerical and textual Both numerical and textual 4. CBR-based approach for NTM processes selection Although CBR approach has already been successfully applied in various fields of mechanical engineering, such as material selection, design selection, parts selection for automobile industries etc., no attempt has still been made for its application in the domain of NTM process selection. The CBR approach has the potential to provide complete information about a case where minimum information is available to the user. It yields the best results when the user provides detailed query information. Advances in Production Engineering & Management 11(4) 2016 315 Boral, Chakraborty While selecting the most suitable NTM process for a particular machining application, the process engineer has to consider several machining characteristics of the available NTM processes. In the developed CBR approach-based decision making model, nine NTM processes, i.e. abrasive jet machining (AJM), abrasive water jet machining (AWJM), electric discharge machining (EDM), laser beam machining (LBM), ultrasonic machining (USM), electrochemical machining (ECM), electrochemical discharge machining (ECDM), plasma arc machining (PAM) and wire electric discharge machining (WEDM) are taken into consideration. As the process characteristics, type of the workpiece material, shape feature to be generated, material removal rate (MRR) (in mg/min), surface roughness (SR) (in [im), surface damage (SD) (in [im), tolerance (Tol) (in mm), overcut (OC) (in mm), corner radii (CR) (in mm), taper (TP) (in mm/mm), cost (C) (in relative (R) priority scale), power (P) (in kW) and safety (S) (in R scale) are considered. For cost, the R scale is set as 1 - lowest, 2 - very low, 3 - low, 4 - medium, 5 - high, 6 - very high and 7 - highest. On the other hand, for safety, the R scale is set as 1 - highly safe, 2 - safe and 3 - attention required. As work materials, a) aluminium, b) aluminium alloys, c) ceramics, d) composites, e) glass, f) steel, g) superalloys and h) titanium are considered in this model. The above-mentioned NTM processes can generate a) hole (precision) (0.03 mm < D < 0.13 mm), b) hole (standard) (L/D < 20), c) hole (standard) (L/D > 20), d) through cut (shallow) (t/w < 2), e) through cut (deep) (t/w > 2), f) through cavity (standard) (t/w > 10), g) through cavity (precision) (t/w < 10), h) pocket (shallow) (t < 1 mm), i) pocket (deep) (t > 1mm) and j) surface of revolution feature on the work material (where L is the length of the hole, D is the diameter of the hole, t is the thickness and w is the width of the machined feature). The relevant machining characteristics data for different NTM processes are accumulated from experimentations, machining data handbooks and other reliable resources to create the corresponding case-base. The collected data are then organized in a structured manner in MS Access. The step-wise operational procedures of the developed CBR system for selecting the best suited NTM process for a particular machining application are depicted as follows: Step 1: When the developed CBR system starts to run, the first screen, as shown in Fig. 2, appears to the end user where the type of the work material to be machined and type of the shape feature to be generated can be chosen from the given options as the primary inputs to the system. Step 2: After clicking 'OK' button, a list of feasible NTM process(es) capable of generating the desired shape on the specified work material is displayed. For this, Eqn. (2) is utilized for filtering and retrieving the data. Step 3: When the user presses 'Next' button, another screen, as shown in Fig. 3, is displayed to identify the most suitable NTM process from the list of feasible processes while satisfying the set machining requirements. Step 4: In this screen, the end user has to choose the desired process characteristics based on which the final NTM process selection is made. Step 5: When 'Enter range' functional key is clicked, the required number of empty cells are automatically generated where the input ranges for the selected NTM process characteristics can be provided. Step 6: After inputting the desired ranges of values, pressing of 'Best NTM process' button identifies the most suitable NTM process for the specified machining application while satisfying the set criteria values. For retrieving the best NTM process in this step, Eqn. (1) is employed. Step 7: The actual retrieved values of all the technical characteristics for the best matched NTM process are also displayed. Step 8: When 'Best NTM process' button is clicked, the technical details (tentative settings of the associated process parameters) of the best matched NTM process are also available, as shown in Fig. 4. 316 Advances in Production Engineering & Management 11(4) 2016 A case-based reasoning approach for non-traditional machining processes selection Although in step 5, there is an option for entering ranges of process characteristic values, but if the developed CBR system does not find any data within those ranges from the case-base, it would retrieve the possible data nearest to the query set. For a particular NTM process selection problem, MRR is the sole beneficial attribute where its value is always required to be maximized. On the other hand, SR, SD, Tol, TP, OC, CR, C, P and S are non-beneficial attributes requiring their minimum values. The best matched case should have the highest similarity score, which is calculated with respect to each of the process characteristics. After summing up these similarity scores for the set process characteristics for each case, the NTM process having the highest similarity score is chosen as the most suitable option. 5. Illustrative examples 5.1 Example 1: Standard hole on composite material In this example, standard holes are to be generated on a composite material. After providing the inputs of composite as the work material and hole (standard) as the shape feature options in the primary selection window of Fig. 2, a set of feasible NTM processes consisting of AJM, AWJM, ECDM, ECM, EDM, LBM and USM is displayed, when 'OK' button is clicked. All the processes can generate standard holes on composite materials. In the next window of Fig. 3, MRR, SR, Tol, OC, CR and C are opted as the most important process characteristics based on which the final NTM process selection is to be made. In this example, the desired input ranges for those process characteristics are set as MRR 100-1000 mg/min, SR 2-12 ^m, Tol 0-0.5 mm, OC 0-0.05 mm, CR 0-0.5 mm and C 1-4 (in R scale). Now, when 'Best NTM process' functional button is clicked, LBM process is identified as the best matched case, capable of meeting the set process characteristic values. It is interesting to observe that apart from the set process characteristics, values of the other process characteristics are also available for the best matched NTM process. In this example, the selected LBM process can achieve values of MRR as 286.08 mg/min, SR as 2.63 [im, SD as 102 [im, Tol as 0.02 mm, OC as 0.001 mm, CR as 0.5 mm, TP as 0.05 mm/mm, C as 1 (in R scale), P as 0.23 kW and S as 3 (in R scale). In Fig. 4, the process engineer can also have an idea about the settings of different machining parameters of LBM process. These are the tentative process parametric settings and for achieving the maximum machining performance, fine-tuning of these settings is often necessary. A real time photograph of LBM process is also available in Fig. 4. Fig. 2 Primary selection window for Example 1 Advances in Production Engineering & Management 11(4) 2016 317 Boral, Chakraborty Final selection window using CBR approach Process characteristics w Material removal rate (MRR) P Surface roughness (SR) r Surface damage (SD) I? Tolerance (Tol) p Overcut (OC) p Corner radii (CR) r Taper (TP) p Cost (C) r Power (P) r Safety (S) Select process characteristics and input ranges Material removal rate (mg/min) Surface roughness (pm) Tolerance (mm) Overcut (mm) Corner radii (mm) Cost (1-7) (R scale) Enter range 0.05 Best NTM process LBM R Scale for cost : 1 ^Lowest, 2- Very low, 3= low, 4= Medium, 5 = High, 6= Very high, 7= Highest R Scale for safety : 1 -Highly safe, 2- Safe, 3- Attention required Matched case : Process MRR SR SD Tol OC CR TP C P S 1 LBM 1286.08 2.63 1 102 0.02 0.001 1 0.Q5 1 0.05 I I 1 I 0.23 I 3 I Fig. 3 Best NTM process for Example 1 0 LBM El Laser beam machining (LIBM1 Pulse power (kW) Gas pressure (Bar) Cutting speed (mm/min) Stand off distance (mm) 0.23 9.8 0.5 Pulse frequency (Hz) Pulse width (ms) Nozzle dia (mm) Focal length (mm) i ^ " Process |_BM Workpiece material Composite isl si Exact material AA7075/SÍC MMC m * ^ Gas used Nitrogen Shape Hole (Standard) [LiD<20] 230 0.2 1.2 116 Fig. 4 Details of LBM process 5.2 Example 2: Standard through cavity on ceramics Here, the process engineer wants to generate a standard through cavity on a ceramic work material. In the primary selection window, as shown in Fig. 5, the developed CBR approach first extracts five NTM processes, i.e. AJM, AWJM, EDM, USM and WEDM as the feasible options satisfying the said work material and shape feature combination requirement. In Fig. 6, MRR, SR, Tol, OC, CR, C and S are chosen as the most important process characteristics based on which the final NTM process needs to be selected. Based on the ranges of values for these process characteristics, USM process is identified as the best matched case for this machining application. For USM process, the attainable process characteristics are MRR as 131.96 mg/min, SR as 0.66 [im, SD as 25 [im, Tol as 0.014 mm, OC as 0.15 mm, CR as 0.08 mm, TP as 0.005 mm/mm, C as 5 (in R scale), P as 0.4 kW and S as 1 (in R scale). 318 Advances in Production Engineering & Management 11(4) 2016 A case-based reasoning approach for non-traditional machining processes selection El. Primary selection window ^ i ■Work material- Ceramic Shape Through cavity (Standard) [t/w>10] OK Feasible NTM process(es) AJM AW J M EDM USM WEDM Next Fig. 5 Primary selection window for Example 2 In Fig. 7, the tentative parametric settings and the technical specifications of USM process along with its actual photograph are displayed to guide the process engineer to achieve the best machining performance. 5 Final selection wl ndow us i ng CBRapp roach H I i Select process characteristics and input ranges Process characteristics P Material removal rate (MRR) w Surface roughness (SR) r Surface damage (SD) P Tolerance (Tol) W Overcut (OC) F Corner radii (CR) r Taper (TP) W Cost (C) r Power (P) W Safety (S) Enter range Material removal rate (mg/min) 10 100 Surface roughness (^im) 2 20 Tolerance (mm) 0 0.05 Overcut (mm) 0 0.01 Corner radii (mm) 0 0.05 Cost (1-7) (R scale) 1 4 Safety (1-3) (R scale) 1 2 i Best NTM process! USM R Scale for cost : 1=Lowest, 2- Very low, 3= low, A- Medium, 5 = High, 6= Very high, 1- Highest R Scale for safety : 1=Highly safe, 2- Safe, 3= Attention required Matched case : Process MRR SR SD Toi OC CR TP USM 131.946 0.66 25 0.014 0.15 0.08 0.005 0.4 Fig. 6 Best NTM process for Example 2 Advances in Production Engineering & Management 11(4) 2016 319 Boral, Chakraborty q, usm Ultrasonic machining (USM) Process Shape |USM Through cavity (Standard) [t/w>10] Workpiece material Exact material Tool material Abrasive used Slurry media Feed rate (mm/min) Amplitude of vibration (|jm) Ceramic Zirconla Stainless Steel Boron Carbide Water 1.08 32 Slurry concentration (%) Frequency of vibration (Hz) Material thickness (mm) 20 Fig. 7 Details of USM process 5.3 Example 3: Shallow through cutting on steel In this example, a shallow through cutting operation needs to be performed on a standard steel plate. For this work material and shape feature combination, the CBR system first recognizes AJM, AWJM, ECM, EDM, LBM and PAM as the six feasible NTM processes, as shown in Fig. 8. Then, in Fig. 9, seven process characteristics, i.e. MRR, SR, SD, Tol, OC, CR and C are identified by the process engineer for the final selection of the most suited NTM process for the considered machining application. In this window, the ranges of values of the set process characteristics are also provided. 5 Primary selection window (Si \w » I rWork material - Steel Shape Through cutting (Shallow) [t/w<2] OK Feasible NTM process(es) AJM - AWJM ECM — EDM - LBM - Next Fig. 8 Primary selection window for Example 3 320 Advances in Production Engineering & Management 11(4) 2016 A case-based reasoning approach for non-traditional machining processes selection Final selection window using CBR approach Process characteristics W Material removal rate (MRR) W Surface roughness (SR) R Surface damage (SD) F Tolerance (Tol) W Overcut (OC) W Corner radii (CR) r Taper (TP) I? Cost (C) r Power (P) r Safety