Online optimization of multipass machining based ...

Int J Adv Manuf Technol DOI 10.1007/s00170-012-4163-7

ORIGINAL ARTICLE

Online optimization of multipass machining based on cloud computing M. Chandrasekaran & M. Muralidhar & U. S. Dixit

Received: 29 October 2011 / Accepted: 10 April 2012 # Springer-Verlag London Limited 2012

Abstract Cloud computing provides on-demand computing services and data. The cloud infrastructure can be owned by a private organization, a group of organizations, or the public. In this work, the feasibility of using cloud computing for the optimization of machining processes is explored. A web system is developed in which a main server keeps the repository of data and carries out optimization. The main server can provide the optimized process parameters on demand to various clients. The clients can use the optimized data and fine-tune them if necessary. The clients send the feedback to the server, which is utilized with the help of a probability-based approach. The case examples show the feasibility of using the system for helping in agile manufacturing. Keywords Multipass machining . Web-based machining . Optimization . Cloud computing

1 Introduction Starting from the late 1990s, web-based manufacturing is steadily gaining popularity. With the help of World Wide Web, information distributed at different locations can be M. Chandrasekaran : M. Muralidhar North Eastern Regional Institute of Science and Technology, Nirjuli 791 109, India U. S. Dixit (*) Department of Mechanical Engineering, Indian Institute of Technology Guwahati, Guwahati 781 039, India e-mail: [email protected] U. S. Dixit e-mail: [email protected]

accessed and shared by users anywhere in the world using web tools such as web browsers [1]. Recently, cloud computing has evolved from the web-based technologies. The cloud computing is a concept of providing computing services and data on demand. The National Institute of Standards and Technology [2] defines cloud computing as “a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management efforts or service provider interactions.” The following are the four deployment models of cloud computing [2]: 1. Private cloud: “The cloud infrastructure is operated solely for an organization. It may be operated by the organization or a third party and may exist on premise or off premise.” 2. Community cloud: “The cloud infrastructure is shared by several organizations and supports a specific community that has shared concerns. It may be managed by the organizations or a third party and may exist on premise or off premise.” 3. Public cloud. “The cloud infrastructure is made available to the general public or a large industry group and is owned by an organization selling cloud services.” 4. Hybrid cloud. “The cloud infrastructure is a composition of two or more clouds (private, community, or public) that remain unique entities but are bound together by standardized or proprietary technology that enables data and application portability.” From this brief introduction of cloud computing, it is evident that private, community, public, or hybrid cloud computing model can be used for getting the machiningrelated information and optimization of the machining performance.

Int J Adv Manuf Technol

When the cloud computing runs on the internet, it is vulnerable to some security threats from hackers. A number of authors have analyzed security aspects of the cloud and suggested suitable measures to counter security threats. Rosenthal et al. [3] suggested selective application of cloud such that highly confidential data are not put on the web, while other type of data is placed on the web cloud. Jamil and Zaki [4] suggested forming private cloud for selective data, which can interact with other types of clouds through a secure bridge with firewall. Zissis and Lekkas [5] proposed introducing a trusted third party, tasked with assuring specific security issues within a cloud environment. Optimization of machining process is one of the most widely investigated topic in the field of manufacturing, and researchers have used a number of traditional and nontraditional optimization techniques for obtaining optimum parameters in single and multipass machining [6–8]. However, the prediction of tool life, surface roughness, and cutting force is difficult due to complexity involved in machining processes, although the predictive models are very much needed for the optimization. There should also be a provision to fine-tune the offline optimization results during machining as there are a number of machine, tool, and cutting environment-related parameters that the predictive models cannot account for. Considering the availability of cloud computing infrastructure, it is possible to develop reliable predictive models and carry out the offline optimization in a cloud. The offline results supplied by the cloud can be fine-tuned at the client side, and feedback can be provided to the cloud. In the present work, the feasibility of such a system is assessed by developing a private cloud-based online optimization system. Compared to other types of clouds, a private cloud is much secure. It can also interact with outside world as per the requirement. Suitable security measures can be adopted. The developed private cloud uses a probability-based procedure for processing the information at the server side. The rest of the paper is organized as follows. The review of literatures in the area of web-based manufacturing and machining optimization are presented in Section 2. Section 3 discusses the structure of the developed system and the method of getting optimized parameters (with a probability-based approach) from cloud is discussed in Section 4. Online fine-tuning and feedback procedure is discussed in Section 5. Section 6 demonstrates use of online optimization method applicable to turning and milling processes with case examples. Conclusions are presented in Section 7.

2 Literature review This section presents a brief review of literature concerning the work of the present paper. The review is presented in two subsections. Section 2.1 discusses web-based manufacturing, and Section 2.2 discusses machining optimization.

2.1 Review on web-based manufacturing The application of computer and web technology in manufacturing industries connects different manufacturer and customers globally on wire. Yang and Xue [1] carried out a comprehensive review of recent research on developing web-based manufacturing systems. Several researchers have applied web technologies for various applications such as mechanical design, quality modeling, production scheduling, process planning, etc. Chen et al. [9] developed an Integrated Concurrent Engineering Design for Mechanical Parts system. In this system the user can access product information from different domains of computerized tools by interacting with the system through graphical user interface (GUI) interactive screens. Also the design can be evaluated based on criteria of design for manufacturing and design for assembly. Huang and Mak [10] developed a web-based quality function deployment (QFD) modeling system. It is a three-tier architecture system. The first tier connects QFD application clients with web server. The QFD web server and QFD application server are two kinds of middle tier, and QFD database server is third tier. The application server is responsible for dealing with database server for synchronization and computational activities. Li et al. [11] developed a web-based system for supply chain management employing heuristic method for collaborative decision making. Web-based production scheduling introduced by Jia et al. [12] obtains the optimal schedule considering different requirements and manufacturing resources distributed at different locations. Thakur and Pande [13] developed an internet-based system for feature modeling and process planning of sheet metal components commonly manufactured by blanking, shearing, and bending processes. In the area of process optimization and information sharing using web technologies, less numbers of research papers are available. Wang and Jawahir [14] developed web-based, user interactive expert system for optimization of single-pass milling for the selection of optimum cutting conditions using genetic algorithms. The system allows user to access relevant databases and calculate results according to user’s request/ input and generate HTML documents to send back corresponding information. Wu and Liao [15] developed an internet-based machining parameter optimization and management system using genetic algorithm (GA) for highspeed machining process. Zheng et al. [16] developed a web-based machining parameter selection system for agile turning which provides user appropriate selection of machine tools, cutting tools, and cutting parameters for reduction in product life cycle cost and increase in product quality. 2.2 Review on machining optimization The selection of optimum process parameters is an important step in machining, considering economic aspects. Researchers

Int J Adv Manuf Technol

have investigated single and multipass machining problems mostly for minimizing production cost or time. They used various conventional optimization techniques such as geometric programming, the sequential unconstrained minimization technique, goal programming, and dynamic programming [17–19]. These techniques may be useful for some specific problems, but they are inclined to converge to local optimal results. Due to complexity and uncertainty of the machining processes, various soft computing techniques are used by researchers for predicting the performance of the machining processes and optimizing them. Lee et al. [20] developed fuzzy nonlinear programming model for optimizing cutting conditions for a turning process. Onwubolu and Kumalo [21] used GA to optimize unit production cost and compared the result with simulated annealing (SA). Tandon et al. [22] optimized an end milling process with the combined feed forward neural network and the particle swarm optimization (PSO) for minimizing production cost. Saravanan and Sachithanandam [23] developed GA-based optimization procedure to optimize cutting parameters in surface grinding process. Results show that GA performs better than traditional quadratic programming. Wang and Jawahir [24] proposed a methodology using GA for the allocation of total depth of cut in multipass turning. Abburi and Dixit [25] developed an optimization methodology, which is a combination of a real coded genetic algorithm and sequential quadratic programming, to obtain Pareto optimal solutions for minimizing production cost. Rao and Pawar [26] optimized multipass milling process using three nontraditional methods to minimize production time subjected to various constraints. They found that the convergence rate for PSO and artificial bee colony algorithm is greater than SA. Shutong and Yinbiao [27] proposed a methodology for optimizing cutting speed, feed, and depth of cut for rough and finish pass to minimize unit cost. They used a pass enumerating method, in which the number of all possible rough cuts is calculated to divide the whole complicated problem into several subproblems, which in turn are solved by GA. Bharathi Raja and Baskar [28] optimized face milling process during machining of aluminum bar using carbide tool. They obtained optimum machining parameters to minimize machining time subjected to the desired surface finish. Apart from various optimization algorithms proposed, the implementation of optimal cutting conditions during machining is very important. However, shop floor applications of optimization methods are limited due to nonavailability of required information about the process readily. The approach for online optimization of machining process has not received attention of the researchers. Very recently, the concept of cloud computing and cloud manufacturing has started gaining popularity [29, 30]. Similar to cloud computing, cloud manufacturing aims to share manufacturing resources to make “manufacturing as service” environment. However, as rightly pointed out by Sun et al. [31],

“many issues about the cloud manufacturing are still confusing and more attention should be paid to advance progress in this field.” The present work is a step in that direction and concentrates on demonstrating the feasibility of applying cloud computing concepts to optimization of machining parameters.

3 System structure In the proposed cloud computing-based optimization (CCBOD) system, the cloud supports machining database and computing facilities. The user can get data as well computational service on demand. The real machining datasets obtained by online optimization are stored in “machining database” and maintained by web server. The physical setup of CCBOD is shown in Fig. 1. The server which hosts the CCBOD system has been configured to run in Windows environment and use Apache web server. The CCBOD system consists of five elements. They are the web server, the online optimization system, the machining database, retrieval and sharing of information, and the clients. Being the central link, the client interfaces web server via the Internet/Intranet by typing the internet protocol (IP) address. Every web server has an IP address known as domain name. Entering the uniform resource location (URL), the client’s internet browser sends request to the web server and the server then fetches the page and sends it to the client browser to establish remote link via Internet. 3.1 Online machining optimization The heuristic based online machining optimization system coded in MATLAB® with GUI is operated through interactive screens. The “online_optimization.exe” file is downloaded at client system, and the user/machine tool operator can implement the same for optimization of machining process at their location. With the input information, the user interacts with the system without knowing the details of the heuristic algorithm. The system suggests initial cutting conditions and

Fig. 1 Structure of cloud computing-based optimization

Int J Adv Manuf Technol

accepts feedback such as “actual surface roughness” and “tool life” from real machining environment and thereby optimizes the machining process online. 3.2 Machining database The database consists of organized data structures. The “MySQL” relational database system is used to store information in web server. The machining data viz. surface roughness and tool life obtained during online optimization process, under different cutting conditions, are stored. The clients contribute this information to the database, which is maintained by the server. The optimum cutting parameters are also stored in a separate database file. Different databases are maintained for different processes. 3.3 Retrieval of optimum cutting parameters and feedback The web-based implementation allows the users to access database information. The optimum cutting parameters for different process can be retrieved. For particular machining process, the user submits the input data (i.e., depth of machining and maximum allowable surface roughness value). The hypertext pre-processor (PHP) script fetches the information from the respective database for display on the browser screen. The user may also implement the cutting condition and give feedback. 3.4 System information flow The graphical representation of data flow is shown at two different levels. Figure 2 shows data flow diagram (DFD) at zero level. It depicts that clients/users interact with CCBOD. Initially user enters the URL of the website to enter into the system and accesses the information being transacted by the developed system. The level 1 of the data flow diagram shown in Fig. 3 has three main functions. They are: (1) online machining optimization system, (2) retrieval of database information, and (3) retrieval of optimum cutting parameters and user feedback about the same. The user enters the system and inputs informations such as process, work material type, tool material, work/ cutter diameter, machining length, depth of machining, and maximum allowable surface roughness. The system suggests initial cutting conditions for optimizing finish machining process. Unlike conventional offline optimization which employs

Fig. 2 DFD level 0

Fig. 3 DFD level 1

empirical relation to evaluate surface roughness and tool life in optimizing cutting parameters, the present methodology optimizes the process during production of components and does not use any empirical relations in the optimization process. The feedback information such as “actual surface finish” and “tool life” obtained by machining process are entered into the system. Machining data obtained from different machine tool operators help to develop predictive model by understanding probabilistic nature of the machining process. Based on this information, further cutting conditions are suggested in the future through cloud computing feature. The procedure is similar to the common practice followed by machine tool operator who fine-tunes the cutting parameters based on his experience as well as observation of machining performance. The data obtained by real machining process are stored in a database called “machining database.”

4 Obtaining the optimized parameters from cloud The online machining information related to different machining process is stored in the database. MySQL being open source database management system is used in this work. It is a relational database management system that runs as a server providing access to multiple users and to a number of databases. The system maintains four databases

Int J Adv Manuf Technol

each for turning, end milling, face milling, and grinding. The database contains the values of all parameters (fields) in table forms. The structure of the table consists of different field variables, its type, and extra features such as autoincrement and primary key used for indexing the datasets. The system has a feature for retrieval of the data values fetched from respective database and displayed at client’s system. The system also accepts client’s inquiry and provides updated database information through server calculation and displays optimum cutting parameters for specific machining process. The hypertext markup language (HTML) page will be displayed based on user option on the main page for inputting the information such as total depth of stock to be removed and maximum allowable surface roughness. The web server executes the entered information and retrieves data and submits the result to the browser in the form of HTML, which appears in the client’s system screen. The database is maintained by administrator for necessary updating and deletion of the database information. It is carried out at web server using structured query language statement. Figure 4 shows the coordinated working pages and HTML frame sheet of the developed system. The system manages the process through the front-end screens made up of HTML, cascading style sheets (CSS), JavaScript, and PHP codings. This provides an interface between the database and users to perform various functions. XAMPP being cross-platform software consisting of the Apache web server, MySQL, PHP, and Perl is used for operation of the system. Figure 5 shows typical screen shot of the XAMPP control panel application in the process. The tool life and surface roughness data are obtained from the stored database by applying the techniques of data mining. As the server obtains the feedback from various locations, its predictions are probabilistic in nature. The optimization parameters have to be found with probabilistic input. As a simple example of how it can be done in finish

Fig. 5 XAMPP control panel

machining, consider that the feedback from three machines predicted the tool life function as T1(f, v), and other nine machines predicted a tool life of T2(f, v). Hence, the probability that tool life function is T1 is 0.25 and the probability of tool life function being T2 is 0.75. For a given tool life function T1, the cost per piece is C1(f, v), a function of feed f, and cutting speed v. Similarly for a given tool life function T2, the cost per piece function is C2(f, v). Now, based on the mathematical expectation, the expression for the cost per piece is given as Cðf ; vÞ ¼ 0:25C1 ðf ; vÞ þ 0:75C2 ðf ; vÞ

ð1Þ

The above expression can be minimized to obtain the minimum cost of machining. Once sufficient amount of data is available, the discrete probabilities can be replaced by the probability distribution of tool life. Let p(f, v) be the probability distribution of tool life and CT(f, v, T(f, v)) denotes the cost per piece function for a tool life T. Based on mathematical expectation, the cost per piece is given as Tmax Z ðf ;vÞ

CT ðf ; v; T ðf ; vÞÞpðf ;vÞdT

Cðf ; vÞ ¼

ð2Þ

Tmin ðf ;vÞ

where Tmin(f, v) and Tmax(f, v) are the minimum at a maximum tool life at process parameters f and v. The above expression can be minimized by any optimization method. In this work, because of scarcity of data and simplicity, the discrete model has been used.

5 Online fine-tuning and sending feedback

Fig. 4 Functional working pages of ‘CCBOD’ system

The adoption of cloud computing feature helps to optimize multipass machining of different processes. An online optimization methodology is accessed via web/internet or intranet technologies and applied to turning, end milling, face

Int J Adv Manuf Technol

milling, and grinding processes. The machining feedback is sent to the server through clouds. The real machining data provided by different clients/users are stored in database server. It helps to build up predictive models needed for optimization. The information is fine-tuned and provides useful data to users through cloud services. 5.1 Methodology The clients at different locations access online optimization system through web browsers and operate at their locations for optimization of machining process. The presentation layer, i.e., namely GUI facilitates easy interaction with the system by the machine tool operator. On entering basic informations such as name of the machining process, tool and work material, work or cutter diameter, length of machining, total depth of stock to be removed, maximum allowable cutting force, maximum allowable surface roughness, and limits of cutting parameters, the system suggests the suitable initial cutting conditions for rough machining. The result of machining process such as actual cutting force and surface roughness is fed back into the system. Based on this information, improved cutting conditions are suggested and the process is optimized for roughing as well as finishing process during production of components. The optimization objective is to minimize the cost applied to multipass machining process.

In a multipass machining process, the production time to produce a component [32] is given by   tc pL Tp ¼ 1 þ ðmD0  mðm  1ÞdR Þ TR vR fR   tc pL ðDf þ 2dF Þ þ TL þ tts ; ð3Þ þ 1þ TF vF fF where L is the machining length and D is the work/cutter diameter. vR, fR, dR, m, and TR are cutting speed, feed, depth of cut/pass, number of passes, and tool life for roughing passes; vF, fF, dF, and TF are cutting speed, feed, depth of cut, and tool life for finishing pass. D0 and Df are initial and final diameter of work piece. TL is loading and unloading time, tc is the tool/cutter change time, and tts is total tool reset time. The total production cost is given by   TtR TtF ; ð4Þ þ CT ¼ C0 TP þ Ctc TR TF where C0 is the operating cost in US dollars per minute, and Ctc is the tool change cost in US dollars per cutting edge. TtR and Ttf are the machining times in roughing and finishing respectively. Substituting the value of Tp in Eq. (4), the total production cost being the summation of cost for roughing passes, cost for finishing pass, and cost involved for loading and unloading is given by [33]:

h i pL ðmD  mðm  1Þd Þ þ ðC t þ C Þ ðmD  mðm  1Þd Þ þ C ðm þ 1Þt CT ¼ CR þ CF þ CL ¼ C0 vpL 0 R 0 c tc 0 R 0 s v R f R TR R fR h i pLðDf þ2dF Þ pL þ C0 vF fF ðDf þ 2dF Þ þ ðC0 tc þ Ctc Þ vF fF TF þ C0 TL For constant values of L, D, tc, C0, and Ctc, the total production cost is minimized to obtain optimum cutting parameters for roughing and finish passes satisfying allowable cutting force and surface roughness. Considering statistical variation in machining, it is aimed that the actual surface roughness in finish pass lies between the acceptable minimum and maximum limits. On rough machining at selected cutting condition, if the actual cutting force is within the maximum allowable value, machining is continued to produce components and tool life data are obtained. During finish machining the process is continued if the selected cutting condition satisfies cutting force constraint and the actual surface roughness falls within the limits. The data cloud obtained for different cutting conditions and from different machine tool operators helps to develop predictive tool life model which is used for optimization. Considering tool life variation due to its probabilistic nature, the production cost per piece is evaluated. The system suggests improved cutting conditions and obtains feedback. The process is optimized at the server

ð5Þ

using suitable optimization technique. The user can access the information for necessary implementation at their location. 5.2 Web implementation The developed cloud-based system used for machining optimization is based on the Internet. The system has client– server architecture; both the client programs and the server (administrator) programs are preserved in the server side with the feature that the client system automatically downloads informations connecting via web browser. The system uses HTML, CSS, and PHP scripts for clients/users to interact with the system. Being the central link, the client interfaces web server via the Internet typing IP address. Every web server has an IP address known as domain name. Entering the URL, the client’s internet browser sends request to the web server and the server then fetches the page and sends it to the client browser to establish remote link via Internet. Figure 6 visualizes the communication structure, describing the different feature of the CCBOD system.

Int J Adv Manuf Technol Table 1 Input details Description

Typical input data

Machining details Process

Turning

Workpiece material

Steel (0.4 % C)

Tool material Diameter of workpiece

TiN-coated carbide 30 mm

Length of machining

Fig. 6 Communication structure of ‘CCBOD’ system

The user interactive online optimization of multipass machining scheme is coded in MATLAB® with GUI. The user can browse the system for implementation at their location and submit feedback data. The informations are stored in a database server, which can be accessed by the user. Figure 7 shows main screen of developed CCBOD system. The system can be accessed through common web browsers such as Windows IE, Mozilla Firefox, Google Chrome, Safari, and Opera at client system. Server side database is maintained using MySQL which stores all information in “machining database” at server and managed by the administrator. The web server handles clients request for obtaining optimum cutting parameters. The information is provided to the user by fetching data from the database. The user also sends the feedback of the same through HTML and PHP pages.

6 Examples In this section the web implemented online optimization methodology for obtaining optimum cutting parameters for two common machining processes, viz., turning and milling process is demonstrated.

Fig. 7 Main screen of ‘CCBOD’ system

Maximum depth of stock to be removed Constraints details

100 mm 3.0 mm

Allowable cutting force

1,000 N

Allowable surface roughness Cutting speed (lower bound)

3.25 μm 140 m/min

Cutting speed (upper bound)

240 m/min

Feed (lower bound) Feed (upper bound)

0.04 mm/rev 0.32 mm/rev

Depth of cut (lower bound)

0.1 mm

Depth of cut (upper bound)

1.0 mm

6.1 Online optimization of multipass turning process Consider a multipass turning process to be optimized for obtaining optimum process parameters. Let total depth of stock to be removed be 3.0 mm and allowable surface roughness value ( R*a ) 3.25 μm. The maximum allowed cutting force is 1,000 N. The variable bounds are 140≤v (m/min)≤240, 0.04≤f (mm/rev)≤0.32, 0.1≤d (mm)≤1.0. The input information for the optimization problem is shown in Table 1. Figures 8 and 9 show snapshots of the input screens. In this work, actual machining has not been carried out. Instead a code was written for making a semivirtual lathe. Actual machining experiments are replaced by simulated results. The semivirtual lathe consists of three modules: (1) cutting force prediction module based on empirical relation, (2) surface roughness prediction module based on a neural network (NN) model, and (3) tool life prediction model based on empirical relation. The cutting force is obtained by using empirical relation [32] and it is given as:

Fig. 8 Input screen 1

Int J Adv Manuf Technol

Fig. 9 Input screen 2

Fz ¼ 858:33f 0:248 d 1:6819

ð6Þ

where Fz is the cutting force in newtons, f is the feed in millimeters per revolution, and d is the depth of cut in millimeters. The NN module is developed using the experimental data selected from the work of Kohli and Dixit [34] for dry turning of steel using TiN-coated carbide tools. For the given cutting condition (v, f, d), the network predicts the surface roughness. The evaluation of tool life is obtained by using Taylor’s extended tool life equation and it is given as: vT n1 f n2 d n3 ¼ C;

ð7Þ

where v is in meters per minute, tool life T is in minutes, f is in millimeters per revolution, d is in millimeters, C is constant, and n1, n2, and n3 are exponents. The online optimization roughing process starts with selection of initial cutting condition suitably so that actual cutting force is within limit. The depth of cut in all roughing

process is taken as equal. The number of passes is decided such that the maximum depth of cut constraint is not violated. In the present problem, the depth of cut for finishing and two roughing passes comes out as 1.0 mm. As the optimization function is a two-variable (v, f) problem, three initial machining conditions are generated and optimized using simplex search method. There are many nonconventional optimization methods proposed in the literature. However in this work, the simplex search method is used to optimize the parameter. To initiate optimization, three cutting conditions are selected as (v1, f1), (1.1v1, f1), and (1.1v1, 1.1f1). For the present problem, initial cutting condition v1 (0vmin) as 140 m/min and f1 (0fmin) 0.04 mm/rev is chosen. The cutting force is obtained as 386.3 N, which is within limit. The machining is continued with first set of cutting condition (v 0140 m/min, f 0 0.04 mm/rev, d01 mm/pass) till the tool fails and obtained tool life 315.2 min is entered in the system. The snapshots of the interactive screens are as shown in Fig. 10. The display of spindle speed along with cutting speed facilitates operator for quick setting of cutting parameters. As different machine tool operators employ this methodology, the real tool life data are supplied to the clouds. These data are fine-tuned at the server side by adoption of cloud computing feature and help to optimize the machining process. The feedback is provided to the users via cloud service. The use of cloud computing combines multiple machining data and provide comprehensive result to the users. As an example consider tool life data of 12 machine tool operators, in which 3 machines predicted the tool life that follow Taylor’s extended tool life function as vT 0:2 f 0:15 d 0:15 ¼ 273

ð8Þ

and the other 9 machines follow tool life function as vT 0:2 f 0:18 d 0:12 ¼ 276:

ð9Þ

For equal depth of cut in all roughing passes, the tool life becomes function of T(f, v). Also the rough machining cost per piece for m02 and dR 01.0 mm/pass is given as   pL pLv4R fR0:25 CR1 ¼ C0 ð2D0  1Þ þ ðC0 tc þ Ctc Þ ð2D  1Þ þ 3C t 0 0 s vR fR 2735

ð10Þ and

  pL pLv4R fR0:1 CR2 ¼ C0 ð2D0  1Þ þ ðC0 tc þ Ctc Þ ð2D  1Þ þ 3C t 0 0 s vR fR 2765

ð11Þ

Fig. 10 Snapshots of the screen during roughing process: a display of initial cutting condition and b tool life entry

for three and nine machines respectively. C0 is the operating cost in US dollars per minute, and Ctc is the tool change cost in US dollars.

Int J Adv Manuf Technol Table 2 Iterative results of optimization for roughing process Actual cutting force (Fact) (N)

Roughing cost per piece (CR) (US$)

0.07 0.10

455.6 484.3

1.64 1.57

0.14

527.1

1.31

Iteration no.

Suggested iterative cutting conditions (vR, dR, fR) (m/min, mm, mm/rev)

1 2

168.0 118.0

1.0 1.0

3

129.0

1.0

(Optimum)

Based on probabilities of tool life functions of 0.25 and 0.75, the roughing cost per piece is h i pLv4R fR0:25 CR ¼ 0:25 C0 vpL ð2D0  1Þ þ ðC0 tc þ Ctc Þ 273 ð2D0  1Þ þ 3C0 ts 5 R fR h i pLv4R fR0:1 ð2D0  1Þ þ ðC0 tc þ Ctc Þ 276 ð2D0  1Þ þ 3C0 ts : þ0:75 C0 vpL 5 R fR

ð12Þ Substituting L0100 mm, D0 030 mm, C0 00.5 US$/min, Ctc 02.5 US$/min, tc 05 min, and ts 00.51 min/pass, Eq. (12) is simplified as CR ¼

9:27 28:9v4R fR0:25 86:84v4R fR0:1 þ þ þ 0:765: vR f R 2735 2765

ð13Þ

To optimize the function, three sets of cutting conditions are guessed as (140.0, 0.04), (154.0, 0.04), and (154, 0.05), respectively. The evaluated costs are US$2.43, US$2.34, and US$2.03. Simplex search optimization [35] is used for optimization. Normally, it takes three iterations to reach optimum. The iterative cutting conditions and roughing costs are shown in Table 2. The cutting force obtained is within the limit of 1,000 N. The optimum process parameters are obtained as vopt 0129.0 m/min and fopt 00.14 mm/ rev. The optimum cost for roughing/piece is US$1.31. The proposed scheme obtains reliable optimum cutting parameters by its direct implementation without deviating cutting force constraint. Figure 11 shows the final display

Fig.12 Snapshots of the screen during finish machining process: a display of initial cutting condition, b surface roughness entry, and c tool life entry

screen for roughing process showing the real machining data clouds stored in database of the system. In optimization of finish turning process, the initial cutting condition are selected as v1 (0vmax) and f1 (0fmin), and the constant depth of cut is equal to dmax. For the present problem v1, f1, and d1 are 240 m/min, 0.04 mm/rev, and 1 mm, respectively. The allowable surface roughness value is (R*a ) 3.25 μm. Considering some statistical variation, the ba ) is considered between the actual surface roughness ( R Table 3 Iterative results of optimization for finishing process

Fig. 11 Final display for roughing process

Iteration no.

Suggested iterative cutting conditions (vF, dF, fF) (m/min, mm, mm/rev)

1 2 3 4

134.5 198.3 153.0 214.1 (Optimum)

1.0 1.0 1.0 1.0

0.07 0.07 0.11 0.10

Surface roughness obtained ba ) (μm) (R

Cost per piece (CF) (US$)

3.07 2.47 2.93 2.47

0.45 0.34 0.27 0.23

Int J Adv Manuf Technol

ba < 1:2Rt where Rt lower and upper limits such that 0:8Rta < R a a ba , the feed is increased may be taken as 0.8R*a. If 0:8Rta > R using the following relation [36]:  t 1:5 Ra f1 ; ð14Þ ðf1 Þnew ¼ ba R

availability of more and more real life machining data, the system database enriches with information and helps to develop predictive models for machining optimization. It is also fine-tuned at server side based on feedback information provided by the user.

where f1 is feed in millimeters per revolution. The exponent 1.5 has been chosen based on some experimental study. If ba > 1:2Rt , the system displays “Surface roughness cannot R a be achieved with this set of work tool combination.” ba ) may lie between 2.08 and The actual surface roughness (R 3.12 μm. The machining of component is performed, and the obtained cutting force and CLA surface roughness are 386.3 N and 2.12 μm, respectively. If surface roughness falls below lower limit, the feed is increased suitably with v remaining same. Now machining is continued till the tool fails and obtained tool life data are entered in the system. The snapshots of the interactive screens during finish machining process are shown in Fig. 12. Based on probabilities of tool life functions of 0.25 and 0.75, the finishing cost per piece is

Consider the optimization of multipass end milling process. Aluminum alloy workpiece is machined with 20-mm diameter high-speed steel (HSS) end mill cutter. The cutter has 4 number of tooth (z). Length of machining is 100 mm. Total depth of stock to be removed is 5 mm, and allowable surface roughness value (R*a ) is 2.75 μm. Maximum allowable cutting force is 500 N. The machining is performed on semivirtual milling which consists of the following modules: (1) surface roughness prediction module based on NN, (2) tangential cutting force prediction module based on empirical relation, and (3) tool life module based on empirical relation. NN module is developed using an experimental data from the work of Lou and Chen [37] for end milling of aluminum alloy using HSS end mill cutter. The tangential cutting force is evaluated from empirical relation obtained from experimental data presented in [38]. Tool life is evaluated by the relation proposed by Tolouei-Rad and Bidhendi [39], and it is given as 0  g 11=n 0:2d 60 @ C fz A ð17Þ T¼ ðdfzÞw v Q

CF ¼ 0:25CF1 þ 0:75CF2

ð15Þ

where CF1 and CF2 are finishing costs corresponding to tool lives that follows Eqs. (8) and (9), respectively. Substituting L0300 mm, Df 024 mm, C0 00.5 US$/min, Ctc 02.5 US$/ min, tc 05 min, and ts 00.51 min/pass, Eq. (15) is simplified as CF ¼

4:08 10:21v4F fF0:25 30:62v4F fF0:1 þ þ vF f F 2735 2765

ð16Þ

To optimize the function (Eq. 16), three sets of cutting conditions are guessed as (240, 0.04), (216, 0.04), and (216, 0.05), respectively. The evaluated costs are US$0.57, US$0.56, and US$0.47. Simplex search optimization takes four iterations to reach optimum. The iterative cutting conditions and costs are shown in Table 3. The surface roughness obtained is within the limit of 2.08 and 3.12 μm. The optimum process parameters for finish machining are obtained as vopt 0214.1 m/min and fopt 0 0.10 mm/rev. The optimum cost for finishing/piece is US$0.23. Total production cost per piece is US$1.92. The proposed scheme obtains reliable optimum cutting parameters by its direct implementation without causing any loss/rejection of components due to surface finish. With the  C ¼ CR þ CF þ CL ¼ C0

6.2 Online optimization of multipass end milling process

where T is the tool life in minutes, v is the cutting speed in meter per minute, C is the constant for 60 min of tool life when the area of cut is 1 mm2, f is the feed in millimeters per tooth, z is the number of tooth, d is the depth of cut in millimeters, and g, w, and n are exponents for different tool and work material combination. The factor Q is the contact proportion of cutting edges with workpiece per revolution. The ranges of cutting parameters are as follows: Cutting speed (v)—45–90 m/min Feed (f)—0.025–0.23 mm/tooth Depth of cut (d)—0.25–1.27 mm The optimization problem is to determine optimum cutting velocity and feed/tooth through online machining process for minimizing total production cost based on cloud computing feature. The total production cost is given by

 pLDm pLDm þ ðC0 tc þ zCtc Þ þ C0 ðm þ 1Þts 1000vR fR z 1000vR fR zTR   pLD pLD þ C0 TL þ ðC0 tc þ zCtc Þ þ C0 1000vF fF z 1000vF fF zTF

The online optimization of roughing process starts with selection of initial cutting condition suitably so that actual

ð18Þ

cutting force is within limit. The machining is initialized with v1 (0vmin) as 45 m/min and f1 (0fmin) as 0.025 mm/

Int J Adv Manuf Technol

tooth. The radial depth of cut (d) is 1.24 mm and number of passes are 3. The obtained cutting force is 106 N, which is within limit. Thus machining is continued with first set of cutting condition (v045 m/min, f00.025 mm/tooth) till tool fails, and obtained tool life data are entered in the system. Consider the tool life provided by eight machine tool operators. Two machines predicted tool life that follows tool life function of Eq. (17) as  0:14 11=0:15 0:2d 33:98 fz 60 B C T1 ¼ A @ 0:29 ðdfzÞ0:35 v 0

ð19Þ

The tool life predicted by other six machines follows tool life function as  0:16 11=0:15 0:2d 33:98 fz 60 B C T2 ¼ A @ 0:38 0:29 ðdfzÞ v 0

ð20Þ

Considering the probability of tool life function of 0.25 and 0.75, the rough machining cost per piece is given by CR ¼ 0:25CR1 þ 0:75CR2

CF ¼ 0:25CF1 þ 0:75CF2 ¼

0:785 þ 7:2 vF fF

ð23Þ

2:27 2:6 þ 4:36  1010 v5:67 1010 v5:67 F fF F fF

The optimization is initialized to minimize the function Eq. (23) with three set of cutting conditions as (90, 0.042), (81, 0.052), and (73, 0.052). Optimum cutting parameters vopt 063.2 m/min and fz(opt) 00.11 mm/tooth are obtained in three iterations. The optimum cost for finishing/piece is US $0.21. Total production cost/piece is US$2.09. The cutting force and surface roughness obtained at optimum cutting condition are 395 N and 2.46 μm, respectively. Figure 13 shows display of optimum cutting conditions for finish milling process. The iterative cutting conditions are provided by the system in a user-friendly interactive GUI screen based on optimization source code written in MATLAB® 7.10 running at Pentium-4 with 512 MB RAM. In the proposed scheme, optimization is carried out during machining by selecting cutting conditions suitably.

ð21Þ 7 Conclusions

With the constant values of costs, length, and diameter, the objective function to minimize roughing cost is expressed as CR ¼

2:355 2:27 þ 2:115  109 v5:67 þ 3:606 R fR vR f R 2:6  1010 v5:67 R fR þ 1:02:

ð22Þ

The optimization is initialized with three set of cutting conditions as (45, 0.025), (49.5, 0.025), and (49.5, 0.028). Optimum cutting parameters vopt 053.4 m/min and fz(opt) 0 0.12 mm/tooth are obtained in four iterations. The optimum cost for roughing/piece is US$1.51. The optimization of finish milling process considers the initial cutting condition as v1 (0vmax), and f1 (0fmin) with depth of cut equal to dmax. The initial cutting parameters are 90 m/min, 0.025 mm/tooth, and 1.27 mm. The machining of component is performed, and the obtained cutting force and CLA surface roughness are 106 N and 1.43 μm, respectively. The limit of ^ a ) is between 1.76 the actual surface roughness value (R and 2.64 μm. Now the feed is increased following Eq. (14) with cutting velocity remaining the same. This increases the surface roughness value, and thus, the first set of cutting condition is obtained as 90 m/min and 0.042 mm/tooth. The CLA surface roughness is 2.26 μm. Considering the probabilities of tool life function as 0.25 and 0.75, the finish machining cost per piece is given by

The developed CCBOD system provides an efficient userfriendly environment for optimization of machining process. The proposed optimization scheme suggests cutting conditions for shop floor machine tool operator, accepts feedback, and optimizes the machining process during the production of components. The methodology coded in MATLAB® with GUI helps user to interact with the system easier. The optimization model does not use any empirical relation and/or offline experimentation for evaluation of tool life and surface roughness in the optimization process. The web-enabled feature helps for accessing and sharing information among the users located at anywhere in the globe via web browsers. The client can also retrieve optimum cutting parameters using PHP scripts, and MySQL

Fig. 13 Display of optimum cutting parameters for finish milling process

Int J Adv Manuf Technol

queries with the result can be viewed at clients’ screen. The system works on private cloud and is available at IIT Guwahati intranet with URL http://172.16.72.110/home.html. In the future, it is planned to place this system on the internet. The datasets obtained through online machining can also be used to develop an artificial intelligence-based online learning methodology for optimizing cutting parameters. Finally, the system has advantage of open access, applicable to optimization of different process, and simplified interactive access and provides opportunity for sharing the process data information among different manufacturers. Acknowledgment The authors gratefully acknowledge the financial help provided by All India Council for Technical Education (AICTE) from the project AICTE: 8023/RID/BOIII/NCP (21) 2007–2008, project identification number at Indian Institute of Technology, Guwahati being ME/P/USD/4.

16.

17. 18. 19.

20.

21. 22.

23.

References

24.

1. Yang H, Xue D (2003) Recent research on developing Web-based manufacturing systems: a review. Int J Prod Res 41(15):3601–3629 2. Mell P, Grance T (2009) Perspectives on cloud computing and standards. National Institute of Standards and Technology (NIST). Information Technology Laboratory 3. Rosenthal A, Mork P, Li MH, Stanford J, Koester D, Reynolds P (2010) Cloud computing: a new business paradigm for biomedical informatics. J Biomed Inform 43:342–353 4. Jamil D, Zaki H (2011) Cloud computing security. Int J Eng Sci Tech 3(4):3478–3483 5. Zissis D, Lekkas D (2012) Addressing cloud computing security issues. Future Generat Comput Syst 28:583–592 6. Aggarwal A, Singh H (2005) Optimization of machining techniques— a retrospective and literature review. Sadhana 30(6):699–711 7. Mukherjee I, Ray PK (2006) A review of optimization techniques in metal cutting processes. Comput Ind Eng 50(1–2):15–34 8. Chandrasekaran M, Muralidhar M, Krishna CM, Dixit US (2010) Application of soft computing techniques in machining performance prediction and optimization: a literature review. Int J Adv Manuf Technol 46(5–8):445–464 9. Chen KH, Chen SJG, Lin L, Changchien SW (1998) An integrated graphical user interface (GUI) for concurrent engineering design of mechanical parts. Comput Integrated Manuf Syst 11:91–112 10. Huang GQ, Mak KL (2001) Issues in the development and implementation of Web applications for product design and manufacture. Int J Comput Integrated Manuf 14:125–135 11. Li D, McKay A, DePennington A, Barnes C (2001) A Web-based tool and a heuristic method for cooperation of manufacturing supply chain decisions. J Intell Manuf 12:433–453 12. Jia HZ, Fuh JYH, Nee AYC, Zhang YF (2002) Web-based multifunctional scheduling system for a distributed manufacturing environment. Concurrent Eng 10:27–39 13. Thakur AD, Pande SS (2006) Web based feature modeling and process planning for sheet metal components, 1st International & 22nd AIMTDR conference. IIT Roorkee, Roorkee 14. Wang X, Jawahir IS (2004) Web-based optimization of milling operations for the selection of cutting conditions using genetic algorithms. Proc IME B J Eng Manufact 218:648–655 15. Wu MP, Liao WH (2005) Internet-based machining parameter optimization and management system for high-speed machining.

25. 26.

27.

28.

29. 30. 31. 32.

33. 34.

35. 36.

37.

38.

39.

Trans of Nanjing University of Aeronautics & Astronautics 22 (1):42–46 Zheng L, Yang XM, Zhang ZH, Liu TI (2008) A web-based machining parameter selection system for life cycle cost reduction and product quality enhancement. Comput Ind 59:254–261 Ermer DS (1971) Optimisation of the constrained machining economics problem by geometric programming. ASME J Eng Ind 93:1067–1072 Sundaram RM (1978) An application of goal programming technique in metal cutting. Int J Prod Res 16:375–382 Gopalakrishnan B, Khayyal FA (1991) Machine parameter selection for turning with constraints: an analytical approach based on geometric programming. Int J Prod Res 29:1897–1908 Lee YH, Yang BH, Moon KS (1999) An economic machining process model using fuzzy non-linear programming and neural network. Int J Prod Res 37:835–847 Onwubolu GC, Kumalo T (2001) Optimization of multipass turning operations with genetic algorithm. Int J Prod Res 39:3727–3745 Tandon V, El-Mounayri H, Kishawy H (2002) NC end milling optimization using evolutionary computation. Int J Mach Tools Manuf 42:595–605 Saravanan R, Sachithanandam M (2001) Genetic algorithm (GA) for multivariable surface grinding process optimisation using a multiobjective function model. Int J Adv Manuf Technol 17:330–338 Wang X, Jawahir IS (2005) Optimization of multi-pass turning operations using genetic algorithms for the selection of cutting conditions and cutting tools with tool-wear effect. Int J Prod Res 43:3543–3559 Abburi NR, Dixit US (2007) Multi-objective optimization of multipass turning processes. Int J Adv Manuf Technol 32:902–910 Venkat Rao R, Pawar PJ (2010) Parameter optimization of a multipass milling process using non traditional optimization algorithms. Appl Soft Comput 10(2):445–456 Shutong XIE, Yinbiao GUO (2011) Intelligent selection of machining parameters in multi-pass turnings using a GA-based approach. J Comp Inf Sys 7(5):1714–1721 Bharathi Raja S, Baskar N (2012) Application of Particle Swarm Optimization technique for achieving desired milled surface roughness in minimum machining time. Expert Syst Appl 39:5982–5989 Xu X (2012) From cloud computing to cloud manufacturing. Robot Comput Integrated Manuf 28:75–86 Marston S, Li Z, Bandyopadhyay S, Zhang J, Ghalsasi A (2011) Cloud computing—the business perspective. Decis Support Syst 51:176–189 Sun H, Neng W, Chang Z, Mo R (2011) Approach to optimization of part machining service combination. Int J Adv Manuf Technol 56:767–776 Ojha DK, Dixit US, Davim JP (2009) A soft computing based optimization of multi-pass turning processes. Int J Mater Processs Technol 35(1/2):145–166 Yeo SH (1995) A multipass optimization strategy for CNC lathe operations. Int J Prod Econ 40:209–218 Kohli A, Dixit US (2005) An neural-network-based methodology for the prediction of surface roughness in a turning process. Int J Adv Manuf Technol 25:118–129 Deb K (2005) Optimization for engineering design—algorithms and examples. Prentice-Hall, New Delhi Chandrasekaran M, Muralidhar M, Krishna CM, Dixit US (2012) Online machining optimization with continuous learning. In: Davim JP (ed) Computational methods for optimizing manufacturing technology: models and techniques. IGI Global, Hershey Lou SJ, Chen JC (1999) In process surface roughness recognition (ISRR) system in end-milling operation. Int J Adv Manuf Technol 15:200–209 Wan M, Lu MS, Zhang WH, Yang Y (2012) A new ternary-mechanism model for the prediction of cutting forces in flat end milling. Int J Mach Tools Manuf. doi:10.1016/j.ijmachtools.2012.02.003 Tolouei-Rad M, Bidhendi IM (1997) On the optimization of machining parameters for milling operations. Int J Mach Tools Manuf 37(1):1–16