International journal of computer integrated manufacturing , tập 24, số 6, 2011

International Journal of Computer Integrated Manufacturing Vol 24, No 6, June 2011, 517–534 Integration of process planning and scheduling: a state-of-the-art review Rakesh Kumar Phanden, Ajai Jain* and Rajiv Verma Department of Mechanical Engineering, National Institute of Technology, Kurukshetra, India (Received 12 July 2010; final version received February 2011) Process planning and scheduling functions strongly influence profitability of manufacturing a product, resource utilisation and product delivery time Several researchers have addressed the need for integration of process planning and scheduling (IPPS) functions to facilitate flexibility and for improving profitability of manufacturing a product, delivery time as well as creation of realistic process plans that can be executed readily on shop floor This article presents a state-of-the-art review on IPPS Three common integration approaches, non-linear approach, closed loop approach and distributed approach, are discussed with their relative advantages and disadvantages and reported research is classified accordingly It also identifies several future research directions Keywords: integration; process planning; scheduling; review Introduction Process planning and scheduling are two most important tasks in a manufacturing system These tasks strongly influence profitability of manufacturing a product, resource utilisation and product delivery time (Yang et al 2001) Process planning is the systematic determination of methods by which a product is to be manufactured economically and competitively The primary goal of process planning function is to generate process plans, which specifies raw material/components needed to produce a product as well as processes and operations necessary to transform raw materials into the final product Thus, outcome of process planning is the information required for manufacturing processes, including identification of machines, tools and fixtures Scheduling assigns a specific task to a specific machine in order to satisfy a given performance measure It is bound by process sequencing instructions that the process plan dictate and by the time-phased availability of production resources Thus, both process planning and scheduling involve assignment of resources and are highly interrelated Conventionally, process planning and scheduling are carried out in two distinct, sequential phases, where scheduling is done separately, after the process planning This approach is based on the concept of subdividing the tasks into smaller and separated duties to satisfy the requirements of suboptimisation and suitable for mass production (Larsen and Alting 1992) However, today’s manufacturing environment is quite different from traditional one It *Corresponding author Email: ajayjainfme@nitkkr.ac.in ISSN 0951-192X print/ISSN 1362-3052 online Ó 2011 Taylor & Francis DOI: 10.1080/0951192X.2011.562543 http://www.informaworld.com is characterised by decreasing lead time, exacting standards of quality, larger part variety and competitive costs In such manufacturing environment, it is difficult to get a satisfactory result using traditional approach due to following reasons: (Larsen and Alting 1992, Zhang and Merchant 1993, Gindy et al 1999, Morad and Zalzala 1999, Baykasoglu and O¨zbakır 2009, Li et al 2010a,b,c) (1) Process planner assumes that shop floor is idle and unlimited capacities of resources are always available in the shop Thus, process planner plans for the most recommended alternative resources This leads to the process planner favouring to select the desirable resources repeatedly Moreover, the resources are never always available on shop floor Therefore, unrealistic process plan will generate that may not be readily executed on shop floor (2) In conventional approach, fixed process plans restrict the schedule to only one machine per operation Therefore, possible choices of schedule using alternative machines are ignored (3) Even if the dynamic shop status is considered during process planning phase, the constraints considered in planning phase may change greatly because of time delay between planning phase and scheduling phase Thus, the generated process plan may become sub-optimal or invalid (4) Both, process planning and scheduling focus on single criterion optimisation to determine 518 R.K Phanden et al optimal solution However, real manufacturing environment involves more than one optimisation criterion The short comings of traditional approach can be overcome by considering an integrated approach to process planning and scheduling An integrated approach can respond better than traditional approach to present manufacturing environment and facilitate flexibility, improves profitability of a product, resource utilisation, product delivery time and creation of realistic process plans that can readily be executed without frequent alterations (Chryssolouris and Chen 1985, Sundaram and Fu 1988, Saygin and Kilic 1999, Lee and Kim 2001, Kumar and Rajotia 2003) Thus, integration of process planning and scheduling (IPPS) is essential to achieve eventually integrated manufacturing and to dismiss conventional manufacturing approach The purpose of this article is to present a state-ofthe-art review in the area of IPPS by synthesis the information available in literature The various approaches for IPPS have been discussed with their advantages and disadvantages and reported research is classified accordingly It also identifies potential future research directions Thus, this article not only provides a platform to novice researchers but also assist in stimulating further research in the area of IPPS This article is organised with the following sections Section presents the various approaches to integration along with related contributions Section 2.1 discusses nonlinear approach (NLA) Section 2.2 presents closed loop approach (CLA) Section 2.3 deals with distributed approach (DA) Section 2.4 presents the other approaches for IPPS that are followed by researchers Section presents the conclusion and shows some potential future research directions Literature review The best way for IPPS is to merge both process planning and scheduling functions into one However, as process planning and scheduling individually are non-polynomial (NP)-hard, the resulting problem is also NP-hard (Khoshnevis and Chen 1990) Moreover, process planning and scheduling department in a company have to be completely dismantled and reorganised Thus, it cannot be implemented in a company with existing process planning and scheduling departments Tan and Khoshnevis (2000) attempted in this direction with a limited success Another way of IPPS is to increase information exchange between process planning and scheduling functions Several classification schemes are suggested by various researchers following this approach (Zhang and Merchant 1993, Huang et al 1995, Gaalman et al 1999, Gindy et al 1999, Zhang et al 2003a, Shen et al 2006, Baykasoglu and O¨zbakır 2009, Guo et al 2009, Wang et al 2009, Li et al 2010b) However, present work follows the most commonly used classification among researchers (Larsen and Alting 1990, Zhang and Merchant 1993, Huang et al 1995, Gaalman et al 1999, Gindy et al 1999, Baykasoglu and O¨zbakır 2009) Accordingly, there are three main approaches of integration viz., NLA, CLA and DA These IPPS approaches and related contributions are discussed below 2.1 Non-linear approach Here, multiple process plans (MPP) for each part before it enters to shop floor are created by considering operation flexibility (possibility of performing an operation on more than a machine), sequencing flexibility (possibility of interchanging the sequence in which required manufacturing operations are performed) and processing flexibility (possibility of producing the same manufacturing feature with alternative operations or sequence of operations) (Benjaafar and Ramakrishnan 1996) The underlying assumption is that all problems that can be solved ahead of time should be solved before the manufacturing starts Thus, NLA is based on static shop floor situations (Zhang and Merchant 1993, Gaalman et al 1999) All these possible process plans are ranked according to process planning criterion (such as total machining time and total production time) and stored in a process planning database The first priority plan is always ready for submission when the job is required and then scheduling makes the real decision If the first priority plan does not fit well in the current status of shop floor, the second priority plan is provided to scheduling This procedure is repeated until a suitable plan is identified from already generated process plans The criteria for decisions are due dates and batch size of order, capacity of workshop and optimisation criterion for schedule (throughput, lead time, etc.) Figure shows the NLA NLA has one-way of information flow, i.e from process planning to production planning, and thus, it may be impossible to achieve full optimal results in integrating the two functions (Kempenaers et al 1996) Moreover, modern production systems maintain MPP (Kim and Egbelu 1999), and it seems to be a proper Figure NLA (Zhang and Merchant 1993) International Journal of Computer Integrated Manufacturing means to realise the integration between process planning and scheduling (Kempenaers et al 1996) Also, it can be implemented in a company with existing process planning and scheduling department When there are large numbers of parts, the number of process plans tends to increase exponentially and can cause a storage problem (Usher 2003) Also, some of the process plans created are not feasible according to realtime shop status and considering all possible process alternatives for resource allocation may enormously increase the complexity of process plan representation (Zhang and Merchant 1993, Huang et al 1995) Chryssolouris and Chan (1985) proposed manufacturing decision-making approach (MADEMA), the first approach available in literature for IPPS It considered a set of alternative resources for execution of a particular production task Decision matrix was formed for the selection of alternatives, where row represents alternative while column represents attribute and entry was value of attribute for corresponding alternative MADEMA concept contained five consecutive steps: (i) determine alternatives, (ii) determine attributes, (iii) determine consequences with respect to attributes for each alternative, (iv) apply decision rules for choosing the best alternative and (v) select the best alternative The alternative resources were chosen by evaluating the contribution on some decision making established criterion such as a linear combination of attributes with weights or alternatives with greater chance to produce higher utility value Sundaram and Fu (1988) developed a scheduling method through outcome of process planning for minimisation of makespan and to balance loads for machines in job shop environment For schedule improvement, authors used an automated system based on group technology (GT) called integrated computer-aided process planning and scheduling They used a group scheduling algorithm called key machine loading and combined it with process planning generator and operation planner A key machine was loaded with jobs such that it was kept busy continuously Tonshoff et al (1989) presented FlexPlan for IPPS Authors created all MPP before manufacturing starts The scheduling function selected the suitable process plan according to the availability of resources This approach covers reactive replanning to allow reaction of disturbances occurring on shop floor Srihari and Greene (1990) proposed a prototype computer-aided process planning (CAPP) for a Flexible Manufacturing System (FMS) to integrate scheduling function GT coding system was used to input information of parts parameters Heuristic knowledge was used to decide sequence of operations and route through system, on basis of the flow time of jobs 519 A dynamic shop status module of prototype CAPP system maintained dynamic shop status overtime Every alternative route was evaluated with respect to minimisation of flow time of jobs The queues at every machine were modelled in pseudofacility that monitored in terms of time units Prismatic and rotational parts were planned and tested with proposed approach Authors concluded that proposed CAPP system for an FMS with multiple machining centres maintained dynamic shop floor conditions in order to decide the actual sequence of operations and the final job route Jablonski et al (1990) proposed a flexibly integrated production planning and scheduling system having three modules First was an automated feature recognition module in order to identify geometric features of a part and generate a production-oriented part representation in term of basic manufacturing operations Second was a process planning module (PPM) in order to generate all possible resources combinations for production of the part Third was a scheduling and dispatching module, which select the best resources combination to produce the part according to some user defined strategy such as manufacturing operations/features and resources on the shop floor Authors showed that flexible and reactive scheduling approach on feature-based process planning was feasible Palmer (1996) proposed a simulated annealing (SA) based approach for IPPS It contained three types of configuration alterations; (i) reverse the order of two sequential operations on a machine, (ii) reverse the order of two sequential operations within a job and (iii) change the method used to perform an operation The cost functions such as, tardiness, mean flow time, makespan and a combined function of mean flow time and tardiness were considered The performance of SA and dispatching rules were compared Authors concluded that solution quality of SA was remaining high across varying situations, and it was effective means for IPPS Also, it outperformed the use of dispatching rules Kim and Egbelu (1999) proposed a mathematical approach to develop a scheduling tool for multiple jobs with each having MPP in a job shop environment Authors claimed that the proposed methodology minimise throughput/makespan for part mix It contained two sub-systems viz., process plan selection subsystem (PPSS) and shop scheduling subsystem (SSS) PPSS selected a set of process plans for each part type to be scheduled The selected set of process plans was passed to SSS to generate a feasible schedule Performance measures determined by scheduling system were passed back to process planning system to modify its process plan selection This 520 R.K Phanden et al iterative process continues until no further improvement in schedule was identified The scheduling problem was solved with two algorithms viz., preprocessing algorithm and heuristic/iterative algorithm ‘Pre-processing algorithm’ combined features of branch and bound and integer programming techniques while heuristic/iterative algorithm combined features of branch and bound technique and earliest completion time dispatching rules They concluded that computational time of pre-processing algorithm was substantially lower than that of mix integer programming technique but higher than that of heuristic model As number of jobs increases, solution quality obtained by heuristic got worse, but as the number of machine increases, it had no clear effect on performance of heuristics The increase in number of process plans per job had a negative effect on the solution quality of the heuristic Aldakhilallah and Ramesh (1999) proposed an architecture and framework called computer-integrated process planning and scheduling (CIPPS) which consists of three modules viz., super relation graph, cover set model and cover set planning and scheduling module (CSPS) in order to recognise polyhedral depression features and extract prismatic features from CAD database using artificial neural network (ANN) and computational geometry techniques Moreover, CIPPS framework contained three modes of operations viz., dynamic support for design decision (DSDD), runtime intelligent operational control (IOC) and data consolidation and integration (DCI) DSDD mode supports decisions during design process IOC mode worked up for automatic shop floor management, when changes occurred in the environment DCI mode performed the interfacing and integration of CIPPS with other functions in manufacturing environment Weintraub et al (1999) proposed a procedure for scheduling jobs, to minimise manufacturing cost while satisfying due dates by taking into account alternative process plans of jobs, in a large scale manufacturing job shop An iterative simulation-based scheduling algorithm was developed to minimise lateness and applied in virtual factory, which was a Windows-based software package In order to further reduce lateness, a Tabu search (TS)-based algorithm was applied to identify process plans with alternative operations and routings The numbers of alternative process plans were fixed for the job at two Process plans with alternative routings, operations and sequences were selected according to the current shop status They concluded that scheduling with alternatives can greatly improve the ability to satisfy due dates under varying shop status Also, scheduling with alternative operations had largest schedule improvement and schedule with alternative sequence had smallest schedule improvement Saygin and Kilic (1999) proposed a framework to integrate MPP with predictive (off-line) scheduling in an FMS, in order to minimise completion time The framework worked up in four stages: (i) machine tool selection, (ii) process plan selection, (iii) scheduling and (iv) re-scheduling module Dissimilarity maximisation method (DMM) was used for selection of appropriate process plans of a part mix Rescheduling strategy was developed in order to reduce waiting time of parts and algorithms were based on mathematical and heuristic approaches They concluded that idle time of machine tool was reduced to 30 units from 81units and waiting time of parts was dropped to 50 units from 74 units Also, optimal process plan that might have shortest processing time (SPT) or least number of operations may not guarantee best system performance Lee and Kim (2001) proposed a method for IPPS, using simulation based on genetic algorithm (GA) Simulation module computes performance measures based on process plans combination created by GA instead of process plan alternatives and output the near-optimal process plan combination prior to execution on shop floor The performance measures were makespan and lateness based on SPT and earliest due date (EDD) dispatching rules They concluded that about 20% reduction of makespan was possible when compared with random selection of process plan combination Yang et al (2001) proposed a feature-based multiple-alternative process planning system with scheduling verification The process plan was generated directly from part design and available resource data information The system had four components viz., a relational manufacturing database, form feature recognition, process alternative generation and scheduling state evaluation A 3D model and blank raw material model was entered by using initial graphics exchange specification data format The manufacturing features were decomposed by using graph-based and rule-based algorithm After generating MPP, each plan was allocated to scheduling, and a candidate process plan was retrieved on the basis of required due date Authors concluded that proposed prototype contained choice of process sequence with verification of delivery time for all feasible set of process sequences Moon et al (2002) proposed a GA-based IPPS model for multi-plant supply chain A mathematical model was formulated with consideration of alternative machines and sequences, sequences dependent setup and due dates to minimise tardiness Operations sequencing problem was formulated as a multiple travelling salesman problem (TSP’s) and each TSP International Journal of Computer Integrated Manufacturing determine machine operation sequences for each part type A topological short technique (TST) was used to obtain all flexible sequences in directed graph Authors concluded that proposed GA approach was more efficient than TS approach in terms of computational time and problem size Also, population size and number of generations were main factors that affect performance of the proposed approach Kim et al (2003) proposed an artificial intelligence (AI) search technique called symbiotic evolutionary algorithm (SEA) to simultaneously deal with process planning and job shop scheduling in FMS SAE was based on the fact that parallel searches for different pieces of solution were more efficient than a single search for the entire solution They considered operation flexibility, sequencing flexibility and processing flexibility during process planning The job-shop scheduling determines both process plan for each job and corresponding schedule, while optimising two types of objectives: minimising makespan and minimising mean flow time SEA was tested on 24 test-bed problem set and found better outcomes than existing cooperative co-evolutionary GA (CCGA) (Potter 1997) as well as hierarchical approach Kumar and Rajotia (2003) suggested a method for on-line scheduling in a CAPP system for a job shop environment A scheduling factor was used to make operation-machine assignment The operations were assigned to machines with highest value of actual scheduling factor The scheduling criterions were flow time and number of tardy jobs They concluded that the proposed method helps in on-line determination and assignment of loads on various machines Further, Kumar and Rajotia (2006) proposed a framework for IPPS system in job shop environment It considered machine capacity and cost while assigning operation to machines The proposed framework contained two controlling modules viz., process plan generator and scheduler Both modules were interacting with a decision support system (DSS) DSS interacts with various databases such as machine tool database, toolwork material database and machining parameters database A generative scheme was used to develop process plan for axis-symmetric components The scheduling factor as reported in earlier authors work (Kumar and Rajotia 2003) was used to assign setup to respective machine tool Zhao et al (2004) proposed a GA-based approach for IPPS in a job shop environment A fuzzy inference system was used to select alternative machines It was based on fuzzy logic toolbox by MATLAB Based on the capability of machines, GA was used to balance load for all machines Gliffer and Thompson’s algorithm (Gliffer and Thompson 1960) was used to evaluate fitness of chromosome in schedule builder 521 The scheduling objectives were to minimise makespan, minimise number of rejects and minimise processing cost Zhao et al (2006) extended their earlier work and used particle swarm optimisation (PSO) algorithm for balancing load on each machine Moreover, Zhao et al (2010) proposed an IPPS applicable to Holonic Manufacturing System (HMS) in which they used a hybrid PSO and differential evolution algorithm in order to balance the load for all machines Grabowik et al (2005) proposed an integration methodology utilising MPP of a product in order to respond in disturbances during manufacturing The proposed methodology represented a predictive-reactive and event-driven approach to rescheduling They concluded that the availability of processes routes expanded flexibility of control system and increases efficiency of rescheduling Choi and Park (2006) proposed a GA-based method for IPPS, that minimise makespan of each job order, considering alternative machines and alternative operations sequences in integrated manufacturing environment An operation-based representation was used to construct chromosomes The performance of proposed method was evaluated in a job shop environment evolving MPP Authors concluded that the proposed approach shows the possibility of improving makespan Jain et al (2006) proposed an integration scheme that can take advantage of flexibility on the shop floor and can be implemented in a company with existing process planning and scheduling departments The proposed methodology was able to take advantage of MPP, while following a real-time strategy for scheduling suitable for changing workshop status The proposed system was composed of two basic modules: process plans selection module (PPSM) and scheduling module (SM) PPSM selects best four process plans for each part type and stores them in a database SM performs part scheduling for using best four process plans The effectiveness of MPP over single process plan was assessed through makespan and mean flow time Authors concluded that the availability of MPP during FMS scheduling improves makespan and mean flow time Li and McMahon (2007) proposed a SA-based approach for IPPS in a job shop environment Processing, operation sequencing and scheduling flexibility were used to explore search space of proposed algorithm The algorithm was defined in two sets of data structures The first set represents process plans and the second set specifies the schedule of a group of parts The performance measures were makespan, balanced level of machine utilisation, job tardiness and manufacturing cost The proposed algorithm was compared with GA, TS and PSO algorithms The authors concluded that the proposed algorithm performed satisfactorily and was 522 R.K Phanden et al able to choose one or more specific performance criterion to address practical requirements Moon et al (2008) proposed an evolutionary search method based on TST for IPPS in supply chain A mixed integer programming model was formulated, which incorporate process planning of resources selection and sequence of operation as well as determination of their schedule to optimise makespan They conducted three experiments with considerations of various orders, operations and resource sizes and concluded that proposed approach was capable of producing optimal schedule and robust in generating the best makespan through varied genetic environment and under various order environments with precedence constraints Li et al (2008b) proposed a GA-based approach to facilitate IPPS They developed an efficient genetic representation and operator scheme The first part of chromosomes composed of alternative process plan string and second part composed of scheduling plan string They assumed job shop environment to minimise makespan They found that the value of makespan without integration was worse than proposed integration model Li et al (2010c) proposed a hybrid approach, which synthesises advantage of GA and TS to solve IPPS problem The first part of chromosome was alternative process plan string, second part was scheduling plan string and third was machine string Third part selects machine set of corresponding operations of all jobs to minimise makespan Authors concluded that the proposed algorithm was effective and acceptable for IPPS problem Wang et al (2008) proposed an IPPS approach in a batch manufacturing environment by utilising process plan solution space A heuristic was developed to minimise tardiness and also, to maintain cost of process plan involved in modification of process plan An SA algorithm was used to find optimal process plan for prismatic parts only They concluded that tardiness of jobs was improved, while the cost of process plan was maintained at low level, because PPM optimises the route with minimum processing cost Haddadzade et al (2009) proposed an approach for IPPS, in a job shop for prismatic components that can be implemented in a company with existing departments The model consisted of PPM and SM PPM generates all possible alternative plans, then SM ranked these based on minimum cost while due date was considered The proposed approach took advantage of MPP to fulfil due dates using overtime It can optimise cutting parameters for milling operations only Authors concluded that the proposed approach can determine machining parameter, tool, machine and amount of overtime within minimum cost objective and due date Baykasoglu and O¨zbakır (2009) proposed an IPPS model that comprises of two parts First part was a generic process plan (GPP) generator to generate final process plan Second part was dispatching rule based heuristic to generate feasible schedules A multiple objective tabu search algorithm was employed to find an optimal schedule for two objectives that were ‘flow time’ and ‘cost of process plan’ They concluded that process plan cost decreases as process plan flexibility increases Rajkumar et al (2010) proposed a multi-objective greedy randomised adaptive search procedures (GRASP) in order to minimise makespan, maximum workload, total workload, tardiness and total flow time evolving flexible job shop environment It consists of two phases viz., construction phase and local search phase Authors focussed on construction phase through computational experiments to solve IPPS problem The IPPS framework consisted of PPSM and SM The proposed algorithm was validated with four benchmarking problems Authors concluded that the proposed GRASP was effective to solve IPPS problem Leung et al (2010) proposed an IPPS approach utilising ant colony optimisation (ACO) algorithm based on multi-agents system (MAS) in order to minimise makespan evolving job shop environment They considered processing flexibility of alternative routing and alternative machines AND/OR graphs were used to represent MPP Authors concluded that the proposed agent-based ACO approach was feasible to solve IPPS problem 2.2 Closed loop approach Here, process plans are generated by means of a dynamic feedback from production scheduling and available resources Production scheduling tells process planning regarding availability of different machines on shop floor for the coming job, so that every plan is feasible with respect to current availability of production facilities Every time an operation is completed on shop floor, a feature-based work piece description is studied in order to determine next operation and allocate the resources This approach takes dynamic behaviour of the manufacturing system into consideration Thus, real-time status is crucial for CLA (Zhang and Merchant 1993) It is also referred to as real-time approach or dynamic approach Figure shows CLA In order to take full advantage of CLA, process planning and scheduling departments in a company may have to be dismantled and reorganised (Iwata and Fukuda 1989) Moreover, it requires high-capacity International Journal of Computer Integrated Manufacturing Figure CLA (Zhang and Merchant 1993) software and hardware (Zhang and Merchant 1993) and adaptation of step-by-step local view that limits the solution space for subsequent operations (Gaalman et al 1999) However, this approach is unrealistic as the complexity of manufacturing processes might be unavoidable in achieving real-time process plan generation (Joo et al 2001) Dong et al (1992) proposed a dynamic featuresbased IPPS The product features were extracted using an AI-based feature extractor with respect to shop floor conditions Then, rough process plan for a product was prepared It considers all possible manufacturing ways for each operations volume (operation features) that can be produced in one machine setup considering shop floor capabilities Moreover, during rough process plan generation, geometric constraints decide the priority of manufacturing for each operation volume Rough process plan with alternative was input to scheduling Smallest slack time criterion was considered for scheduling of a batch size manufacturing shop Concurrent Manufacturing Planning and shop control for batch production (COMPLAN), a European ESPRIT project 6805 during the period 1992– 1995, integrates process planning and workshop scheduling using MPP The COMPLAN approach was an extension of FlexPlan (Kruth and Detand 1992) The goal of COMPLAN project was to develop a software system prototype that was capable of carrying out manual and automatic process planning and scheduling based on MPP, in a small batch manufacturing of complex products in a job shop It contained PPM and workshop scheduling system (WSS) PPM was capable of handling MPP, and it could use projected resources load while developing MPP This module described feasible manufacturing alternative that provided flexibility to workshop scheduling WSS followed a hierarchical approach Usher and Fernandes (1996) proposed a process planning architecture for integration with scheduling system It used feature-based approach to planning and has two phases namely ‘Static Planning’ and ‘Dynamic Planning’ ‘Static Planning’ phase involved selection, assignment and sequencing of processes and machines that exist within the shop The output of ‘Static Planning’ phase was a set of alternative macrolevel plans ‘Dynamic Planning’ phase considered availability of shop floor resources and objectives specified by scheduler The proposed system was able 523 to perform for both prismatic and rotational parts Authors concluded that the proposed two-phased approach was able to reduce work load when realtime portion of planning activities were carried out in second phase Cho et al (1998) proposed a prototype Block Assembly Process Planning and Scheduling system in shipbuilding, which consists of a block assembly PPM, a SM, a bottleneck block selection module and a process-replanning module Rule-based reasoning technology was applied to determine optimal assembly units and assembly sequences in generating initial process plans For SM, a schedule revision heuristic was developed for efficient reallocation of blocks to alternative assembly shops For bottlenecks, block selection that plays a central role in bridging process planning and scheduling, a heuristic was developed by employing an entropy-based partitioning method to identified bottleneck periods Thus, initial process plan was repeatedly modified and improved by iterating ‘scheduling’ – ‘bottleneck block selection’ – ‘process replanning’ cycle until workloads were sufficiently balanced Sugimura et al (2001) proposed an IPPS system applicable to HMS The process planning system selected suitable sequence of machining feature using GA approach The optimum sequence of machining equipment was selected using dynamic programming (DP) after taking into consideration the future schedule of machining equipments, with objectives to minimise total machining and set-up time Machining schedules were determined using a real-time scheduling procedure in which individual job selected suitable machining equipment, in order to carry out next operation based on process plan Sugimura et al (2003) extended their earlier work (Sugimura et al 2001) Here, optimum sequence of machining equipment, two objective functions viz., minimisation of shop time and machining cost were used in DP approach Shop time was combination of machining time, fixturing time, tool changing time, transportation time and waiting time The machining time was calculated on the basis of manufacturing process time and operation cost per unit time of manufacturing resources Authors concluded that the proposed approach was capable to select both suitable sequence of machining equipment and machining schedule concurrently Shrestha et al (2008) developed an IPPS system for HMS using DP method-based modification of process plans Two types of holons viz., job holons and scheduling holons were considered for process planning and scheduling, respectively Based on feasible process plans of all jobs, the scheduling holon generates the schedule of all equipments Makespan, total machining cost and weighted tardiness cost were 524 R.K Phanden et al assumed as an objective function for scheduling A GA-based method was adopted for selecting a combination of process plans Schedules were generated using a set of dispatching rules viz., SPT, SPT/ Total Work Remaining and Apparent Tardiness Cost (ATC) Feedback processes were considered in order to modify the process plans based on load balancing of machining equipments Two approaches viz., centralised approach and DA were developed in order to modify process plans In centralised approach, the feedback information from scheduling holon after scheduling was transferred to job holons and one modified set of process plans obtained with consideration of constraints of machining equipments In DA, the job holons modify their process plans without any centralised control of scheduling holons Results were compared with and without modification approaches Authors concluded that in centralised approach of process plan modification, the makespan was improved for the case where there was concentration of the machining load on some machining equipments Also, in the DA, the makespan and weighted tardiness cost were improved for both cases where there is and there was no concentration of machining loads on the machining equipment Zhang et al (2003a) proposed an IPPS scheme for batch manufacturing of prismatic parts This approach is similar to COMPLAN except, the process planning system could generate whole solution space based on operations for a given part using SA and GA in order to find optimal plan (Kempenaers et al 1996) An ‘intelligent facilitator’ was used to generate instructions for process plan modifications The integration was achieved through an ‘intelligent facilitator’ that provide feedback to PPM of a particular job They developed two algorithms viz., ‘machine utilisation’ and ‘tardy job’ based on heuristic and concluded that algorithms based on heuristic were suitable to achieve a satisfactory schedule Wang et al (2008) have extended the work of Zhang et al (2003a) They proposed two heuristic algorithms namely fine-tuning (FH-Tardy) and quick-tuning (QH-Tardy) In FHtardy algorithm, solution space of selected operation of selected tardy job was modified in each iteration In QH-Tardy algorithm, solution space of selected operation was modified in each iteration for each tardy job They concluded that the proposed heuristic was able to explore process plan solution space in order to reduce job tardiness Wong et al (2006a) proposed an agents-based multi-stage negotiation protocol scheme for IPPS in a job shop environment The proposed system comprised of part agents and machine agents to represent parts and machines, respectively Part and machine agents negotiate to establish schedule using process plans and operation details from AND/OR graph The negotiation protocol was established to handle multiple task and many-to-many negotiation A currency conversion function, which incorporates processing time and due date, was adopted for bidding A javabased simulation model multi-agents negotiation (MAN) was used to implement the proposed approach Authors concluded that in the pursuit of local objectives such as parts flow time, the proposed approach performs better than meta-heuristics Wong et al (2006b) extended their earlier work (Wong et al 2006a) and proposed a hybrid-based multi-agent system called Online Hybrid Agents-based Negotiation (oHAN) It comprised of local agents (part and machine agents) and a supervisor agent The supervisor agent was used for global rescheduling process and influenced the decisions made by local agents It acted as a system coordinator manager in between local agents for global rescheduling objective Authors concluded that hybrid approach is effective for larger scale rescheduling and provided a better global performance They suggested a mobile agent in oHAN in order to handle job order details 2.3 Distributed approach It performs both process planning and production scheduling simultaneously It divides process planning and production scheduling tasks into two phases The first phase is preplanning In this phase, process planning function analyses the job based on the product data The features and feature relationships are recognised, and corresponding manufacturing processes are determined The required machine capabilities are also estimated The second phase is the final planning, which matches required job operations with the operation capabilities of available production resource The integration occurs at the point when resources are available and the job was required The result is dynamic process planning and production scheduling constrained by real-time events This approach is also referred to as just-in-time approach or phased or progressive approach Figure shows DA This approach is the only one that integrates the technical and capacity-related planning tasks into a dynamic fabrication planning system (Larsen and Alting 1990) However, this approach requires high capacity and capability from both hardware and software Moreover, scope of DA is limited within some specific CAPP functions such as process and machine selection as detailed process planning tasks are shifted down to manufacturing stages for enhancing flexibility (Joo et al 2001) From implementation viewpoint, both process planning and scheduling International Journal of Computer Integrated Manufacturing Figure 525 DA (Zhang and Merchant 1993) departments in a company have to be dismantled and reorganised (Haddadzade et al 2009) Aanen et al (1989) proposed a hierarchical approach for IPPS in a FMS The components of FMS were integrated with software called supervisory control system The primary objective was to satisfy due dates of the order and secondary was to minimise change over and idle times of machines within time horizon Initially, planning function was solved and resulting output becomes the input for scheduling Feedback information was provided to planning level, if output of scheduling was not satisfied Authors planned and scheduled two types of activities viz., machining activities and operator activities The time horizon (of about 10 days) at the planning level was divided in periods of day For each day, machining activities to be performed were assigned The resulting list was called a day list List of activities for 1st day was the input to scheduling level Scheduling criterion was to minimise makespan of the day list The scheduling was performed in two steps: first, scheduling of machining activities (using branch-and-bound method) and second, scheduling of operators activities Zhang and Merchant (1993) proposed an integrated process planning model The modules were integrated in three levels viz., pre-planning module at initial integration level, pairing planning module at decisionmaking level and final-planning module at functional integration level Pre-planning module performed three activities viz., feature reasoning, process recognition and setup determination The output of pre-planning was possible setups, machining operation and associate times Pairing planning module worked in three steps: machining selection, tooling and fixture selection and exact time selection Final planning level worked in three steps viz., operational tolerance analysis, operation sequencing and overall time and cost calculations In pre-planning level, SM provides available equipment in next time window In decision-making level, the available equipments were matched with requirement of setup and matching processes Decision-making module (DMM) was central element to perform by means of real-time information Real-time machine database constructed base in the task and time assignment of the machine that contained information about all machines in shop floor Huang et al (1995) proposed a progressive approach containing three phases namely pre-planning, pairing planning and final planning The activities within each phase takes place in different time periods Pre-planning was executed in early stage, as soon as product design finished Pairing planning executed in later stage, when an order has been released to the shop and final planning was executed just before the manufacturing begins The interaction between process planning and scheduling takes place in all three phases The model consisted of PPM and SM PPM was responsible for generating process plans according to part design specifications The criterion for process plan selection was manufacturing lead time SM was responsible for allocating resources in the shop and overall management of flow of production orders They solved IPPS problem by developing first mathematical models and then using optimisation < 0:001 j ¼ 1; ; mp ð10Þ where W{ is the left pseudo-inverse of W, namely W{ : (WTW)71WT which is equivalent to specifying that each parameter should be refined to 0.1% precision, that is, DbðsÞ ¼ Wy DpðsÞ bj Àbj International Journal of Computer Integrated Manufacturing ðsÞ ð1Þ ðsÞ ð0Þ ð1Þ ð0Þ 567 jbj À bj j < 0:001 Á jbj À bj j % 0:001 Á jbj À bj j; where bð1Þ is the asymptotic value of the parameter that after calibration (Figure 10, bottom right) the workpiece is better finished 3.3 Calibration results The calibration procedure described in this work has been run on an Intel1 Core Duo PC with Matlab1 2007c A total of 108 points have been measured, that is, a regular mesh of 36 grid points (6 6) was commanded for each plane in Figure Figure shows the behaviour of the proposed calibration algorithm for the studied workcell The calibration process shows a good convergence The stop criterion given by Equation (10) is met at the 18th iteration and the value achieved for the parameters is b(18) ¼ b(0) þ [0.05 0.01 0.06 0.01 0.07 0.08]T (mm, rad) To validate the improvement obtained with the calibration procedure, two comparisons are considered For the first, a total of nv ¼ 96 points on the table plane were measured using the laser sensor and varying external joints randomly from to 1000 mm for dL and from to p rad for yM Figure shows the performance for this test with and without calibration After calibration the accuracy of the data clearly improves: the average value is reduced from 73.342 to 70.600 mm; the root mean square (RMS) error is reduced from 5.429 mm to 0.998 mm; and the standard error deviation is reduced from 4.279 to 0.797 mm The second comparison consists of the milling of a small workpiece, see Figure 10 Although the motion of the additional joints is not required for this milling task, the operator manually displaces (by means of the control panel of the robot) the linear track 500 mm and rotates the table p/4 rad at a certain time while the lateral walls of the workpiece are being milled Note Inverse kinematic control of the redundant workcell The inverse kinematic problem (IKP) is relevant because CAM systems specify the toolpath at the Cartesian workspace O, whereas the robot controller works in the joint space = The IKP can be solved for each pose of the path, although at the displacement level the IKP is cumbersome because an infinite number of solutions exist for redundant manipulators Instead, an iterative approach at the joint rate level is usually used (Whitney 1969) The forward kinematics between the EE velocity t and the joint velocity q_ is represented by the linear _ where J is the m6n Jacobian matrix, equationt ¼ J Á q, which is a non-linear function of the joint angles that can be computed with the geometric procedure outlined by Angeles (2003), among others For the redundant workcell considered in this work, the Jacobian matrix maps the 9-dimensional vector q_ ¼ ½y_ M ; d_L ; y_ ; ; y_ ŠT of joint rates into the EE velocity vector t ¼ [o v]T, with o ¼ [ox oy oz]T and v ¼ [vx vy vz]T denoting the linear and angular velocities of the EE reference frame relative to the base frame {B}, respectively The objective of the IKP at the joint rate level is to obtain the required joint velocities q_ to achieve the desired EE velocity t for a given Jacobian matrix J In the case of redundant robots, for which the Jacobian matrix J is non-square (n m), the right pseudo-inverse J{ : JT(JJT)71 gives the minimum least-squares solution q_ that fulfils the desired EE velocity t, namely, q_ ¼ Jy Á t Moreover, an j ðsÞ Figure Left, the value bj of each parameter tends to an asymptotic value; right, the calibration algorithm meets the stop criterion at the 18th iteration 568 Figure Figure 10 J Andres et al Validation of the calibration: scattergraph and histogram of the distance measurements DP to the table surface Milling of a small workpiece: bottom left, result without calibration; bottom right, result with calibration arbitrary vector from the Null Space of J, namely @ðJÞ, can be added to achieve a secondary goal: q_ ¼ Jy Á t þ ðI À Jy JÞh ð11Þ where I is the identity matrix of dimension n, (I J{J) is the projection operator on @ðJÞ, and h is the performance vector for the secondary task At this point it is worth mentioning that, as the Jacobian matrix depends on the DH parameters, the workcell calibration is crucial for a proper robot control In the RRS given by Equation (11), the performance vector h can be considered as a virtual force that attempts to push the configuration of the manipulator away from a critical area in = (Nemec and Zlajpah 2008) as a secondary task The most widespread method used to select h is the gradient projection method (GPM) (Lie´geois 1977), which minimises a position-dependent scalar, the performance index p, by means of its gradient vector:   @pðqÞ @pðqÞ @pðqÞ T h ¼ Àkrp; with rp ¼ ; ; ; @q1 @q2 @qn ð12Þ In this work, two performance criteria are combined into vector h: h ¼ hnt þ hcond ¼ Àrðpjnt þ pcond Þ ¼ ÀðHjnt ðq À qref Þ þ Hcond kðq À qTs ÞÞ ð13Þ where H is the weighting matrix for each criterion, for example, constant diagonal matrices are considered in this work Hjnt ¼ Hcond : diag(0.01) Note that the International Journal of Computer Integrated Manufacturing performance vector hjnt given by performance index pjnt tends to maintain the workcell as close as possible to a reference posture qref (e.g the HOME posture q0 shown in Figure 2) to avoid joint limits (Huo and Baron 2008) The performance vector hcond given by performance index pcond tends to maintain the workcell as far as possible from ill-conditioned postures (Angeles and Lo´pez-Caju´n 1992), that is, postures with poor kinematic performances The second performance vector hcond is activated when the value of the condition number k of the Jacobian matrix (e.g obtained using the Euclidean norm) passes over a preset threshold value, for example, 0.5 At that instant, the corresponding configuration qTs is recorded to evaluate at = the distance to the actual posture Post-processor implementation for a redundant workcell 5.1 CAM system for toolpath generation The software NX1 is a digital development system from SiemensTM that integrates the tasks of design (CAD), simulation (CAE) and planning of milling tasks (CAM) NX interacts with the programme codes in TCL and Cþþ that manipulate the path data (event handler) and gives a convenient format to the generated output (definition file), see Siemens Corp (2009) and Figure 11 The treatment of the data has been programmed in Matlab1 code with the aid of the Hemero toolbox (Maza and Ollero 2001), whereas the Robomove1 software (Qdesign S.R.L 2007) has been used to visualise the robot postures resulting from the implemented RSS 5.2 Continuous path tracking Once the CAM system has generated the toolpath as a discrete set of close-enough poses at O, the EE of the robot must track this path A tangent, normal and 569 binormal unit Frenet-Serret vectors, that is, {t, n, b}, is associated with every sample point of the trajectory and this indicates the required pose, see Figure 12 These trajectory data are stored as TCAM The joint angles of the robot have to be computed along this continuous set of poses of the EE The IKP can be solved at each sampled pose, although the resolution is cumbersome because an infinite number of solutions exist for redundant manipulators Instead, an iterative approach can be applied by using J (Angeles 2003):   Qk vectðQTk Qd Þ Jðqi ÞDqi ¼ Dti ; with Dti  ð14Þ Dp where Qk (i.e the first three rows and columns of the matrix T introduced in Section 2) is the current rotation matrix from {B} to {EE} (see Figure 1), and Qd is the desired rotation matrix Both matrices are related by: Qd ¼ Qk Á DQ ! DQ ¼ QTk Á Qd ð15Þ Function vect(DQ) represents the axial vector of the rotation matrix DQ and is computed as: DQ32 À DQ23 14 vectðDQÞ  ð16Þ DQ13 À DQ31 DQ21 À DQ12 Vector Dp is defined as the difference between the prescribed value pd of the TCP position vector and its current value pk The relations among Qd, pd, Qk, pk, DQ and Dp are shown in Figure 12 Algorithm below is programmed with these premises It starts from the HOME posture (q0, Figure 12) and iteratively uses the kinematic models of the single KR15/2 manipulator (sub-index 6R) and the complete workcell (sub-index Workcell) to Figure 11 Integrated post-processing in NX1: the ‘definition file’ and the ‘event handler’ adapt the toolpath to the controller’s Kuka Robot Language (KRL) 570 J Andres et al Figure 12 Highlight of the loop leading from an initial current pose (k) to a desired final pose (k þ 1), both specified by the Frenet-Serret vectors {t, n, b} evaluate the condition number k and manage the motion of all the joints Algorithm Algorithm J0 (transposr of J in Matlab) (1) JT (2) [Q R] ¼ qr(JT) (Matlab’s orthogonal-triangular QR-decomposition) (2.1) H Q0 (2.2) U R(1:6,:) (3) r Dt-JÁ h (4) y1 forward (U0 r) (UTÁy1 ¼ r, solved by forward substitution) (5) y [y1; zeros(3,1)] (6) k H0 Á y (7) Dq ¼ k þ h (1) q q0 for (each i-point of the trajectory, TCAM(i)) (2) {pd, Qd} TCAM while jjDqjj e (3) fp; Qg ( DKðq; DH À workcell Þ (4) DQ ( QT Á Qd (5) Dp ( pd À p   Q Á vectðDQÞ (6) Dt Dp (7) Jworkcell ( DKðq; DH-WorkcellÞ (8) Determination of kF (8.1) q6R {0, q(4), .,q(8)} (8.2) J6R ðqÞ ( DKðq6R ; DH À KR15=2Þ (8.3) H6R Jg 6R L (8.4) kF H6R (9) Dq ( RRS (10) q ( q þ Dq endwhile endfor The RRS in the ninth step of Algorithm is programmed using the Householder reflections Direct calculation of the pseudo-inverse of the Jacobian J{ is avoided to keep the round-off errors as low as possible (Arenson 1998) Algorithm below summarises this calculation using Matlab1 code Application The production system described above was implemented to mill an 13 m orographic model of the tail end of a reservoir on the river Mijares (Spain) The design is imported from AUTOCAD1 (widely used in topography) to the NX1-CAD module to restore the contour lines and interpolate a surface mesh The quality of the resulting CAD file determines the efficiency of the results obtained in the subsequent steps of the milling process As shown in Figure 13, the model is made by assembling 120 blocks of 6 0.5 m of EPS The toolpath generated in NX1 has been simulated with and without the implemented post-processor in the graphical Robomove1 simulator and starting from the same HOME posture In Figure 14 it can be seen that International Journal of Computer Integrated Manufacturing Figure 13 The orographic model is produced by assembling 120 blocks of m 10.5 m of EPS Figure 14 Toolpath post-processed without (left) and with (right) the proposed RRS 571 Figure 15 Left, milling process of an EPS block; right, final 1:75 model used for flowing simulation (real dimensions of m 13 m) 572 J Andres et al the end of the workpiece (red part) cannot be reached without moving the additional joints, that is, dL and yM A better performance is achieved when the toolpath is post-processed with the proposed RRS, which moves the additional joints to maintain a wellconditioned posture during the machining process Figure 15 (right) shows the final orographic model at the Hydraulic Engineering Department (DIHMA) of the Universidad Polite´cnica de Valencia, which is used for simulating water courses under various climatic conditions Conclusions After introducing the capabilities of an industrial robotic workcell, two necessary requirements for setting up the workcell have been discussed in this article: the in situ calibration of the robot’s external joints and the management of the redundancies (caused by the additional joints and the symmetry of the cutter tool) For the calibration task, a novel non-contact planar constraint procedure has been developed on the basis of a planar pattern and a laser displacement sensor and has been successfully validated with two illustrative examples The proposed calibration method is relatively cheap, can be implemented autonomously in most industrial robots, and is fast enough to be used in situ at regular intervals Further work must be done to integrate the laser measuring device within the tool holder, in order to bridge the gap between the theoretical and practical applications on the shop-floor Moreover, a functional post-processor has been programmed inside the CAM system for the control of redundancies during milling tasks A real case study has been considered to validate the effectiveness of these production systems for the milling of large prototypes using soft materials The proposed postprocessor is expected to be easily applicable on any industrial robot, and useable for different applications such as welding or painting The implementation details to adapt CAM modules from machining tasks to welding and painting tasks is a field for further research References Angeles, J., 2003 Fundamentals of robotic mechanical systems: theory, methods and algorithms New York: Springer Angeles, J and Lo´pez-Caju´n, C.S., 1992 Kinematic isotropy and the conditioning index of serial robotic manipulators The International Journal of Robotics, 11 (6), 560– 570 Arenson, N., 1998 Real-time redundancy-resolution schemes for robotic manipulators Thesis (PhD) McGill University Bidanda, B., Narayanan, V., and Rubinovitz, J., 1993 Computer-aided-design-based interactive off-line programming of spray-glazing robots International Journal of Computer Integrated Manufacturing, (6), 357–365 Driels, M.R and Swayze, W., 1994 Automated partial pose measurement system for manipulator calibration experiments IEEE Transactions on Robotics and Automation, 10 (4), 430–440 Driels, M.R., Swayze, L.W., and Potter, L.S., 1993 Fullpose calibration of a robot manipulator using a coordinate measuring machine International Journal of Advanced Manufacturing Technology, (l), 34–41 Fassi, I., et al., 2006 Calibration of serial manipulators: theory and applications In: L.K Huat, ed Industrial robotics: programming, simulation and applications Angsburg Germany: Pro Literatur Verlag, 147–170 Feng-Yun, L and Tian-Sheng, L., 2005 Development of a robot system for complex surfaces polishing based on CL data International Journal of Advanced Manufacturing Technology, 26 (9), 1132–1137 Hartenberg, R.S and Denavit, J., 1955 A kinematic notation for lower pair mechanisms based on matrices Journal of Applied Mechanics, 22 (2), 215–221 Huang, H.-K and Lin, G.C.I., 2003 Rapid and flexible prototyping through a dual-robot workcell Robotics and Computer Integrated Manufacturing, 19 (3), 263– 272 Huo, L and Baron, L., 2008 The joint-limits and singularity avoidance in robotic welding Industrial Robot: An International Journal, 35 (5), 456–464 International Standard ISO 9283:1998 Manipulating industrial robots performance criteria and related test methods Khalil, W and Dombre, E., 2002 Modeling identification and control of robots New York: Taylor & Francis Books Khalil, W., Garcia, G., and Delagarde, J.F., 1995 Calibration of the geometric parameters of robots without external sensors In: IEEE International Conference on Robotics and Automation Vol New York, NY: IEEE, 3039–3044 KUKA Corp., 2005 KUKA System software (KSS): expert programming (KRC2/KRC3), Release 5.2 [online] Available from: http://www.kuka-robotics.com/en/ downloads/ [Accessed 18 April 2011] Lee, R.-S and She, C.-H., 1997 Developing a postprocessor for three types of five-axis machine tools International Journal of Advanced Manufacturing Technology, 13 (9), 658–665 Lie´geois, A., 1977 Automatic supervisory control of the configuration and behavior of multibody mechanisms IEEE Transactions on Systems, Man and Cybernetics, (12), 868–871 Lo´pez, De Lacalle, L.N., et al., 2005 The CAM as the centre of gravity of the five-axis high speed milling of complex parts International Journal of Production Research, 43 (10), 1983–1999 Maza, J.I and Ollero, A., 2001 Herramienta MATLABSimulink para la simulacio´n y el control de robots manipuladores y mo´viles, Actas de las XXI Jornadas de Automa´tica, Sevilla Nemec, B and Zlajpah, L., 2008 Robotic cell for custom finishing operations International Journal of Computer Integrated Manufacturing, 21 (1), 33–42 Patel, R.V and Shadpey, F., 2005 Control of redundant robot manipulators: theory and experiments Berlin: Springer International Journal of Computer Integrated Manufacturing Patrick, W., 2005 Shop-floor productivity Manufacturing Engineering [online], 135 (1) Available from: http:// www.sme.org/cgi-bin/find-articles.pl?&ME05ART38&ME& 20050710&&SME& [Accessed 12 August 2010] Prenninger, J.R., Vincze, M., and Gander, H., 1993 Contactless position and orientation measurement of robot end-effector IEEE International Conference on Robotics and Automation, 1, 180–185 Qdesign, S.R.L., 2007 CAD-CAM off line programming for industrial robots ROBOmove on line help v 2.0 [online] Available from: http://www.qdrobotics.com/eng/robomove.php [Accessed 18 April 2011] Siemens Corp, 2009 NX Documentation In: {$UGII_base_ dir}\UGDOC Stone, H.W., Sanderson, A.C., and Neuman, C.P., 1986 Arm signature identification In: IEEE International Conference on Robotics and Automation, San Francisco, CA New York: IEEE Press, 41–48 Sultan, I.A and Wager, J.G., 2001 A technique for the independent-axis calibration of robot manipulators with experimental verification International Journal of Computer Integrated Manufacturing, 14 (5), 501–512 Tsai, M.J., Lin, S.-D., and Chen, M.-C., 1992 Mathematical model for robotic arc-welding off-line programming system International Journal of Computer Integrated Manufacturing, (4), 300–309 573 Vincze, M., Prenninger, J.R., and Gander, H., 1994 A laser tracking system to measure position and orientation of robot end effectors under motion International Journal of Robotics Research, 13 (4), 305–314 Whitney, D.E., 1969 Resolved motion rate control of manipulators and human prostheses IEEE Transactions on Man-Machine Systems, 10 (2), 47–53 Whitney, D.E., Lozinsky, C.A., and Rourke, J.M., 1986 Industrial robot forward calibration method and results ASME Journal of Dynamic Systems, Measurement and Control, 108 (1), 1–8 Zhong, X.L and Lewis, J.M., 1995 A new method for autonomous robot calibration In: IEEE International Conference on Robotics and Automation New York, NY: USA, 1790–1795 Zhong, X.L., Lewis, J.M., and Rea, H., 1994 Neuroaccuracy compensator for industrial robots In: IEEE International Conference on Neural Networks, Vol New York: IEEE Press, 2797–2802 Zhu, Z., et al., 2004 Calibration of laser displacement sensor used by industrial robots Optical Engineering, 43 (1), 12–13 Zhuang, H., Motaghedi, S.H., and Roth, Z.S., 1999 Robot calibration with planar constraints In: IEEE International Conference on Robotics and Automation, 10 May 1999–15 May 1999, Detroit, MI, USA Vol New York: IEEE Press, 805–810 International Journal of Computer Integrated Manufacturing Vol 24, No 6, June 2011, 574–592 Integration of simulated annealing and genetic algorithm to estimate optimal solutions for minimising surface roughness in end milling Ti-6AL-4V Azlan Mohd Zaina*, Habibollah Harona and Safian Sharifb a Soft Computing Research Group, Department of Modeling and Industrial Computing, Faculty of Computer Science and Information System, Universiti Teknologi Malaysia, 81310 UTM Skudai Johor, Malaysia; bDepartment of Manufacturing and Industrial Engineering, Faculty of Mechanical Engineering, Universiti Teknologi Malaysia, 81310 UTM Skudai Johor, Malaysia (Received April 2010; final version received 23 February 2011) In this study, simulated annealing (SA) and genetic algorithm (GA) soft computing techniques are integrated to search for a set of optimal cutting conditions value that leads to the minimum value of machining performance Two integration systems are proposed; integrated SA–GA-type1 and integrated SA–GA-type2 The considered machining performance is surface roughness (Ra) in end milling The results of this study showed that both of the proposed integration systems managed to estimate the optimal cutting conditions, leading to the minimum value of machining performance when compared to the result of real experimental data The proposed integration systems have also reduced the number of iteration in searching for the optimal solution compared to the conventional GA and conventional SA, respectively In other words, the time for searching the optimal solution can be made faster by using the integrated SA–GA Keywords: integration systems; minimum machining performance; optimal cutting conditions Introduction Two huge issues in machining that were carried out and discussed at great length by authors in the machining studies are modelling and optimisation of the machining process (Bouaziz and Zghal 2008, Shabaka and ElMaraghy 2008, Samanta 2009, ElMounayri and Deng 2010, Jeang et al 2010, MunozEscalona and Maropoulos 2010) Generally, modelling is the process of estimating the potential minimum value of the machining performance, while optimisation is the process of estimating the potential minimum value of machining performance at the optimal point of cutting conditions The most common machining performance that is of interest to most machinists are cutting force, power, torque, tool-wear, tool-life, chipform, chip breakability, surface integrity, surface roughness and part accuracy (Jawahir et al 2003) Improvement in the quality could be indicated by referring to a performance measurement known as surface roughness (Ra) Generally, the Ra value is influenced by many factors such as machining parameters, cutting phenomena, work piece properties and cutting tool properties As shown in Figure 1, tool angle, one of the machining parameter factors is an effective parameter that affects the Ra There are various types of angles in machining cutting tools, such as the axial rake angle, the radial rake angle, the side *Corresponding author Email: azlanmz@utm.my ISSN 0951-192X print/ISSN 1362-3052 online Ó 2011 Taylor & Francis DOI: 10.1080/0951192X.2011.566629 http://www.informaworld.com rake angle, etc As shown in Figure (Benardos and Vosnaikos 2003), the tool geometry is one of the significant parameters that affects the Ra The various types of tool angles in end milling are the axial rake angle, radial rake angle, helix angle, etc It was found that the optimisation of cutting conditions for Ra in end milling involving radial rake angle is still lacking, in particular when dealing with titanium alloys As such, optimisation of the cutting conditions, which include radial rake angle, cutting speed and feed on the Ra in end milling of Ti-6A14V can be considered as a new contribution to machining research Modelling techniques in estimating the potential minimum value of machining performance such as Ra can generally be classified into two groups With conventional approaches such as Regression technique, explicit models are developed that required complex physical understanding of the modelling process With non-conventional or artificial intelligence (AI) approaches such as artificial neural network (ANN), fuzzy logic (FL) and genetic algorithm (GA)based modelling, implicit models are created within the weight matrices of the net, rules and genes that are easier to be implemented Similar to modelling techniques, optimisation techniques to estimate the potential minimum value of machining performance at the optimal point of cutting conditions also can be International Journal of Computer Integrated Manufacturing Figure Parameters that affect the surface roughness, Ra (Benardos and Vosnaikos 2003) classified into conventional approaches and AI approaches The techniques that are classified as the conventional optimisation approaches are the Taguchi technique, Factorial technique and response surface methodology (RSM) technique (Mukherjee and Ray 2006) Some established AI approaches applied by previous works to suggest the optimal cutting conditions in machining problem are GA, simulated annealing (SA), Tabu search (TS), ant colony optimisation (ACO) and particle swarm optimisation (PSO) (Aggarwal and Singh 2005, Mukherjee and Ray 2006) In this study, regression modelling technique combined with GA and SA optimisation techniques are applied in order to estimate the minimum value for machining performance at the optimal cutting conditions Some of the advantages of GA in optimising cutting conditions for machining problems may include (Cus and Balic 2003, Mukherjee and Ray 2006): (i) preferred when near-optimal conditions instead of the exact optimal solution are cost effective and acceptable for implementation by the manufacturers; (ii) a derivative-free approach for near-optimal point(s) search direction; (iii) able to handle objective functions of any complexity with both discrete (for example, integer) and continuous variables; (iv) a simple complementation of the model by new input parameters without modifying the existing model structure; (v) an automatic search for the non-linear connection between the inputs and outputs; (vi) a fast and simple optimising technique Some of the advantages of SA in optimising cutting conditions for machining problems may include (Tarng et al 1995, Khan et al 1997, Wang et al 2005): (i) SA algorithm is very easy to program, typically it takes only a few hundred lines of a computer code, (ii) implementation of a new problem often only takes very little modification of the existing code, (iii) SA algorithm does not need to calculate the gradient descent that is required for most traditional optimisation algorithms; 575 this means that the SA algorithm can be applied to all kinds of objective and constraint functions, (iv) SA algorithm can find the global minimum more efficiently instead of trapping in a local minimum where the objective function has surrounding barriers, and (v) SA search is independent of initial conditions Recently, the new trend taken up by some researchers in estimating the potential minimum value of the machining performance with the recommended optimal machining cutting condition is by using the integration system Integration system is a combination of two or more techniques with the target to obtain a more successful result for the investigated problem For example, the ANN model integrated with PSO optimisation technique was used to optimise the cutting parameters subject to a comprehensive set of constraints, which are feed and speed for a typical case found in industry, namely, pocket-milling (Tandon et al 2002) The ANN model and PSO optimisation technique were integrated to determine the optimum solution for the computer numerical control (CNC) ball end milling, and it was found that the feed rate has a dominant effect on the machining efficiency (El-Mounayri et al 2005) The capability of GA as the optimisation technique in investigating the optimal cutting conditions for Ra value in milling process was given by Oktem et al (2005) and Palanisamy et al (2007) This technique was integrated with regressionbased model such as second order regression RSM model and dynamic programming model In this study, integration system was introduced to observe the possible improvement in the result that may be obtained particularly for the end milling machining problem In order to estimate the optimal cutting conditions that lead to the minimum value of the Ra machining performance, this study focuses on two soft computing techniques, SA and GA, that will be integrated to form a integration system, named integrated SA–GA With integrated SA–GA, it is expected that it can produce more significant result for the Ra value compared to the experimental result The proposed approach The proposed approach involves five modules, which are experimental data, regression model, GA singlebased optimisation (conventional GA), SA singlebased optimisation (conventional SA) and integrated SA–GA-based optimisation The objectives of the proposed integrated SA–GA are: (1) to estimate the minimum value of the machining performance compared to the machining performance value of the experimental data, regression, conventional GA, conventional SA 576 A.M Zain et al and RSM (machining case study by Mohruni 2008) (2) to estimate the optimal cutting conditions values that have to be within the range of the minimum and maximum coded values for cutting conditions of experimental design that are used for experimental trials (3) to estimate the optimal solution with the small number of iterations compared to the optimal solution with conventional GA and conventional SA The steps for development of the proposed integrated SA–GA-based optimisation are given as follows: Step 1: Define the objective function State the predicted equation of the regression model to define the objective function of optimisation Step 2: Define the limitation constraints Propose the optimal cutting conditions values of GA combined with the optimal cutting conditions values of SA to define the limitation constraints of optimisation, then move to Step (labelled as SA–GA-type1), or propose the optimal cutting conditions values of GA combined with the optimal cutting conditions values of SA to define the limitation constraints of optimisation (labelled as SA–GA-type2) Step 3: Define the initial constraints Propose the optimal cutting conditions values of GA to define the initial points of optimisation Step 4: Generate the optimal points Estimate a minimum predicted machining performance value at the optimal solution milling of titanium Ti-6Al-4V is given in Table From this table, the coded variables used in the 23-factorial design are only for levels 71, and þ1 The whole experiments were carried out under flood conditions (6% concentration of water base coolants) with mm constant axial depth of cut and mm constant radial depth of cut A CNC MAHO 700S machine centre was used for the milling experiments The Ra value of the machined work piece was measured using the Taylor Hobson Surftronic þ3 with resolution 0.01 mm Initially, the instrument was calibrated using a standard roughness specimen to ensure the consistency and accuracy of Ra values Five measurements were conducted at the location of the length of cut on the work piece and the average Ra value was recorded, and the experimental set-up is shown in Figure 3.2 Related to the investigated problem, an experiment that dealt with the Ra measurement was conducted In the experiment, 24 experimental trials were executed that were based on eight data of two levels of DOE 2k full factorial analysis with four centre data and twelve axial data All the data were tested in real machine to show the actual value (experiment result) of Ra for three different cutting tools, which are uncoated, TiA1N-coated and SNTR-coated cutting tools The Ra values of each type of cutting tool that were observed for the selected cutting conditions in the experiment are given in Table Overall, the flow to develop the proposed integration systems, SA–GA-type1 and SA–GA-type2, are illustrated in Figure Experimental results Regression modelling In general, the measurement of Ra in end milling in relation to the independent variable is expressed mathematically as follows: Ra ¼ cvk f l gm e0 ð1Þ Experimental data The machining experiment conducted by Mohruni (2008) to measure the Ra value in the end milling operation was considered in this study The work piece used in the experiments was an annealed alpha–beta titanium alloy, Ti-6AL-4V Three types of end mills were used in the experiments, namely uncoated carbide (WCCo) and two TiAlN base coated carbide tools which include common PVD-TiAlN coated carbide tool and PVD with enriched Al-content TiAlN coated carbide tools (also called Supernitride coating or SNTR) where Ra is the surface roughness value in mm, v is the cutting speed in m/min, f is the feed rate in mm/tooth, g is the radial rake angle in8, e0 is the experimental error and c,k, l, m are the model parameters to be estimated using the experimental data To develop the regression model for estimating the Ra value, the mathematical model given in Equation (1) is linearised by performing a logarithmic transformation as follows: 3.1 Experimental design According to the design of the experiment for the three independent variables, the coding variables for the end Equation (2) can be rewritten as: ln Ra ¼ ln c þ k ln v þ l ln f þ m ln g þ ln e0 y ¼ b0 x0 þ b1 x1 þ b2 x2 þ b3 x3 þ e ð2Þ ð3Þ 577 International Journal of Computer Integrated Manufacturing Figure Table The flow of the proposed integrated SA–GA Coded values for cutting conditions of experimental design Level in coded form Independent variables Units 71.4142 71 Cutting speed, v Feed rate, f Radial rake angle, g m/min mm/tooth 124.53 0.025 6.2 130.00 0.03 7.0 where y is the logarithmic value of the experimental Ra, x0 ¼ is a dummy variable, x1, x2 and x3 are the cutting condition values (logarithmic transformations) of v, f 144.22 0.046 9.5 þ1 þ 1.4142 160.00 0.07 13.0 167.03 0.083 14.8 and g, respectively, e is the logarithmic transformation of experimental error e0 and b0, b1, b2 and b3 are the model parameters to be estimated using the experimental data 578 A.M Zain et al Next, Equation (3) can also be written as follows: 4.1 y^ ¼ y À e ¼ b0 x0 þ b1 x1 þ b2 x2 þ b3 x3 ð4Þ where y^ is the logarithmic value of the predictive (estimated) Ra Subsequently, this equation will be proposed as the objective function of the optimisation solution Regression model for each cutting tool Regression models for each cutting tool that are given in Equation (4) are developed by using Statistical Package for the Social Science (SPSS) software, based on the data for the real machining experiments given in Table The values of coefficients for the model parameters of uncoated, TiA1N coated and SNTR coated cutting tools are given in Table 3, Table and Table 5, respectively By transferring the values of coefficients of each cutting tool from Tables 3–5 into Equation (4), the regression model equations can be written as follows: y^1 ¼ R^a uncoated ¼ 0:451 À 0:00267x1 þ 5:671x2 þ 0:0046x3 ð5aÞ y^2 ¼ R^a TiAlN ¼ 0:292 Figure Table No À 0:000855x1 þ 5:383x2 À 0:00553x3 ð5bÞ y^3 ¼ R^a SNTR ¼ 0:237 À 0:00175x1 þ 8:693x2 À 0:00159x3 ð5cÞ Equations (5a)–(5c) are applied to calculate the predicted Ra values of regression, and the results are summarised in Table Experimental set-up (Mohruni 2008) Ra values for real machining experiments Setting values of experimental cutting conditions Data source DOE 2k Centre 10 11 12 13 Axial 14 15 16 17 18 19 20 21 22 23 24 Ra (minimum) Ra (maximum) Experimental Ra value (mm) v (m/min) f (mm/tooth) g (8) Ra_uncoated Ra_TiA1N Ra_SNTR 130 160 130 160 130 160 130 160 144.22 144.22 144.22 144.22 124.53 124.53 167.03 167.03 144.22 144.22 144.22 144.22 144.22 144.22 144.22 144.22 0.03 0.03 0.07 0.07 0.03 0.03 0.07 0.07 0.046 0.046 0.046 0.046 0.046 0.046 0.046 0.046 0.025 0.025 0.083 0.083 0.046 0.046 0.046 0.046 7 7 13 13 13 13 9.5 9.5 9.5 9.5 9.5 9.5 9.5 9.5 9.5 9.5 9.5 9.5 6.2 6.2 14.8 14.8 0.365 0.256 0.498 0.464 0.428 0.252 0.561 0.512 0.464 0.444 0.448 0.424 0.328 0.324 0.236 0.240 0.252 0.262 0.584 0.656 0.304 0.288 0.316 0.348 0.236 0.656 0.32 0.266 0.606 0.476 0.260 0.232 0.412 0.392 0.324 0.380 0.460 0.304 0.360 0.308 0.340 0.356 0.308 0.328 0.656 0.584 0.300 0.316 0.324 0.396 0.232 0.656 0.284 0.196 0.668 0.624 0.280 0.190 0.612 0.576 0.329 0.416 0.352 0.400 0.344 0.320 0.272 0.288 0.230 0.234 0.640 0.696 0.361 0.360 0.368 0.360 0.190 0.696 [...]... manufacturing control In: Proceedings of Rensselaer’s 2nd international conference on computer integrated manufacturing, 21–23 May, Troy, NY, USA, 444–450 International Journal of Computer Integrated Manufacturing Jain, A ., Jain, P ., and Singh, I ., 2006 An integrated scheme for process planning and scheduling in FMS International Journal of Advanced Manufacturing Technology, 3 0, 1111–1118 Joo, J .,. .. 2–14 Wang, L ., Feng, H.Y ., and Cai, N ., 2003 Architecture design for distributed process planning Journal of Manufacturing Systems, 22 (2 ), 99–115 Wang, L ., Jin, W ., and Feng, H.Y ., 2006 Embedding machining features in function blocks for distributed process planning International Journal of Computer Integrated Manufacturing, 19 (5 ), 443–452 Wang, L ., Song, Y ., and Shen, W ., 2005 Development of a function... systems International Journal of FMS, 1 1, 5–17 Gindy, N ., Saad, S ., and Yue, Y ., 1999 Manufacturing responsiveness through integrated process planning and scheduling International Journal of Production Research, 37 (11 ), 2399–2418 Gliffer, B and Thompson, G.L ., 1960 Algorithms for solving production scheduling problems International Journal of Operational Research, 8, 161–171 Grabowik, C ., Kalinowski, K .,. .. Brazil 2 7, 87–94 Liao, T.W ., et al ., 1993 Modification of CAPP systems for CAPP/scheduling integration Computers & Industrial Engineering, 25 (1–4 ), 203–206 Mamalis, A.G ., Malagardis, I ., and Kambouris, K ., 1996 On-line integration of a process planning module with production scheduling International Journal of Advance Manufacturing Technology, 1 2, 330–338 Moon, C ., Kim, J ., and Hur, S ., 2002 Integrated. .. scheduling Computer and Industrial Engineering, 1 5, 296–307 Tan, W and Khoshnevis, B ., 2000 Integration of process planning and scheduling – a review Journal of Intelligent Manufacturing, 1 1, 51–63 Tonsho , H.K ., Beckendor , U ., and Andres, N ., 1989 FLEXPLAN: a concept for intelligent process planning and scheduling In: CIRP International Workshop, 21–2 2, September, Hannover, Germany, 87–106 Ueda, K ., Fuji,... Design, 13 (2 ), 65–82 Huang, G.Q ., Yee, W.Y ., and Mak, K.L ., 2001 Development of a web-based system for engineering change management Robotics and Computer- Integrated Manufacturing, 1 7, 255–267 Huang, G.Q ., Yee, W.Y ., and Mak, K.L ., 2003 Current practice of engineering change management in Hong Kong manufacturing industries Journal of Materials Processing Technology, 139 (1–3 ), 481–487 551 Jarratt, T.A.W .,. .. Applications, 6 (2 ), 88–100 Saygin, C and Kilic, S ., 1999 Integrating flexible process plans with scheduling in flexible manufacturing systems International Journal of Advanced Manufacturing Technology, 1 5, 268–280 Shao, X ., et al ., 2009 Integration of process planning and scheduling: a modified genetic algorithm-based approach Computers and Operations Research, 3 6, 2082– 2096 Shen, W ., Wang, L ., and Hao, Q ., 2006... optimization of an integrated process planning and scheduling model International Journal of Production Research, 47 (1 ), 119–142 Chan, F ., Zhang, J ., and Li, P ., 2001 Modelling of integrated, distributed and cooperative process planning system using an agent-based approach Proceedings of the Institution of Mechanical Engineering, Part B: Journal of Engineering Manufacturing, 215 (10 ), 1437–1451 Cho, K.K ., et... Integrated Manufacturing, 20 (1 ), 80–95 Li, W ., et al ., 2008a Game theory-based cooperation of process planning and scheduling CSCWD200 8, In: 12th international CSCWD, 16–18 April, Xi’an, China, 841– 845 Li, X ., et al ., 2010b A review on integrated process planning and scheduling International Journal Manufacturing Research, 5 (2 ), 161–180 Li, X ., et al ., 2008b A genetic algorithm for integration of process... In: Proceedings of IEEE international conference on mechanical and automation, 25–28 June, Luoyang, China, 2290–2295 Zijm, W.H.M and Kals, H.I.J ., 1995 The integration of process planning and shop floor scheduling in small batch part manufacturing CIRP Annals - Manufacturing Technology, 44 (1 ), 429–432 International Journal of Computer Integrated Manufacturing Vol 2 4, No 6, June 201 1, 535–551 A method