application of taguchi method to analyze the …

102

Int. J. Mech. Eng. & Rob. Res. 2013 S Vijayan et al., 2013

APPLICATION OF TAGUCHI METHOD TOANALYZE THE IMPACT OF COMMON PROCESSES

IN MULTISTAGE PRODUCTION SYSTEM

S Vijayan1*, R Ashok Raj1, T Parun1 and K Periasamy1

*Corresponding Author: S Vijayan,[email protected]

This paper examines the optimum level of batch size in bottleneck facility and to analyze theeffect of common processes on throughput and cycle time in a production system under uncertainsituations created by machine breakdown. Few simulation models are carried out by comparingthe existing model and proposed model of the production system in the company. The modelsare verified and validated with the historical data from the company. Taguchi approach fororthogonal array is used in designing experiment. The experimental settings are executed inWITNESS, a simulation package. The delivery performance, average production cycle time andproduction quantity for both line 1 and line 2 are examined. It is observed that the variation inlevel of common process in the system has significant impact on the production quantity andcycle time. The main contribution of this paper is determination of the optimal level of batch sizein a bottleneck resource. This approach can be generalized to any multistage production system,regardless of the precedence relationships among the various production stages in the system.

Keywords: Batch size, Throughput, Cycle time, Taguchi approach, WITNESS

INTRODUCTIONMultistage production planning is a systemwhich transforms or transfer inventories througha set of connected stages to produce thefinished goods. The stages represent thedelivery or transformation of raw materials,transfer of work-in progress betweenproduction facilities, assembly of componentparts, or the distribution of finished goods. The

ISSN 2278 – 0149 www.ijmerr.comVol. 2, No. 4, October 2013

© 2013 IJMERR. All Rights Reserved

Int. J. Mech. Eng. & Rob. Res. 2013

1 Department of Mechanical Engineering, J.J College of Engineering & Technology, Trichy.

fundamental challenge of multi-stageproduction is the propagation andaccumulation of uncertainties that influencesthe conformity of the output. The present studyis concern with such a multistage system andsimulation is chosen to analysis the objectives.A simulation is a surrogate for experimentingwith a real manufacturing system. It is ofteninfeasible or not cost-effective to do an

Research Paper

103


experiment in a real process. Few simulationmodels are used to analyze various effects ofuncertain factors namely machine breakdown.

Machine breakdown means the failure orstoppage of machines for unknown reasonsand a representation of interruption in theprocess. It yields a reduction of capacity leveland delays the release of products or subassemblies. In this production system there isno alternative machines are available if theexisting machines fail and no alternativerouting can be executed if an order needs tobe expedite.

THE PRODUCTION SYSTEMThe above figure shows the production line ofstator core only. There are two different endproducts, product SA15 (line 1) and A115 (line2) of this system. Parts are placed in twoseparate production lines. Each productionline contains 7 (seven) different processing

Figure 1: Existing Production Layout

Figure 2: Proposed/Modified Production Layout

namely blanking, riveting, coining, T’Boltmilling, second coining, OD-Turning and slotinsulation and ended up with single endproducts after the assembly operation. Figure1 is showing the existing production layout ofthe company. Presently the company use theconventional production processes with knownlead time.

EXPERIMENTAL DESIGNFew simulation models are carried out basedon the existing production layout (Figure 1) ofthe company. The existing layout is modifiedto introduce common processes in the system.

Figure 2 shows the proposed layout thatincorporates commonality dimension. Twomodels are developed in WITNESS simulationpackage.

In this paper, two factors are consideredand the effects of these factors on the systemperformance are tested. The levels of common

104


process and production batch size atblockage station are considered as controlfactor or decision variable is shown in Table1. The machine breakdown is considered asnoise factor. Analyze of mean value, signal tonoise ratio and ANOVA are used to analyze

the effect of batch size and common processon production cycle time and throughputquantity. Before confirmed the results,interaction effect are observed to make surethat the characteristic of the control factors isadditive. The ranges of factor levels areselected based on capacity limitation and inconsultation with the engineers in thecompany. Since this study contains two controlfactors of three levels.

Each experiment is simulated, theaverage value and its signal to noise ratioare obtained and analyzed. In order toevaluate the experimental results statistically,analysis of variance (ANOVA) is applied. Thesame are used to see the effect of theinteraction.

Figure 3: Experimental Design

Table 1: Control Factors and Their Levelsfor Taguchi method

105


DATA COLLECTION ANDVALIDATIONIn order to build the simulation models and toset the initial level of various factors in themodel, data were collected. The data includescycle time at each stages and setup time.

The historical data shows that the cycle timeand the setup time for OD-Turning station aremuch higher than the others. It is the bottleneckof the system shown in Table 2. Therefore inthis paper different levels of batch size areconsidered to analyze the effects of productionquantity and cycle time.

DATA ANALYSIS ANDDISCUSSIONTables 3 and 4 shows the summary ofexperimental results for the averageproduction quantity and production cycle timewith corresponding S/N ratio for each exerciseof Line 1 and Line 2 are observed. The smaller

the better characteristic is used for cycle timeand in calculating the corresponding S/N ratio.The larger the better principle is adopted forproduction quantity and for corresponding S/N ratio.

Since the experiment design is orthogonal,the effect of batch size and common processfor different levels are separated out. Table 5shows the response for mean and S/N ratiofor production quantity and the same forproduction cycle time is shown in Table 6 forboth of the production lines. Since thecharacteristic of these factors for productionquantities are larger the better, the batch sizesare chosen based on larger mean and S/Nratio for production level and for productioncycle time, the smaller the better policy is used.The selection in later case is based on thesmaller mean and larger S/N ratio. Becausethe larger the S/N ratio the smaller the varianceare around the desired value.

Table 2: Cycle Time and Setup Time at Each Stage

106


It shows that an increase in batch size yield anincrease in production level in the system, butthe capacity restrains the further increase inthe batch size. Thus, based on responseTables 5 and 6, the batch size and commonprocess are chosen as 900 and 4 respectively.

Figures 4 and 5 shows the interactioneffects of variation in levels of control factorsfor (a) mean value and (b) S/N ratio ofproduction quantity for Line 1 and Line 2.Figure 4 shows there is an interaction at level3 (i.e., in level 1500), similarly the interaction

Table 3: Experimental Results of Production Quantity with Corresponding S/N Ratio

Table 4: Experimental Results of Production Cycle Time with Corresponding S/N Ratio

107


Table 5: Response Table for Production Quantity for Line 1 and 2 (the Larger the Better)

Table 6: Response Table for Production Cycle Time for Line 1 and 2(the Smaller the Better)

Figure 4: Interaction Plot for (a) Mean Value and (b) S/N Ratio of Production Levelfor Line 1

Batch Size

SNra

tios

1500900450

86

85

84

83

82

Common

4

Process02

Interaction Plot (data means) for SN ratios

Signal-to-noise: Larger is better

Batch Size

SNra

tios

1500900450

86

85

84

83

82

Common

4

Process02



108


Figure 5: Interaction Plot for (a) Mean Value and (b) S/N Ratio of Production Levelfor Line 2

Batch Size

Mea

ns

1500900450

27000

26000

25000

24000

23000

22000

21000

20000

Common

4

Process02

Interaction Plot (data means) for Means

Batch Size

SNra

tios

1500900450

88.5

88.0

87.5

87.0

86.5

86.0

Common

4

Process02



Figure 6: Interaction Plot for (a) Mean Value and (b) S/N Ratio of Production Cycle Timefor Line 1

Figure 7: Interaction Plot for (a) Mean Value and (b) S/N Ratio of Production Cycle Timefor Line 2

109


plot for S/N ratio (in level 1500), the interactionis found. There is no interaction is found inFigure 5.

Figures 6 and 7 shows the interactioneffects of variation in levels of control factorsfor (a) mean value and (b) S/N ratio ofproduction cycle time for Line 1 and Line 2.Figure 6 shows the interaction at level in themean plot. The interaction plot explains, as atbatch size level 450 to 1500 there is aninteraction with the level of common process0, 2 and 4. The interaction plot Figure 7explains, as at batch size level 1500 there isan interaction with the level of commonprocess 2 and 4. The production quantity ispeak and cycle time is least when the batchsize is 900 and common process is at thehighest level.

Tables 7 and 8 shows the ANOVA forproduction quantity in mean and S/N ratio forLine 1 and Line 2. The same for productioncycle time for both of the lines are shown inTables 9 and Table 10. These tables show therelative importance of the control factorsaffecting the throughput and cycle time. Bothmean and signal to noise ratio ANOVAindicates that batch sizes in OD-Turning station(factor A) and the use of common process(factor B) is statistically significant. The factorshave impacts on production quantity and cycletime.

Based on ANOVA (Tables 7, 8, 9 and 10)and response Tables 5 and 6, it is obvious thatbatch size of 900 in the OD-Turning station and4 common processes yield the lowest cycletime and maximum throughput in the system.

Table 7: ANOVA for Mean Value and S/N Ratio of Production Quantity of Line 1

Table 8: ANOVA for Mean Value and S/N Ratio of Production Quantity of Line 2

110


CONCLUSIONThe developed simulation models for themanufacturing of the company underconsideration are verified and validated asdescribed

• Batch size in the bottleneck (that is, OD-Turning) in combination of common processdrastically improves the measured systemdeliveries. ANOVA for mean and S/N ratiofor cycle time and through put indicate thatno important factor is omitted from theexperiment.

• There is a significant interaction among thecommon process and the batch size in thebottleneck facility (that is OD-Turning station).It is observed that under the combined effectof common process, the cycle time reduced

by 22.6% and 17.40% respectively andthroughputs increased by 38% and 28%.

• Based on the least manufacturing cycle timeand maximum throughput the optimumbatch sizes 900 in the bottleneck (that isOD-Turning) when four common processesensure the best outcome of the system.

• The mean value and S/N ratio of throughputand cycle time in batch size and commonprocess P-value is less than 1%. Thereforebatch size and common process isstatistically significant.

• The interaction effect of batch size andcommon process is statistically significant.Therefore the optimal level yields thelowest cycle time and highest productionquantity.

Table 9: ANOVA for Mean Value and S/N Ratio of Production Cycle Time of Line 1

Table 9: ANOVA for Mean Value and S/N Ratio of Production Cycle Time of Line 2

111


REFERENCES1. Amirul Khair Salehun M D and Awaluddin

Mohamed Shahroun (2005), “Guide Linefor Improving Capacity of a Job ShopBased on Selected PerformanceMeasures”, Journal of ProductionResearch, Vol. 43, No. A, pp. 87-104.

2. Atul Kumar Sahu (2009), “EfficientHeuristics for Scheduling Tasks on a FlowShop Environment to OptimizeMakespan”, Journal of ProductionResearch.

3. Chou-Jung Hsu, Wen-Hung Kuo, Dar-li Yangand Maw-Shengchern (2006), “Minimizingthe Makespan in a Two Stage Flow ShopScheduling Problem with a FunctionConstraints on Alternative Machines”,Journal of Marine Science andTechnology, Vol. 14, No. 4, pp. 213-217.

4. Gur Mosheiov and Assaf Sarig (2010),“Minimum Weighted Number of Tardy

Jobs on an m-Machines Flow Shop witha Critical Machine”, Journal of OperationResearch.

5. Helmut Bley, Claas Christian Wutike andWolfgang Massberg (1997), “DistributedSimulation Applied to ProductionSystems”, Journal of ManufacturingTechnology, Vol. 46, pp. 361-364.

6. Ouelhaoj D and Petrouic S (2009), “Surveyof Dynamic Scheduling in ManufacturingSystems”, Journal of AutomatedScheduling, Optimization and PlanningResearch.

7. Wazed M A, Ahmed S and Nukman Y(2010), “Application of Taguchi Method toAnalyze the Impacts of Commonalities inMultistage Production System UnderBottleneck and Uncertainty”, TheInternational Journal of the PhysicalScience, Vol. 5, No. 10, pp. 1576-1591.

application of taguchi method to analyze the …

Documents