Vous êtes sur la page 1sur 5

Selection of a Conveyor Belt Material under Multi-criteria Decision Making Environment

Dr S Bhattacharya, Non-member B Sarkar, Associate Member R N Mukherjee, Non-member


Conveyor belts are required to transport different types of materials in industries under varying operating conditions. A number of materials for such belts are available in the market. However, it is impossible to identify one such material which will perfectly respond to all such requirements in a given environment. Hence, arises the question of optimisation. The present article is based on one such optimisation technique known as analytical hierarchy process (AHP). The algorithm used here is obtained by integrating Brown and Gibson model and AHP. It considers a number of alternative materials as well as design criteria (respectively 4 and 5 in the present paper) and arrives at the most suitable material. The algorithm calculates the priority vector (PV) values by eigen-vector method of pair-wise comparison. These PV values are calculated through an iterative process, and are accepted if the concerned matrices yield consistency ratios (CR) below a specific limit. Then the composite priority (CP) values are calculated, and the alternatives are ranked according to their CP values. Next, the sensitivity analysis is taken up where the CP values are used as subjective factor measures (SFM) in determining belt material measure (BMM). The algorithm is coded in FORTRAN language. The program is general, and can be used to handle smaller or larger number of criteria and alternatives as well.
Keywords: Conveyor belt; Analytical hierarchy process; Brown and Gibson model; Belt material measure; Sensitivity analysis

INTRODUCTION The belt conveyor system is the most common method used in the industry for conveying bulk materials1. Belt conveyors are presently utilised for material handling of all kinds of pulverised, granular and lumpy materials. There are mainly four types of belts used for power transmission flat, V, ribbed and toothed2. Of these, the first two are most widely used. The selection of a belt drive depends upon (i) power to be transmitted, (ii) speeds and orientations of the driver and driven shafts, (iii) speed ratio, (iv) centre distance and (v) space available etc. The different types of flat belts are (i) open, (ii) crossed, (iii) quarter twist, (iv) right angled, (v) stepped, (vi) fast and loose, (vii) compound, and (viii) reverse, etc. Choosing an appropriate material for a conveyor belt to suit some particular application is a problem with no definite or unique solution. Usually the demands on the belt performance are many, and the choice of materials wide-ranging. Among the most important characteristics of belt materials are3 (i) strength (modulus of elasticity, ultimate tensile strength, flexibility, etc), (ii) friction, (iii) slip, (iv) creep, (v) density, (vi) durability, and (vii) cost, etc. Some of the popular belt materials are leather, fabric and canvas, rubber, balata, woven cotton, etc. However, none of these caters to all the requirements equally well. The demands
Dr S Bhattacharya is with the Department of Mechanical Engineering and Dr B Sarkar is with the Department of Production Engineering, both at Jadavpur University, Kolkata 700 032; while Dr R N Mukherjee is with the Department of Mathematics, Burdwan University, West Bengal. This paper (revised) was received on December 23, 2004. Written discussion on the paper will be entertained till December 31, 2005.

are often conflicting too. The requirement of high flexibility of the belt material (to absorb shocks and damp out vibrating forces so as to increase the life of the driving machine), for example, is often countered by the high cost of such materials. In older designs, the cost-reduction element used to be recognised as the most important factor in the selection of belt materials4. Presently the approach is different. In cases of conflicting demands and varying choices, the designer usually assigns relative priority values to the concerned parameters and choices, and tries to arrive at a good and not the best solution. Such design procedures are generally called multi-criteria design making (MCDM) process. PROBLEM DEFINITION Multi-criteria Decision Making Process MCDM may have as its input several kinds of problems. However, there are some common characteristics of such problems. These include: l Multiple objectives/attributes: Each problem has multiple objectives/criteria, which are often conflicting. However, the objectives or attributes must be precisely defined. Incommensurable units: Each objective/attribute has a different unit of measurement. Design/selection: The aim of MCDM is to finally select one alternative which fits best with the criteria requirements. A criterion is in fact a measure of effectiveness. These emerge as forms of attributes or objectives in the actual problem setting. IE (I) JournalMC

l l

140

Table 1

The nine-point scale of pair-wise comparison Definition Equally important Moderately preferred Essentially preferred Very strongly preferred Extremely preferred Intermediate importance between two adjacent judgments

Intensity of relative importance 1 3 5 7 9 2, 4, 6, 8

measure (CFM) and objective factor decision weight (a ) are related by the equation:
BMM i = CFM i [( OFM i ) + ( 1 ) SFM i ]

(2)

where i = 1, 2, 3, . . . , n are the number of alternative materials and 0 1. The philosophy embedded in the development of the algebraic expressions [equations (1) and (2)] is a hybrid of cardinal as well as ordinal factors. While selecting a proper belt material, it considers all aspects, and is indeed a holistic and multidimensional approach. Usually the value of a is taken as 0.67. It may be mentioned here that OFMi values are tangible, whereas, SFMi values are intangible. The CFM values vary as the location changes. Its value may be either 0 or 1. CFM is the product of the individual critical factor indices for each value of i with respect to critical factor j. In the present methodology, the CFM value as taken as 1 since location does not affect the selection of belt material. The basis for taking the value of a (objective factor decision weight), a synergised value in the Brown and Gibson model, as 0.67, is the brain-storming session where experts from design, manufacturing, operations and maintenance sections were present. THE BELT MATERIAL SELECTION PROBLEM Five criteria (eg, strength, friction, slip, density, durability) have been considered and coded as C1, C2, C3, C4 and C5. Four materials are considered and coded as M1, M2, M3 and M4. It is presumed that the behaviours/performances of these four alternatives (materials) with respect to each of the five criteria are known. Then AHP is used to arrive at a decision, based on the methodology given here. The Proposed Methodology The proposed methodology is the integration of Brown and Gibson model and AHP. This methodology is applied to calculate the priority weights for functional, design factors and other important attributes by eigen vector method for each pairwise comparison matrix. Next, global priorities of various attributes are found by using AHP. These global priority values are used as the SFM in the Brown and Gibson model. The pairwise comparison matrices for the alternatives and five different factors (Tables 3 to 8) are constructed on the basis of Saatys nine-point scale (Table 1). The different cost attributes related to the selection of belt material are shown in Table 2. Four
Table 2 Cost factor components Belt material M1 Cost of acquisition, 105 Rs 105 Rs 1.89 1.35 3.66 2.34 9.24 0.19 Cost of manufacturing, M2 1.21 3.78 1.31 0.68 6.98 0.25 M3 1.26 2.00 1.12 0.89 5.27 0.34 M4 3.05 0.65 0.55 3.91 8.16 0.22

A MCDM problem involves at least two conflicting criteria and at least two alternative solutions to the problem. There are several types of methods for MCDM processes. One of them is analytical hierarchy process (AHP), which has been adopted in the present case. Analytical Hierarchy Process AHP is a powerful tool in formulating and handling complex, multi-person and multi-period problems hierarchically. This process is used in a large number of applications including engineering design, economic planning, energy policy, project selection, budget allocation, etc. It can provide useful insight into the trade-offs embedded in a decision-making problem. The three fundamental steps of AHP are: (i) defining a multi-criteria problem hierarchically. (ii) assigning relative priorities to the various elements using pair-wise comparison techniques. (iii) integrating these priorities to converge at an overall evaluation of decision alternatives. For assigning weights to the alternatives as well as criteria for constructing the decision matrix and pair-wise comparison matrices, the concept of relative importance is used to arrive at the decision-makers preferences. The intensity scale of importance (Table 1) introduced by Saaty5 has been used here. The decision matrix and the pair-wise comparison matrices used in AHP are all square matrices. The consistency of the judgment values assigned to the decision alternatives and criteria are checked using eigen values and eigen vectors. Based on these checks, the decision-maker revises and modifies the judgment values, if required. BROWN AND GIBSON MODEL Brown and Gibson developed a model based on the multiattribute preference theory6. In this model, the objective factor cost (OFC) and the objective factor measure (OFM) are interrelated by the expression:
OFM i = [OFC i ( 1 / OFC i )] 1

Cost of operation, 105 Rs Cost of maintenance, 105 Rs Total cost, 105 Rs Objective factor measure (OFM)

(1)

The belt material measure (BMM), objective factor measure (OFM), subjective factor measure (SFM), critical factor Vol 86, October 2005

141

Table 3

Decision matrix I II 1/3 1 3 1/3 1/5 III 1/2 1/3 1 1/4 1/7 IV 1/3 3 4 1 1/3 V 2 5 7 3 1

different types of cost components for belt materials have been considered. These actually are the OFC of the materials. Algorithm of the Proposed Methodology l Step 1 : Listing of the set of alternative materials (M1 to M4 ). l Step 2 : Identification of the design criteria ( C1 to C5 ). l Step 3 : Determination of the OFM values for each of the alternatives of materials, using the relation OFM i = [OFC i (1 / OFC i )] 1 , where OFCi = Objective Factor Component, for i = 1, 2, 3, . . . , n (number of alternative materials). l Step 4 : Assigning weights to each of the criteria based on the relative importance of its contribution according to the nine-point scale. This is the decision matrix. l Step 5 : For each criterion, assigning weights to each of the alternatives, based on its relative importance, according to the nine-point scale. These are called the pair-wise comparison matrices. l Step 6 : Determination of the priority vectors (PV) for decision matrix and for each of the pair-wise comparison matrices. Multiplication of the sum of each column with the corresponding PV value. Calculation of the sum of these products, ie, the principal eigen value max . l Step 7 : Calculation of the Consistency Index (CI) for each of the matrices, according to the formula,
CI = max n n 1

I II III IV V Table 4

1 3 2 3 1/2

Pair-wise comparison matrix for criterion 1 M1 M2 1/5 1 1/5 1/3 M3 1/5 5 1 3 M4 1/3 3 1/3 1

M1 M2 M3 M4 Table 5

1 5 5 3

Pair-wise comparison matrix for criterion 2 M1 M2 5 1 5 3 M3 1/3 1/5 1 1/3 M4 3 1/3 3 1

M1 M2 M3 M4 Table 6

1 1/5 3 1/3

Pair-wise comparison matrix for criterion 3 M1 M2 3 1 5 1/3 M3 1/3 1/5 1 1/5 M4 3 3 5 1

M1 M2 M3 M4 Table 7

1 1/3 3 1/3

(3)

where n is the order of each matrix. l Step 8 : Determination of the Random Consistency Index (RI) for each of the matrices, as per the formula
RI = 1.98 ( n 2 ) n

Pair-wise comparison matrix for criterion 4 M1 M2 1/3 1 5 3 M3 1/7 1/5 1 1/3 M4 1/5 1/3 3 1

(4)

M1 M2 M3 M4 Table 8

1 3 7 5

l Step 9 : Determination of the Consistency Ratio (CR) for each of the matrices, using the relation
CR = CI RI

(5)

Pair-wise comparison matrix for criterion 5 M1 M2 1/3 1 2 1/3 M3 1/7 1/2 1 1/3 M4 2 3 3 1

If CR < 0.1 (6) then the process is accepted. Else, all these steps are reiterated until a CR < 0.1 is reached. All these formulae [equations (3)-(5)] and condition [equation (6)] are adopted from the AHP developed by Saaty5. IE (I) JournalMC

M1 M2 M3 M4

1 3 7 1/2

142

Priority Vectors

l Step 10 : Calculating the composite priority value (SFMi ) for each of the alternatives. This is done for each alternative by multiplying its PV values for a particular criterion with the corresponding PV value of that criterion, and adding up these products. l Step 11 : Determination of the BMM for each of the alternative materials, as per the relation
BMM i = OFM i + ( 1 ) SFM i

0.5 0.4 0.3 0.2 0.1 0 M1 M2 M3 M4

Materials

Priority Vectors

where i = 1, 2, 3, . . . , n are the number of alternative materials and a is the objective factor decision weight [0 1 (usually the value of a is taken as 0.67)]; and SFMi , subjective factor measures, whose values are the respective composite priority values. l Step 12 : Ranking of the alternatives according to the descending order of their BMM values. l Step 13 : Selection of the best alternative according to the highest value of BMM. RESULTS FOUND FROM THE PRESENT METHODOLOGY The algorithm is coded as a FORTRAN program that operates under MS DOS. Tables 3 to 8 show the matrices for the criteria and the alternatives. Figures 16 show the PV values for each of these matrices. Finally, Table 9 shows the CP values and the corresponding Figure 7 shows these values graphically. The material M3 is seen to have the highest CP. Sensitivity Analysis In order to take into account the dependence of the decision variables (materials) on costs, a sensitivity analysis has been carried out based on the method given earlier. The governing
0.5 Priority Vectors

Figure 3 PV values for pair-wise comparison matrix for criterion 2 0.6 0.5 0.4 0.3 0.2 0.1 0 M1 M2 M3 M4

Materials

Figure 4 PV values for pair-wise comparison matrix for criterion 3

0.6 Priority Vectors 0.5 0.4 0.3 0.2 0.1 0 M1 M2 M3 Materials M4

Figure 5 PV values for pair-wise comparison matrix for criterion 4 0.6 Priority Vectors 0.5 0.4 0.3 0.2 0.1 0 M1 M2 M3 Materials M4

0.4 0.3 0.2 0.1 0 1 2 3 Criteria 4 5

Figure 1 PV values for decision matrix 0.6 0.5 0.4 0.3 0.2 0.1 0 M1 M2 M3 Materials M4

Figure 6 PV values for pair-wise comparison matrix for criterion 5

Composite Priorities

0.5 0.4 0.3 0.2 0.1 0 M1 M2 M3 Materials M4

Priority Vectors

Figure 2 PV values for pair-wise comparison matrix for criterion 1

Figure 7 Composite priorities for the materials

Vol 86, October 2005

143

Table 9

Composite priorities M1 M2 M3 M4 0.205 0.160 0.497 0.139

Table 10 Ranking of the materials Material M1 M2 M3 M4 0.50 0.45 0.40 MEMi Values 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 M2 M1 M4 M3 BMMi (based on a = 0.67) 0.194 0.220 0.392 0.193 Rank # 3 2 1 4

In the present methodology, the unit of OFC is Rupee, whereas OFM is non-dimensional. Correspondingly, the BMM values are also non-dimensional. Higher the BMM values, better would be the selection. The value of Objective Factor Decision Weight (a ) lies between 0 and 1. For a = 0, BMM = SFM, ie, selection solely depends upon the SFM values found from AHP, and SFM values depend dominate over OFM values. There is no significance of considering the cost factor components for a= 0. For a = 1, MEM = OFM, ie, OFM values dominate over the SFM values and the material selection is dependent upon OFM values only. For a = 1, the cost factors get priority over the other factors. Hence, the values of a are taken between 0 and 1. To verify the practicality and effectiveness of the final outcome of the present methodology, the sensitivity analysis is performed. CONCLUSION Selection of a belt material under sensitive operating environment is a very complex problem, when looked at from the point of optimisation. The present methodology is a very useful tool first to quantify the apparently intangible factor in a systematic manner and then to find out the best alternative depending upon their cost factors. As contrasted to the traditional procedure of selection using discounted cash flow (DCF), the present method is a sound surrogate to apply under unstructured environment. The practicability and effectiveness of the method is further enhanced by the sensitivity analysis followed at the end. REFERENCES
1. Maitra and Prasad. Handbook of Mechanical Design. Tata McGraw-Hill Publishing Company Ltd, New Delhi, 1995, p 10.87. 2. Sharma and Aggarwal. A Text Book of Machine Design. S K Kataria and Sons, Delhi, Ludhiana, 1991, p 642. 3. K D Sharma. Fundamentals of Machine Design. Asia Publishing House, 1969, pp 172-177. 4. J E Shigley. Mechanical Engineering Design. Mc-Graw Hill Book Company, 1986, p 621-627. 5. T L Saaty. The Analytical Hierarchy Process. Mc-Graw Hill, New York, 1980. 6. E S Buffa. Modern Production/Operations Management. Wiley Eastern Limited, New Delhi, 1993. 7. E S Buffa. Modern Production/Operations Management. Wiley Eastern Limited, New Delhi, 1993. 8. T L Saaty. How to make a Decision: The Analytical Hierarchy Process. European Journal of Operations Research, vol 48, no 1, 1990, pp 9-26. 9. T L Saaty, J W France and K R Valentine. Modelling the Graduate Business School Admissions Process. Socio-economic Planning Sciences, vol 25, no 2, 1999, pp 155-162.

Objective Factor Decision Weight, a Figure 8 Sensitivity analysis

equation for each of the alternative equation becomes, accordingly,


Yi = ( OFM i SFM i ) X i + SFM i

(7)

where Y is belt material measure (BMM); and X, the objective factor decision weight (a). This represents the equation of a straight line. The plots for the four materials are shown in Figure 8. Finally, the BMM values are calculated for a = 0.67, and the result presented in Table 10. Both the plots (Figure 8) and the Table 10 show that the bestsuited material is M3. DISCUSSION The number of criteria or alternatives should be reasonably small to allow consistent pair-wise comparisons. A maximum of seven criteria or alternatives can be used6. The methodology presented here can be further integrated with the Brown and Gibson model6 based on the multi-attribute preference theory7. For more than seven criteria or alternatives, one can group the criteria with respect to a common property and add another level to incorporate the groupings8,9.

144

IE (I) JournalMC

Vous aimerez peut-être aussi