Prioritization of Attributes for Palletizing Robots in Beverage Industry of Pakistan

Robots are extensively used in modern manufacturing industries to perform numerous repetitive operations. The challenge of selecting the most appropriate robot for a particular manufacturing setup is progressively becoming complex as there are numerous selection criteria and more alternatives available in market. Only a limited amount of research is available in literature which focuses on the selection of industrial robots for beverage industry. This study offers a country specific application of AHP (Analytical Hierarchy Process) in problem of palletizing robot selection for beverage industry of Pakistan. The problem is structured in standard AHP hierarchy and equations. The factors initially explored from the concerned literature are prioritized by the industry experts. The available robot alternatives are evaluated for each parameter and results are computed with the help of Exert Choice, a commercial AHP software. It is observed that experts in Pakistan beverage industry are very sensitive to operating costs of the robots and they do not assign as much weightage to technical parameters like repeatability and programmability. The robots with lesser associated costs and better speed and ‘manipulator reach’ are higher in ranking. The findings are beneficial for the international investors and local beverage industry managers to corroborate the current trends and preferences of the said industry.

MCDM models are applied only if there are multiple influencing parameters and multiple alternatives [1]. These methods are either compensatory or non-compensatory.
In non-compensatory methods, there are no mutual flexibilities among the parameters whereas compensatory methods permit some trade-off's [2,3]. Some of the common characteristics which are shared among all MCDM methodologies include multiple objectives, conflicting criteria, different measuring units for the attributes and alternative rankings [4]. Among MCDM techniques, AHP is used most widely. It involves pairwise comparisons of attributes and alternatives which are accomplished by experts of the concerned industry [5].
Despite the fact that robot selection is a complex, cost intensive and irreversible MCDM decision, most of the studies reported in available literature present different other techniques like performance optimization and statistical methods for decisions of industrial robot selection. Only a few researchers utilized some MCDM models. These models normally include both subjective and objective types of attributes [6].
In addition to AHP, some authors have also applied other MCDM techniques like DEA (Data Envelopment Analysis) [7], TOPSIS [8] and graphical methods [9] etc. Recently researchers have started utilizing hybrid methodologies by combining multiple techniques like fuzzy TOPSIS [10], fuzzy-QFD (Quality Function Deployment) [11] and fuzzy-AHP [12]. Most of the research on robots covers their design and planning aspects [13,14]. There are hardly any studies found in literature which emphasize on MCDM perspective of beverage industry palletizing robots particularly in current industrial perspective of Pakistan.
One problem with most of the MCDM techniques is that the data collection from experts is vague and substantial computational efforts are involved. In MCDM problems, the expert opinion is normally required from executive managers who have very busy schedules. AHP requires managers to compare only two parameters at a time which is very convenient and straight forward for these senior professionals. Moreover, AHP is backed by authentic mathematical models founded on matrix algebra which are developed, modified and implemented in other similar fields by scientists and researchers [3,15,16]. The current paper, therefore, utilizes the AHP modelling for selection of palletizing robot in beverage industry of Pakistan.
Next section of this paper offers the general mathematical models constituted for this robot selection problem. The models are simulated in section 3. Quantitative results are presented and discussed in section 4 followed by the concluding section.

THE AHP MODELLING
The standard AHP model presented by Saaty [5] is used in this study to model the problem for selection of suitable robot. After defining our problem next step involves the [ ] (1) The calculation of the weights matrix is the next logical step. Weights can be calculated using the eigenvalue method. This method was used by Saaty for determination of weights vector [5]. To calculate these weights, we first need to normalize matrix A using Equation (2).
where 'i' and 'j' are 1, 2, 3, . . ., n After normalizing the comparison matrix 'A', final priority weights are calculated using Equation (3).  A standard AHP questionnaire adapted fromTomar, and Borad [19] has been developed for pairwise comparison of factors using Satay's scale of 1-9 and presented to the experts of beverage industry. A sample from actual questionnaire is presented in Appendix-I. Researchers in the area of AHP preferably take a small sample size of experts as it has more practical worth than a large sample size [20,21]. This is because of the fact that chances of getting arbitrary and inconsistent answers are increased if a large number of pairwise comparisons are collected, thus leading to mathematically unrealistic result [22].
Unlike statistical techniques, AHP normally does not prefer a large sample size from a big population rather it prefers a smaller one. However, the selected experts must have concrete knowledge and expertise in the chosen field of study. Most of the researchers in similar fields, therefore, normally keep the number of experts lesser than ten. For instance, there were a total of ten experts in studies performed by Cheng and Li [20] and Wong and Li [22] while eight experts were approached by Lam and Zhao [21]. A sample size of three experts from a single industry has been used by Tahriri et al. [23]. In our current research, six top level managers with strong technical and managerial expertise in concerned field from a well-known   e  l  e  e  n  o  g  n  i  r  o  v  a  f  e  c  n  e  d  i  v  e  e  h  T  n  o  i  t  a  m  r  i  f  f  a  f  o  r  e  d  r  o  e  l  b  i  s  s  o  p  t  s  e  h  g  i  h  s  e  u  l  a  v  e  t  a  i  d  e  m  r  e  t  n  i  s  s  e  r  p  x  e  o  t  d  e  s  u  e  b  n  a  c  8  ,  6  ,  4  ,  2 based on decision makers' preferences of all the factors are thus calculated using the mathematical models presented in previous section.
Keeping in view the minimum selection constraints of the problem, seven robots manufactured by different international firms have been selected as candidate alternatives. All of the alternatives are specifically designed for performing pick and place (palletizing) operation. Makes and Models of these robot alternatives are provided in Appendix-II. Quantitative data for these alternatives is collected from the concerned manufacturers.
Data for the subjective factor "Programmability" has been collected from industry experts. The weights for this subjective factor are calculated using standard pairwise comparison matrices. The data for objective factors is summarized in Table 1.