Assessment of Process Capability: the case of Soft Drinks Processing Unit

The process capability studies have significant impact in investigating process variation which is important in achieving product quality characteristics. Its indices are to measure the inherent variability of a process and thus to improve the process performance radically. The main objective of this paper is to understand capability of the process being produced within specification of the soft drinks processing unit, a premier brands being marketed in India. A few selected critical parameters in soft drinks processing: concentration of gas volume, concentration of brix, torque of crock has been considered for this study. Assessed some relevant statistical parameters: short term capability, long term capability as a process capability indices perspective. For assessment we have used real time data of soft drinks bottling company which is located in state of Chhattisgarh, India. As our research output suggested reasons for variations in the process which is validated using ANOVA and also predicted Taguchi cost function, assessed also predicted waste monetarily this shall be used by organization for improving process parameters. This research work has substantially benefitted the organization in understanding the various variations of selected critical parameters for achieving zero rejection.


Introduction
Process capability studies are being incorporated as a successful quality improvement strategic tool in achieving customer satisfaction. Process capability analysis (PCA) and Six Sigma methodology occupy important places in quality and process improvement initiatives. As a fundamental technique in any production, quality and process improvement efforts, PCA is used to improve processes, products or services to achieve higher levels of customer satisfaction. Process refers to some unique combination of machine, tool, method, materials, and people engaged in production. It is often feasible and illuminating to separate and quantify the effect of the variables entering this combination. Capability refers to ability, based on tested performance; to achieve measurable result.Inherent capability refers to the product uniformity resulting from a process that is in a state of statistical control. The product is measured because product variation is the end result.

Significance of "capability of processes"
Process capability measurements allow us to summarize process capability in terms of meaningful percentages and metrics. To predict the extent to which the process will be able to hold tolerance or customer requirements. It helps you choose from among competing processes, the most appropriate one for meeting customers' expectation. Knowing the capability of processes, we can specify better the quality performance requirements for new machines, parts and processes. Process capability is the long-term performance level of the process after it has been brought under statistical control. In other words, process capability is the range over which the natural variation of the process occurs as determined by the system of common causes. Process capability is also the ability of the combination of people, machine, methods, material, and measurements to produce a product that will consistently meet the design requirements or customer expectation.

Research Objectives
i. To evaluate process control limits & actual process capability indices. ii. To assess long term and short term process capability for the process. iii.
To evaluate the Taguchi's loss function of the process for considered parameters i.e gas volume, brix and torque.

LITERATURE REVIEW
For assessing the process capability indices for soft drinks processing unit. The review considered the most cited academic publications covering various journals viz., quality engineering, communication in statistics-Simulation & Computation, Journal of statistical computation & simulation, Communication in statistics -Theory & methods, International journal of production research and Journal of applied statistics. Regarding process capability indices used in various processes, in manufacturing different products. The Boolean keyword combination "(Process Capability or Statistical processes, SPC, SQC,) AND (Soft drinks or food processing)" was applied to conduct the literature search. Keywords such as Process Capability, PCI, Soft drinks, Process capability studies in Indian Soft drinks, were used to search the databases.

Author (s) Findings/Significant Contribution
Chen et al., (2017) [1] An economic of line inspection/disposition approach is proposed, which incorporates manufacturing variation. Djauhari et al., (2016) [2] Method in designing reliable control charts, when observed sensitivity to the change in variance for small or moderate correlation is present, which provides a root, causes analysis of an out of control signal. Maman et al., (2016) [3] Observed the optimal sample size required to achieve a desired error of estimation using absolute percentage error of different Cp estimates. Saleh et al., (2016) [4] Designed the CUSUM chart, observed in-control ARL exceeds a desired value with a specified probability. Using a bootstrap-based design technique adjusted the control limits. Ganji & Gildeh (2016) [5] Introduced a new class of indices, Cp m (u,v)for the processes with asymmetric tolerance. Observed the relation between this index and the departure ratio of the process centering, as well as, the relation between this index and the upper bound of the percentage of non-conforming products. Ali

Author (s) Findings/Significant Contribution
Chen & Tsai (2012) [7] Density function of the natural estimator is derived. The expression is in terms of finite sum and only involves standard normal distribution. Four ensemble estimators for C pk are proposed. Hussein et al., (2012) [8] Proposed a sequential procedure for testing whether two processes are equally capable by using the PCI (Cpm). Niavarani et al., (2012) [9] Three indices referred to as NCpM, MCpM, and NMCPm in order to evaluate process capability in multivariate environment. Daniel Grau (2010) [10] New PCIs proposed based on empirical percentiles, asymmetry of the tolerances as well as the asymmetry of the process distribution has been considered. Perakis (2010) [11] The estimation of the difference between the values of process capability indices C pm or Cpmk for two processes is considered. Point estimators of such types of differences are defined, separately for each of the two indices, and their distributional properties are investigated. Parchami & Mashinchi (2010) [12] Similar to the traditional process capability indices (PCIs), we develop a fuzzy analogue by a distance defined on a fuzzy limit space and introduce PCIs, where instead of precise SLs we have two membership functions for upper and lower SLs. Jeang (2010) [13] Proposed process capability expression revised from the conventional Cpm in consideration of the balance between tolerance cost and quality loss has been developed. Wei Wu & Huang (2010) [14] Investigated the concept of generalized pivotal quantities to derive the generalized confidence intervals (GCI) for the capability ratio and the capability difference between two given suppliers. Albing & Vännman (2009) [15] Investigated a new class of process capability indices for target value 0. Two estimators of the proposed index are studied and the asymptotic distributions of these estimators are derived. Khadse & Shinde (2009) [16] Computing aspects of proposed PCIs are discussed for normal and non normal processes when process tolerance is symmetric as well as asymmetric. Lovelace & Swain (2009) [17] Proposed process capability index estimation methodology for Cp and Cpk for the case of non-normal, zero-bound process data using the delta distribution, a variant of the log normal distribution. Mannar & Ceglarek (2009) [18] Introduced a methodology for functional capability analysis and optimal process adjustment for products with failures that occur when design parameters and process variables are within tolerance limits (in-specs). Chen et al., (2008) [19] Evaluated the capabilities of multi process products together with nominal-thebest specifications, larger-the-better and smaller-the-better specifications. Proposed process capability analysis chart (PCAC/Cpm), process capability analysis chart (PCAC/Cpk) to consider process yield and expected process loss. Saxena & Singh (2006) [20] Studied a class of shrinkage estimators for Cp when a prior guessed value, the Bayesian estimation of Cp has been done under squared error loss function by assuming quasi-prior distributions of sigma and 1/sigma. Pearn et al., (2005) [21] Sensitivity investigation on process capability Cp and C pm in the presence of gauge measurement errors. Pal (2004) [22] Outlined the procedure using the generalized lambda distribution (GLD) curve for modelling a set of process data and for estimating percentile Points in order to compute generalized PCIs. Proposed a procedure that minimizes losses by adjusting the process mean in an economical way. Developed a program that allows the user to find the optimum mean value easily. Perakis & Xekalaki (2003) [24] New index that is a variant of Cpm is introduced proposed index performs satisfactorily for processes with symmetric or asymmetric specifications. Flaig (2002) [25] Presented a multiple objective approach to process capability optimization, that identifies the key input and output variables, and found a mean shift and variation change that optimizes profitability. Nam et al.,(2002) [26] Asymptotic variances of capability indices have been used to derive associated percentile-t boot strap confidence intervals and compared percentile-t bootstrap confidence intervals with Franklin & Wasserman' s standard. Richard et al.,(2002) [27] Proposed multistage process capability analysis algorithm demonstrated with two 2-stage industrial process examples and a 4-stage process example for its expandability.
Zhang & Ni (2002) [28] Introduced and investigated the new PCIs index that has direct quantitative association with the probability or the proportion of the conforming process output. Chen et al., (2001) [29] Developed one process capability analysis chart (PCAC) for precise measurement of an entire product composed of symmetric tolerances, asymmetric tolerances, larger-the-better and smaller-the-better characteristics. Hsin-Hung Wu & Swain (2001) [30] Presented two families of non-normal PCIs observed in the literature, to evaluate how the families of non-normal PCIs perform for non-Non processes using simulation. Niverthi & Dey (2000) [31] Examined the multivariate versions of the common process capability indices (PCI's) denoted by Cp, and C pk Markov chain Monte Carlo (MCMC) methods are used to generate sampling distributions for the various PCI's. Wah Lai & Chew (2000) [32] Studied gauge repeatability and reproducibility (GRR) applied a non parametric method. Asokan & Unnithan (1999) [33] Proposed a procedure for estimating the mean and standard deviation of a process from a sample of a lot truncated at its specifications. Wen & Mergen (1999) [34] Investigated finding the best location for the process mean for a situation where the process is stable but not capable of meeting the specification limits. Alan Veevers (1998) [35] The multivariate viability index V rn is defined, discussed and illustrated using an example from the minerals sector. Pearn (1998) [36] Introduced a new index C"pk , which is shown to be superior to the existing generalizations of Cpk investigated the statistical properties of the natural estimator of C"pk, assuming that the process is normally distributed. Chen (1997) [37] Study of asymptotic behavior the asymptotic distributions of the estimators of a rather wide class of PCI's are investigated in a unified approach. Levinson (1997) [38] Used the non central t distribution to get exact confidence limit. Also discussed simple approximations for the confidence limits. Pearn & Chang (1997) [39] Extended Wright's simulation study to cover some skewed distributions including chi-square, lognormal, and Weibull distributions for some parameter values. Found that the percentage bias of the estimator increases as the skewness coefficient increases.
Encapsulating the literature review it has been observed that researchers used different methodologies to understand/evaluate process capability indices. Some researchers modified existing methods, existing plotting techniques of control charts to interpret the process capability studies as well as various statistical parameters more clearly. To make more clear description, researchers focused on optimal sample size for error approximation, relation between index and departure ratio, capability indices in Bayesian view, PCIs based on empirical percentiles, point estimators of differences while PCI calculation, fuzzy analogue application for PCI, new methodology for functional capability analysis, sensitivity investigation on PCI, multistage process capability analysis algorithm, multivariate versions of common process capability indices, asymptotic behaviour and asymptotic distributions of the estimators and simulation studies on PCI.

SOFT DRINKS INDUSTRY IN INDIA
Pepsi and Coca-Cola are the leading major players in carbonated drinks in the Indian market. The Economic view of Indian Soft drink market best quoted as Duopoly by two major players Coke and Pepsi and both players having concrete monopoly power over the Indian Consumer. Soft drinks have a fairly high price elasticity of demand in which price and sales volume are have been balanced as a major performance indicator among producers. Pricing strategies of both companies is similar and try to gain market share by innovative promotional activities using celebrities as brand ambassadors. Estimated market of 284 million per year of the total soft drink (carbonated beverages and juices). The market follows one of the TCSR of forecasting model i.e is highly seasonal in nature with consumption varying from 25 million crates per month during peak season to 15 million during offseason. The buying behavior of urban population have significant role as major market is predominantly urban with 25 per cent contribution from rural areas. Mineral water market in India is 65 million crates. On an average, the monthly consumption is estimated at 4.9 million crates, which increase to 5.2 million during peak season.

The Indian Beverage Market
India's one billion people, growing middle class, and low per capita consumption of soft drinks made it a highly contested prize in the global CSD market in the early twenty-first century. Coke and Pepsi dominated the market and together had a consolidated market share above 95%. While soft drinks were once considered products only for the affluent, by 2003 91% of sales were made to the lower, middle and upper middle classes. Soft drink sales in India grew 76% between 1998 and 2002, from 5,670 million bottles to over 10,000 million and were expected to grow at least 10% per year through 2012. In spite of this growth, annual per capita consumption was only 6 bottles versus 17 in Pakistan, 73 in Thailand, 173 in the Philippines and 800 in the United States With its large population and low consumption, the rural market represented a significant opportunity for penetration and a critical battleground for market dominance. (FOBO) of premier soft drink (Gama) 2 available globally and also market leader of soft drink in India. This bottling company serves the market of Chhattisgarh as marketing execution strategy. ABCD only executes marketing strategies planned by Gama India for its operational area. It has limited freedom for promotional activities for which it has to get approval from Gama India. ABCD Ltd works on target basis it gets target by Coco Cola India (CCI) based on its previous performance in terms of cases to be sold and on that it received material for manufacturing the CCI beverages.Production Capacity: ABCD Ltd is presently producing 4 million cases (RGB + PET) annually. ABCD Ltd production capability can be viewed in the following

PET Blowing process & fıllıng of PET bottle flow
PET blowing process consists of thirteen steps as follows: Step: 1 preform issue from store Step :2 transfer of preform from store to blowing room Step :3 preform storage in blowing room or day storage Step :4 preform uploading in the hopper by the preform Step :5 filter from jumbo carton Step :6 preform through the elevator Step :7 preform conveyed through conveyor Step :8 heating of preform Step :9 blowing Machine (SIPA)[air pressure= 40nbars, temp = 90-110 0 c] Step :10 checking blown Bottle Specification Step :11 air Flow through Air filter Step :12 conveyed to filler Step :13 rinsing with 1-3ppm Cl2 in soft water & pressure = 1.5kg/cm² Filling process consists of nineteen steps as follows: Step: 1 Treated water intake in deareation tank Step: 2 syrup intake in the syrup tank of paramix Step: 3 mix inline through V-10 proportioning value Step: 4 beverage for chilling through PHE Step: 5 injection of CO2 Step: 6 beverage in the buffer tank Step: 7 proportioning of treated water, syrup & CO2 (chilling, carbonating, mixing temp. = 7 0 C) Step: 8 counter pressure filler Step: 9 filling Step: 10 closure -elevator closure -capping of filled bottle Step: 11 checking of filled bottle torque (5-17Psi) Step: 12 warmer temperature [38 -42•C] Step: 13 labeling Step: 14 data loading Step: 15 filled bottle inspection -removal of half filled and defect bottle Step: 16 corrugated pad -shrink wrapping [temp. 150-160•C] Step: 17 secondary coding on HDPE layer Step: 18 storage of finished product Step: 19 dispatch Through this process PET bottles will be ready for the filling of beverage. From figure 1. methodology adopted for current study in accomplishing the research objectives step by step with statistical inferences as well as operational views as statistical process control aspect.

RESULTS AND DISCUSSION
This section discusses first relevant stastical parameters for four soft drinks, ANOVA results for sample observations for four soft drinks with statistical inferences followed by relevent control charts Step: 1 data collection of four soft drinks for three parameters i.e gas volume, brix and torque, a sample size of 50 Step: 2 established control limits & plotted suitable control chart Step: 4 assessment of process capability indices: Short term & long term.
Step: 5 evaluated taguchi loss function for the process and assessed the monetary value of loss function for the process.
Step: 3analysis of results using ANOVA and suggested exploratory objectives for the processes. with interpretation is also dicsussed,and interpretation of assess ment of process capability indeces : Short term and long term finally evaluated taguchi loss function monetary values of four soft drinks for three technical parameters i.e gas volume, brix and torque are discussed.    figure 2. the process is stable and in control satisfactorily, it has been observed that twenty one points just above the Lower control limit and nine point are just below the upper control limit this signifies that probably process has process variation has been dramatically reduced might look like "too good to be true" or "do not touch the process" and remaining points are on central line which is desirable. In contrast we might see this pattern on process if one or various nozzles become clogged. It is important to analyze the desirability of such change.   Using statistical parameters i.e Upper Specification limit (USL), lower specification limit (LSL) process capability index (PCI) and Mean, from table. 3 plotted the control charts which are shown in figure 5, figure 6 and figure 7.   4. This is uncommon pattern. This can be caused by over adjustment. However, the process is under control as well as stable.  Using statistical parameters i.e Upper Specification limit (USL), lower specification limit (LSL) process Capability index (PCI) and Mean, from table. 4 plotted the control charts which are shown in figure 7, figure 8 and figure 9. in this particular case, when excluded the sample data points which are out of control and re-calculated the control limits in which he process is in control.   figure 10, ten points are on central line, ten points are just above central line, three points are out of upper control limit and one point is out of lower control limit, excluding the points that have fallen out of control re calculated the new control limits, resulted process is under control. This pattern might be indicative of a sudden increase in the process variation. It is possible to have some points in this zone from time to time, but two out of three consecutive points is not desirable. Using statistical parameters i.e Upper Specification limit (USL), lower specification limit (LSL) process Capability index (PCI) which is less than the standard level 1.0 and Mean, from table. 8 plotted the control charts which are shown in figure 10, figure 11 and figure 12.  figure 11. nine points are out of lower control limit thirty two points on central line, nine points are below upper control limit, recalculated the control limits by excluding the data points which are fallen out of control limits, which later process is in control. However, this might an indication of tool wear, poor machine maintenance, and tired operator. It does not represent a sudden change in the process, but a slight and continuous change in it. This kind of patter can be easily detected and acted on before it is too late.  figure 12. six points just near lower control limit, eight points are just near upper control limit, nineteen points above central line, eighteen points below central line. This pattern might be indicative of a sudden increase in the process variation. It is possible to have some points in this zone from time to time, but this kind of patter is not desirable in process however, process is under control.  figure 13. Twelve points are out of control limit, hence process is not in control, cause of data integrity either data collection to be done again for sample size and evaluate control limits for new data.  of assessment of long-term and short-term capability indices reach to conclude that if Standard deviation Sshort and Slong are equal the total process variability independent on time for the process considered in this study parameters for Limca S Short is equal to Slong . Equality Sshort = Slong involves the capacity of adequate short-term and long term capability indices i.e Cp = Pp and Cpk = Ppk. If there is an inequality Sshort < Slong which is observed in all cases can be interpreted as the total process variability is larger than the variability independent on time, so a position of the considered process was being changed significantly in time. from table 2 except for torque as a parameter for all four soft drinks where Sshort < Slong condition not satisfied. Thus inequality S short < Slong condition implies that following relations between shortterm and long-term capability indices : Cp > Pp and Cpk > Ppk . Thus the relationship between shortterm and long-term process capability indices Pp Cpk = Cp Ppk used as verification for the calculation correctness. Calculations done in current study satisfies this condition of correctness. [41]

Taguchi's loss function
As per quality guru Taguchi, monetary losses start as soon as process starts to shift away from the target value. Furthermore, taguchi mentioned that those monetary losses were experienced by society. As we move further away from target value, the monetary loss increases, following a quadratic function. Defined by the following formula L = k (y -T) 2 where L is monetary loss, k is a cost factor, y is actual value, T is target value. Quality loss function from figure.1 Blue line shows difference in loss level between traditional and Taguchi approach. L is always expressed as amount per part regardless of single or multiple unit of part information. If there is more than one part involved then instead of (Y -Y0) 2 Table 9. Loss Function (L) for parameters of four Soft drinks

SUMMARY:
Implications with reference to processing unit management it can be acknowledged that the process meets the expectations regarding the analyzed parameters i.e gas volume, brix, torque for four soft drinks. It is worth noticing that some parameters for some soft drinks not falling within the standard limits of process capability index especially Fanta's PCI is less than 1 which is not desirable. Similarly, Thumpsup's and Fanta's : gas volume, torque processes are out of control, excluding the points which are fallen out of control and recalculated the new control limits resulted the process in control which is an exploratory view of this study. From methodological point of view the assessmnet of a long-term and short term process capability is a single factor analysis of variance. The aim of doing that analysis is to understand and infer the sensitivity of data, observations made are statistically acceptable for further courses of managerial, operational and analytical actions as collected data observed the normal distribution and variance equality. The causes for variation are broadly classified into seven types: equipment, indirect material, direct material, orders, people, methods and working conditions. Implication reference to assessment of short term and long term capability an inequality Sshort < Slong is observed in all cases the total process variability is larger than the variability independent on time, so a position of the considered process was being changed significantly in time. Except for torque, as a parameter for all four soft drinks where Sshort < Slong condition not satisfied. Thus, torque process parameter for all four soft drinks has to be reconsidered to set the new targets for the processes to achieve the desired level of process capability indices. Taguchi's loss function has its own significance in any processing unit which deals with quality as a crux aiming to satisfy customers as well as producers. This study also observed that certain parameters has not been set (targeting mean) exactly in which there is a loss to producer. The loss has been projected monetarily for all selected parameters in the process, for four soft drinks for four different conditions. The loss function projected in this study helps the management in understanding/interpreting the process parameters which are not in within the specified limit which thus research study conducted an exploratory view to management of processing unit.