Computational Intelligence in Industry - Multivariate Statistical Process Control - Research Proposal Example

Computational Intelligence in Industry: Multivariate Statistical Process Control  A Research Proposal Submitted by: HERE December 3, 1. Abstract The focal point of basically all activities of a private company is to preserve their competitiveness in the business in terms of quality, delivery and cost. A company overtakes its competitors only if their operations are conducted in such a way that variances in the output are held to a harmless minimum and problems are anticipated before these results in production lags. One of the best tools in the production line and process operations is multi-variate statistical process control (MSPC) because most processes involve not one but a number of interrelated variables, which should be monitored as a group to sustain the efficiency of a process. The proposed study will attempt to examine the extent of usage of MSPC, as well as the levels of effectiveness and computational intelligence in MSPC among selected companies, en route to the design of a MSPC-based intervention for one process of a selected company with low computational intelligence. A combination of quantitative and qualitative methodologies will be adopted, particularly the survey and observations techniques. Purposive sampling will be used in the selection of institutional participants based on a set of inclusion criteria. Descriptive and inferential statistical measures will be used in the analysis of the data collected. Findings of the descriptive data analysis will form the basis for the selection of a company whose operational process will be analyzed to formulate a MSPC-based intervention. 2. Introduction Statistical process control (SPC) is a technique which “involves the use of various methods to measure and analyse a process” (Groover, 2008, p. 594). Oakland (2008), however, clarified that SPC is neither about statistics nor control, but about competitiveness. Regardless of the type of organisation, competition is focused on three primary issues: quality, delivery and price. Statistical process control, therefore, finds varied applications in both manufacturing and non-manufacturing sectors. The general objectives of SPC are: (1) improvement of the quality of the output of a process; (2) reduction of the variance in a process to establish process stability; and (3) address problems associated with processing operations. The seven most common tools of SPC includes cause and effect diagrams, check sheets, control charts, defect concentration diagrams, histograms, Pareto charts, and scatter diagrams (Groover, 2008). Traditional SPC is univariate, such that mostly the tools are utilised to distinguish unusual changes in independent variables or those which are not affected by the behaviour of other variables in the process. A considerable number of industries, however, rely on processes based on a set of interrelated variables, also termed as multi-variate. Mason and Young (2002) maintained that multi-variate processes are now very common in many industrial settings. These multi-variate variables do not behave independently of each other and should, therefore, be investigated as a group. Hence, familiarity with the tools of multi-variate SPC will be instrumental in maintaining the smooth operation of industrial processes and in avoiding bottlenecks which cause delays in the operation. 1.1. Statement of the Problem This study will venture to examine the extent of usage and effectiveness of multi-variate statistical process control in the operation of various manufacturing and non-manufacturing processes in selected localities. Specifically, the following research problems will be addressed in the study: 1.1.1. What is the profile of the participating institutions in this study in terms of: 1.1.1.1. organisational size, 1.1.1.2. type of industry, and 1.1.1.3. processes operated? 1.1.2. What is the extent of usage if multi-variate statistical process control in these participating institutions? 1.1.3. How effective is multi-variate statistical process control in minimising variance in the operation of the various processes in the participating institutions? 1.1.4. What is the level of computational intelligence of the participating institutions in terms of multi-variate statistical process control? 1.1.5. Are there significant differences in the extent of usage, level of effectiveness and level of computational intelligence of multi-variate statistical process control in the participating institutions in terms of the profile variables considered in this study? 1.1.6. How can at least one of the institutions found to have a low level of computational intelligence in multi-variate statistical process control (MSPC) benefit from the technique? What MSPC intervention may be proposed. 1.2. The Research Being Proposed The research study being proposed is both a descriptive and empirical study which will utilise survey and observation methodologies to answer the issues posed under the problem statement. Based on the responses to researcher constructed survey questionnaires and corresponding observation of industrial processes, a generalization about the extent of usage of MSPC, the level of effectiveness of MSPC and the level of computational intelligence on MSPC of the participating institutions will be formulated. Based on the findings, at least one institution which has a very low level of computational intelligence on MSPC will be selected and a corresponding MSPC procedure for one process will be tailor-fitted for this institution. 3. Background  2.1. Review of Related Literature Statistical process control (SPC) refers to “the use statistical tools and techniques to measure a production process in order to detect change” (Webber and Wallace, 2007, p. 156). Beri (2010) maintained that SPC deals with the application of suitable statistical tools to a process with the objective of ensuring the continuous improvement of the quality of products, services and productivity in the workforce. Statistical process control is just one of the many facets of computational intelligence in industry. Kordon (2009) delineated a multiplicity of reasons for the necessity of intelligent solutions in business, such as to beat competition, accelerate innovation, produce efficiently, distribute effectively, enhance creativity, impress clients, attract investors, etc. The specific objectives of this literature review are: (1) to clarify and focus the research question for the present study; (2) to observe the methodology used by other researchers in the field of SPC in order to gain insight on the most appropriate method to use in the present study; and (3) to present empirical support for the findings of the present study based on existing knowledge about the applications of statistical process control in various industries. 2.1.1. Applications of statistical process control in industry The Mago, Santoso, and McGranaghan (2008) study aims to foster usage of steady state data and statistical process control in the appraisal of feeder voltage regulation. Mago, et al. (2008) enhanced the control charts methods proposed by Zhang and Yu (2000) by developing beta risk and average run length curves to rid the outcome of false positives. The statistical analysis algorithm formulated by Mago, et al. (2008) adopted steady-state 15-minute rms voltage data as input. The researchers maintained consistent data for the analysis by scaling to a 120-V level and setting the mean of the input data is set to zero. The control chart analysis method of appraising feeder voltage regulation, whose special features include upper and lower control limits and a centerline denoting the mean of the data, was formulated to distinguish whether a registered variation in trend resulted from conventional causes or other causes which may necessitate immediate action. The 0.001 probability was set as the threshold of the upper and lower control limits, such that only two of every 1,000 data points will surpass either limit as a consequence of random causes. An important assumption made in this methods is that “the 0.001 probability limits lie very close to three times the within sample standard deviation” (p. 381). In this method, the X-bar, T-chart and S-chart represented control charts for normally distributed and continuously measured data. In each of the three aforementioned charts, control of data variability is undertaken by plotting sample means, sample ranges and within sample standard deviations. The second method in the examination of feeder voltage regulation is called run chart analysis. This method utilizes an algorithm which measures the performance of voltage regulation grounded on the trending of data points situated outside of the threshold limits. Steady state rms data are collected from power quality monitors. Raw data are then processed by setting the mean of the data points to zero, as well as the bandwidth to  1. Analysis of data involves segmentation of the run chart and categorization of the trend, and finally analyzing the trend. Findings of the Mago, et al. (2008) study revealed that statistical process control methods can be utilized as one option to perform trend analysis for the assessment feeder voltage regulation. Using any of the two proposed methods, issues pertaining to poor regulation, regulator problems and other related problems can be detected. Meanwhile, Hossain, Choudhury and Suyut (1996) exhibited the utility of SPC in industrial control on real-time process using the following: process control station, software for SPC; and interfacing hardware. The initial step involved stabilization of the level control loop of the process control station, which run for a substantial period of time prior to collection of data. Sample data from the process were gathered in real-time using Paragon 500, the SPC software tool for analysis. The output from the analysis was presented in the form of X-bar and R charts, histograms and scatter plots. In a nutshell, in the experiments conducted by Hossain, et al. (1996), data for the SPC block was the tank level. Upper and lower specification limits for the tank level were assigned at 50.7500 and 49.2500, deemed to be reasonable for control of the upper and lower tank levels. Using a pump control system, instability in the tank level was introduced. Several problems were encountered during the experiments, including mal-adjusted pump control system, and disturbance in the measured tank level, and the causes of these irregularities were detected through the SPC results. Findings from Mago, et al. (2008) and Hossain, et al. (1996) substantiated the utility of statistical process control in trending analysis from two different fields – power engineering and water distribution. Thor, et al. (2007), Gerard, et al. (2009), and Robinson (2007) unveiled indispensable use of statistical process control in health care improvement, medical physics, and operations research, respectively. 2.1.2. Multivariate statistical process control The study of Senouci, Bendaoud, Tilmatine, Medles, Das, and Dascalescu as a multivariate SPC methodology is presented. In the first phase, Senouci, et al. (2009) attempted to highlight the efficacy of SPC methods in monitoring the results of electronic separation processes. The two phases under control charting using multivariate SPC were presented: (1) analysis of the initial dataset from the process assumed to be in control; and (2) monitoring of the process to check whether or not the process remains in control. In the study, the Hotelling T2 statistic from the first phase measures the covariance of the multivariate normal distribution from the dataset, as apposed to variance which quantifies the change in only one variable. The determinant of the covariance matrix represented as S is designated as the process variance. T2 values for 12 subgroups of data are represented n T2 chart. In the Senouci, et al. (2009) study, the upper and lower control limits are, respectively 0.05 and 3. All data on this chart are above 0.05 and below 3, which implies that the process is in control. The multivariate control charts are then utilized to check if the process is continuously under control. Efficiency of the control charts is tested to discover out-of-control readings by way of two experiments on electrostatic separation in non-standard conditions. Reported outcomes in terms of the applied voltage were reduced from 32 to 30 kV. Based on the method of elliptical control region, an out-of-control state was observed in the 13th experiment from the Hotelling T2 chart, but not in the S chart. It was also found that the mass of middling product triggered the out-of-control event. Based on these experiments, it was found that multivariate control charts can handle correlations between output variables in complex electrostatic separation processes and that these charts can be utilized to monitor the overall performance of the electrostatic separation process and detect out-of-control states. Martin, Morris, and Zhang (1996) demonstrated other applications of multivariate SPC and its non-linear extension for process monitoring. In the first application, changes in raw material, energy supply and internal chemical changes were addressed so that the process operation and the resulting material will not be affected. At least six different regions of operation were evident. Based on the accumulated scores for non-linear principal component, it was recognized that there was a change in the process operating conditions during the first period of the operation being monitored. When the processing conditions were changed back to the plant operating conditions, it was observed that the process eventually moved to the more acceptable region of operation near the so-called original nominal region. In the second application study, a non-linear principal component model was developed to localize four different fault situations: reactor fouling, reactive impurity, solvent problems and combined reactor fouling and reactive impurities. It may be recalled that when linear principal analysis was used, the faults can not be distinguished from the scores. Results indicated that score movement in the north-east direction are the fouling faults, whereas those moving in the north-west direction are combined impurity and fouling faults, whereas, score movement in the south-west direction point to combined impurity and solvent problems. The two application studies by Martin, et al. (1996) showed that multivariate SPC, combined with plant performance monitoring enhances product quality and consistency, reduces plant down-time and rework requirement, as well as discovery of process malfunction and increase usage of resources. Findings of the Martin, et al. (1996) study corroborate the results of Senouci, et al. (2009) that multivariate statistical process control is an indispensable tool in the monitoring of the performance of industrial processes in general. Additionally, Kano and Nakagawa (2008), Kittiwachana, et al. (2008), and Kourti (2006) also found good use of multivariate process control in the steel industry, liquid chromatography and in the pharmaceutical industry, respectively. Statistical process control (SPC) has been around since the 1920s and it can be safely generalized that considerable research on the field has been made since then. The studies of Hossain, et al. (1996) and Mago, et al. (2008) and those of Martin, et al. (1996) and Senouci, et al. (2009) which were 12 years and 13 years apart, respectively, showed evidence that SPC had not lost its vaunted efficacy in monitoring industrial processes, However, the technical quality of the Mago, et al. (2008) and the Senouci, et al. (2009) works illustrated that advances in computer technology further reinforced the usefulness of SPC as a significant part of corporate strategy to minimize and control variability in their operations. 4. Proposed Work 4.1. Statement of the Aims/Objectives The aims of the proposed study are: 4.1.1. Undertake a descriptive and empirical study on multi-variate statistical process control among selected manufacturing/non-manufacturing institutions in a chosen locality in the US. 4.1.2. Use purposive sampling to select participant institutions in the study based on a set of inclusion criteria. 4.1.3. Design a survey questionnaire to inquire about the profile, extent of usage of MSPC, effectiveness of MSPC in the institutions’ operations, and the level of computational intelligence on MSPC. 4.1.4. Administer the questionnaire after the appropriate permissions are granted. 4.1.5. Observe the processes in the various participating institutions and record observations regarding bottlenecks, variances noted on the flow of operations or in the finished product, problems which are encountered during operation, and how MSPC is applied in the operations. 4.1.6. Analyse the data collected using descriptive and inferential measures. 4.1.7. Present the findings and formulate conclusions. 4.1.8. Select an institution with very low computational intelligence in MSPC and design a MSPC-based intervention suitable for the institution. 4.2. Hypotheses The following hypotheses, stated in the null form, will be tested using two-tailed analysis and a level of significance of 0.05 (α=0.05): 4.2.1. There are no significant differences in the extent of usage of MSPC in the participating institutions in terms of organisational size, type of industry, and processes operated. 4.2.2. There are no significant differences in the effectiveness of MSPC in the operations of the participating institutions in terms of organisational size, type of industry, and processes operated. 4.2.3. There are no significant differences in the level of computational intelligence on MSPC in the participating institutions in terms of organisational size, type of industry, and processes operated. 4.3. Methodology The following methodology will be adopted in the conduct of this study: 4.3.1. The study will adopt the descriptive method of research using a combination of quantitative and qualitative methods, particularly the survey and observation techniques. 4.3.2. Purposive sampling will be used in the selection of participating institutions. The inclusion criteria set for the selection are as follows: 4.3.2.1. The institution is a private company licensed to operate in the research locale. 4.3.2.2. The management of the company expressly agrees to allow access to at least one of its process operation within a specified day or time to conduct the study and to administer the survey questionnaire to the operations manager or his equivalent. 4.3.2.3. The raw materials and the product output from the process are not toxic/hazardous chemicals and the operation does not emit toxic fumes. 4.3.3. The survey questionnaire will be designed by the researcher and validated for face validity and internal consistency. Interpretation scales will be researcher-devised based on the normal distribution. 4.3.4 The survey questionnaire will be pilot tested with at least 5 respondents from comparable companies in a neighbouring locality within the same state. The result of this pilot testing will be discussed under the preliminary results. 4.3.5 For the qualitative strand of the study, an observation chart will be prepared for use in the recording of observed bottlenecks, variances noted on the flow of operations or in the finished product, problems which are encountered during operation, and how MSPC is applied in the operations. 4.3.6 Descriptive statistics, particularly, frequency distribution, mean and standard deviation will be used to present the findings of the study with respect to profile, extent of usage of MSPC, effectiveness of MSPC in the institutions’ operations, and the level of computational intelligence on MSPC. 4.3.6 Inferential statistics , particularly, analysis of variance and applicable post hoc analysis will be used to present the findings of the study with respect to significant differences among the companies regarding extent of usage of MSPC, effectiveness of MSPC in the institutions’ operations, and the level of computational intelligence on MSPC in terms of the profile variables considered in this study. 4.3.7 Based on the findings on the level of computational intelligence on MSPC, one company will be selected and provided a MSPC procedure designed specially for one of their operations. 4.4. Preliminary Results The survey questionnaire and the observation charts are currently reviewed by a panel of experts composed of colleagues from at least four companies utilising MSPC in their operations. These instruments will be enhanced based on the recommendations of the panel of experts, and will be presented later to the faculty supervisor for approval and / or comments. The instruments are not yet appended with this proposal. 5. Summary  5.1. Significance of the Research The output from this proposed study will be significant to the following groups: 5.1.1. Operations managers. Findings of the proposed study will be useful for operations managers in as far as effectiveness of multi-variate statistical process control for various industries are concerned. Positive outcomes with respect to effectiveness of the technique will send out signals to operations managers and other personnel charged with the monitoring and maintenance of company processes to adopt or continue harnessing the benefits of MSPC to industrial processes. 5.1.2. Management. As earlier posited, statistical process control is all about competitiveness. Results of this study will, therefore, be meaningful for managers who would like to be presented with first hand evidence that, indeed, MPSC plays an important role in keeping the quality of products, delivery of services and prices of products and / or services competitive. 5.1.3. Industrial engineers. Industrial engineers will also be interested in the output of this study with respect to the tailor-fitting of MSPC procedure for a specific industrial process. Since design of the MSPC procedure will be fully documented in this study, industrial engineers can use the documentation as a guide or example in formulating a MSPC technique for their own operations. Thus, the benefits of using MSPC in monitoring and maintaining smooth and balance flow of operations free from bottlenecks will be extended to other firms. 5.2. Original Contribution of the Research This study is a fresh and novel idea different from most studies already made about statistical process control, which are mainly mathematical and heavily dependent on presenting algorithms in a laboratory controlled environment. The original contribution of this research lies in the novelty of profiling how multi-variate statistical process control are actually used in industry in terms of the extent of its usage, level of effectiveness perceived by knowledgeable professionals in their use of MSPC, and level of computational intelligence on MSPC in real process operation. In majority of the published researches on SPC surveyed during the literature review, the take-off point of the studies was concentrated on gaps in existing knowledge from literature. Another original contribution of this study, therefore, is its use of profiling as a springboard for proposing a MSPC-based intervention. Based on the proposed profiling activity, which will serve as an informal needs analysis, the research will then select one process from a company of very low computational intelligence on MSPC and design a proposed intervention based on the actual operations of a specific process in the company. Finally, the fact that the proposed research will be anchored on the convergence of quantitative and qualitative methods presents a third original contribution of the proposed study. Practically all studies made on multi-variate statistical process control are heavily quantitative. Not that the existing purely quantitative studies lacked in substance, but as far as originality is concerned, the proposed study will deviate from the usual research on statistical process control because it will venture on the benefits of the hybrid technique of quantitative-qualitative approach. References  Beri, G. C., 2010. Business statistics. 3rd ed. New Delhi (IND): Tata McGraw Hill Education. Gerard, K., Grandhaye, J. P., Marchesi, V., Kafrouni, H., Husson, F. and Aletti, P., 2009. A comprehensive analysis of the IMRT dose delivery process using statistical process control. Medical Physics, 36(4), pp. 1275-1285. Groover, M. P., 2008. Automation, production systems, and computer-integrated manufacturing. 3rd ed. Upper Saddle River (NJ): Pearson Education. Hossain, A., Choudhury, Z. A., and Suyut, S., 1996. Statistical process control of an industrial process in real time. IEEE Transactions on Industry Applications, 32(2), pp. 243-249. Kano M. and Nakagawa, Y., 2008. Data-based process monitoring, process control, and quality improvement: Recent development and applications in steel industry. Computers and Chemical Engineering, 32(1,2), pp. 12-24. Kittiwachana, S., Ferrerira, D. L. S., Fido, L. A., Thompson, D. R., Escott, R. E. A., and Brereton, R. G., 2008. Dynamic analysis of on-line high-performance liquid chromatography for multivariate statistical process control. Journal of Chromatography A, 1213(2), pp. 130-144. Kordon, A. K., 2010. Applying computational intelligence: how to create value. Heidelberg (DEU): Springer-Verlag. Kourti, T., 2006. The process analytical technology initiative and multivariate process analysis, monitoring and control. Analytical and Bioanalytical Chemistry, 384(5), pp. 1043-1048. Mago, N. V., Santoso, S. and McGranaghan, M. F., 2008. Assessment of feeder voltage regulation using statistical process control methods. IEEE Transactions on Power Delivery, 23(1), pp. 380-388. Martin, E. B., Morris, A. J., and Zhang, J., 1996. Process performance monitoring using multivariate statistical process control: systems engineering for automation. IEE Proceedings – Control Theory Applications, 143(2), pp. 134-144. Mason, R. L. and Young, J. C., 2002. Multivariate statistical process control with industrial applications. Philadelphia, PA: American Statistical Association, Society for Industrial and Applied Mathematics. Oakland, J., 2008. Statistical process control (6th ed.). Oxford: Butterworth-Heinemann Robinson, S., 2007. A statistical process control approach to selecting a warm-up period for a discrete event simulation. European Journal of Operational Research, 176(1), pp. 332-346. Senouci, K., Bendaoud, A., Tilmatine, A., Medles, K., Das, S., and Dascalescu, L., 2009. Multivariate statistical process control of electrostatic separation process. IEEE Transactions on Industry Applications, 45(3), pp. 1079-1085. Thor, J., Lundberg, J., Ask, J., Ollson, J., Carli, C., Harenstam, K. P., et al. 2007. Application of statistical process control in healthcare improvement: Systematic review. Quality and Safety in Health Care, 16(5), pp. 387-399. Webber, L. and Wallace, M., 2007. Quality control for dummies. Hoboken (NJ): Wiley Publishing. Bibliography Barron, L., 2006. Literature review. In: V. Jupp, ed. The Sage dictionary of social research methods. London: Sage Publications, pp. 162-162. Boudreau, M. A. and McMillan, G. K., 2007. New directions in bioprocess modelling and control: Maximizing process analytical technology benefits. Research Training Park (NC): Instrumentation, Systems and Automation Society. Kell, D., 2009. Recycling and recovery. In: R. E. Hester and R. M. Harrison, eds. Electronic waste management: Design, analysis and application. Cambridge: Royal Society of Chemistry, pp. 91-110. Naikan, V. N. A., 2008. Statistical process control. In: K. B. Misra, ed. Handbook of performability engineering. London: Springer-Verlag, pp. 187-202. Oakland, J. S., 2008. Statistical process control. 6th ed. Oxford: Butterworth-Heinemann. Rauwendaal, C., 2008. Statistical process control in injection molding and extrusion. 2nd ed. Cincinnati (OH): Hanser Gardner Publications. Read More