Process Capability Analysis And Assessment Faced On Continual Quality Improvement

Continual improvement of all processes of the organization is the core of modern quality management philosophy. Process Capability Analysis and Assessment is an important step to improve process quality. By analyzing and assessing process capability, the process status can be realized, the aspect can be determined and the improving results can be assessed, to achieve the goal of decreasing, restrain the variation. Based on the Distinguished Young Scholars "Quality Management" (70125004), the projects "The research on realizing theory and technology from basic standard system to advanced standard system" (70072029) and "The realizing technology and performance assessment of quality control in time-variety discrete process" (70572050) sponsored by National Natural Science Foundation, a deep research is made in the dissertation in the relevant content to process capability. The main work and achievements are as follows:1. Theory base and statistical character of Process Capability IndexBased on new and old quality loss functions, the definition and theoretical foundation of several process capability indices (PCI) are discussed and their relation, difference and characteristic related to the distribution are comprised. It is pointed out that the different choice of PCI shows different quality philosophy and quality goal --- contract request or excellent quality. The concept of quality loss index is initiated. In allusion to being difficult to make sure the lower confidence limit without conforming the distribution form in many practical process outputs, some non-parameter, computerized Bootstrap analysis methods is used and the relevant MATLAB programs are compiled. With C_pk as a sample, the reliability of different Bootstrap lower confidence limits are research by a great deal of simulations with different distribution, parameter and sample size. The best Bootstrap method is confirmed in case of different sample size. The result shows that the PTB method fits better to most distributions. The quality managers always focus to an important character of PCI, the relations between the conformance proportion and PCIs. The relation between Taguchi PCIs (C_pm and C_pmk) and conformance proportion are lucubrated. Their function formula and diagram are set up to promote the using of Taguchi indices in practice. 2. Process capability analysis (PCA) and assessment for skew processesThe common PCA was set up on the assumption that process output is normal. It is difficult to implement PCA in the processes with skew distribution. On the basis of summing up PCA methods in the processes with skew distribution, the study was made in two aspects. First is that the Box-Cox power transformation model is used in data transformation of process capability analysis in allusion to many non-negative zero-bounded processes in chemic, pharmacy etc industries. The results of simulation using Lognormal distributions with different parameters show that the method have higher applicability and reliability to the distribution with different skewness and kurtosis. Because of the bulk of VVV PCI, a new SWV PCI is set up as a common skew PCI based on scaled weighted variance (SWV) method and Clement skew PCI philosophy. The index is validated by the result of simulation using 29 different combinations with different skew distributions and parameters.3, Process Capability Analysis in case of time-variety processesOn a background of variety and automatization of production technology, time-variety exists in more process outputs. The time-variety in PCA has been more focus on by quality experts. The two cases are investigated. The periodic adjusted time-variety processes are investigated. The characteristic of compound distribution is researched while the process output drift in linear mode. The proportion density function system, whose variable is adjusted coefficient, and its standard form of the distribution are set up. New analysis method and formula'are initiated to adapt the case. Because the development of auotcontrol technology in production leads to more processes output is autocorrelated, the effect of autocorrelation on PCA was studied through the simulation, based on stable stochastic models in time-series analysis theory. The result shows that the sampling method, sub-sample size and deviation estimating methods of standard statistical process control method have great effect on process capability in the autocorrelated processes. The correlation should be analyzed before PCA. The process capability analysis approach method is initiated.4. The input's effect on process capability analysisThe effect of process input isn't taken into account in the usual process capability analysis. But the variation of process input is transferred to output in some manner, the effect of input is introduced into process capability analysis. By analyzing the relation between input and output of process, the linear and non-linear (exponential) process,, variation transfer models were developed and the effects of input on process capability was studied. The distribution of the input's effect on process capability and variance transfer was researched in first time when there is non-linear relation between input and output of process. The result indicates that the non-linear relation makes distribution is not normal but heavy tail. The simulation result shows the relation between the variance transferred to output and model coefficient is linear for any input variance. So, the relation of input and output should be analyzed in PCA.5. The extension from process capability assessment to process performance evaluationThe process capability is a quality concept in narrow sense and process performance is quality concept in broad sense. For better to implement the continual process improvement, the process capability assessment is extended to process performance synthetically evaluation, including cost, productivity and quality. The evaluation introduced the benchmarking of multi-processes by different models of data envelopment analysis (DEA). Based on the efficiency value and slack variable of sorted DEA model, the approach for improving and adjusting the inputs and outputs is discussed and illustrated through a practical case in mechanical industry.