ABSTRACT
Chapter 12 considers the most suitable approach to process trouble-shooting and outlines a strategy for process improvement. It examines the use of control charts and for classifying out-of-control processes. The responsibility for trouble-shooting and process improvement should not rest with only one group or department, but the shared ownership of the process.
Problems in process operation are rarely due to single causes, but a combination of factors involving the product (or service), plant, programmes and people. When faced with a multiplicity of potential causes of problems it is beneficial to begin with studies that identify blocks or groups, such as a whole area of production or service operation, postponing the pinpointing of specific causes and effects until proper data has been collected.
There are many and various patterns that develop on control charts when processes are not in control. The taxonomy presented is based on three basic changes: a change in process mean with no change in standard deviation; a change in process standard deviation with no change in mean; a change in both mean and standard deviation. The manner of changes, in both mean and standard deviation, may also be differentiated: sustained shift, drift, trend or cyclical, frequent, irregular.
The most frequently met causes of out-of-control situations may be categorized under: people, plant/equipment, processes/procedures, materials and environment. Machine learning – big data handling problems differ from traditional SPC applications. Noise reduction, data pre-processing and outlier detection are often necessary before any SPC analysis can be done; and the analysis required is more exhaustive than in standard SPC applications. Big data analytical tools are now available that can make these breakthroughs a reality.
