STATISTICAL PROCESS CONTROL MODELS IN AGRO-CHAINS

K.C.B. Roes
Consumer expectations on quality of food products are very high and tend to increase as well as diversify to include flexible and fast delivery, health, safety and minimal environmental impact. Assurance to conform or even exceed these expectations can no longer come from mass inspection, but has to be founded on intelligent process control, process and product design and continuous improvement. Statistical Process Control (SPC) provides the sound basis for this. Although SPC used to be strongly associated with the statistical tools applied, it is now regarded in general as an indispensable approach to managing processes (Deming (1986), Snee (1990), Joiner (1994), Hare et al. (1995), Hoerl (1995), Does et al. (1997), Roes and Dorr (1997) and Roes (1997)). In the broader context this approach is based upon the principles that:
  • all work is a series of interconnected processes;
  • all processes vary;
  • sources of variation can roughly be distinguished as arising from common causes (inherent to the process as designed) and special causes;
  • understanding the origin of each of these sources of variation is the key to reduction of variation;
  • reduction of variation is the key to quality improvement, productivity and profitability.

Statistical methods such as control charts, experimental design, data analysis are applied to uncover causes of variation and thus control and improve the processes. SPC can be implemented on the shop floor by cross-disciplinary teams, called Process Action Teams (PATs). In production processes such teams consist of operators, foremen, process-engineers, maintenance-engineers and other technical personnel involved with the process, and a statistician. A PAT implements SPC for a specific process following a stepwise approach, based on the Plan-Do-Check-Act cycle. This forms a close link between statistical thinking and the scientific method. The main steps are:

  1. Definition of the process to be dealt with
  2. Diagnosis of the process
  3. Actions and measurements
  4. Design of feedback control loop
  5. Implementation and further improvement (back to I)

The result of the phases I through V is usually twofold. The main purpose is to install a control loop with control charts and accompanying out of control action plan (Figure 1). In this control loop, deviations from the normal performance of the process are detected by means of control charts. Subsequently, the shop floor operators follow the out of control action plan to identify and remove the cause as quickly as possible. Concurrently with establishing this control loop, opportunities for improvement arise during process diagnosis and appropriate action is taken or is planned to be taken once control is established.

The process diagnosis is a crucial step and includes describing processes using flow-charts and performing a risk analysis based on the Failure Mode and Effect Analysis (FMEA) technique (see Stamatis, 1995). Possible causes and effects critical quality characteristics

Roes, K.C.B. (1998). STATISTICAL PROCESS CONTROL MODELS IN AGRO-CHAINS. Acta Hortic. 476, 23-32
DOI: 10.17660/ActaHortic.1998.476.2
https://doi.org/10.17660/ActaHortic.1998.476.2
Statistical methods, statistical process control, quality

Acta Horticulturae