Cumulative sum (cusum) control schemes

Abstract

Cumulative Sum (CUSUM) quality control schemes are becoming widely used i n industry because they are powerful, versatile, and easy to use. They cumulate recent process data to quickly detect out-of-control situations. They also serve as a powerful diagnostic tool. There are now more than 10,000 CUSUM control schemes in use daily in Du Pont. This paper describes design and implementation procedures for CUSUM control schemes with emphasison properties that are recorded as counts. The paper will describe recent developments which make CUSUM procedures more useful and more powerful.

The recent developments described are: Fast Initial Response (FIR) CUSUM which gives extra sensitivity to out-of-control situations at start up or after a(possiblyin effective)control action.

A Combined She whart-CUSUM which combines the key features

of CUSUM schemes and She whart Schemes by adding She whart Control Limits to a CUSUM scheme.

ROBUST CUSUM schemes which are no tunduly influenced by a few outliers or fliers occurring in the stream of data.

The philosophy of continual improvement of aprocessisvery compatible with CUSUM procedures.As CUSUM procedures give much more responsive control,a CUSUM signal does notmean that the process is producing bad product.Rather it means that action should be taken sothat the process does not produce bad product.

Keywords:

• fast initial response (FIR)
• combined Shewhart-CUSUM
• robust CUSUM
• average run pength (ARL)