Extracting the "Operator's Know-How" from Operating Records

"Mineral processing and smelting operations have the reputation of being complex processes. Process operators of these unit operations have acquired from experience and basic training a superficial understanding of the relationships between process output and manipulated variables. This understanding, which is a very valuable know-how for process control and optimization as well as for operator training, is however difficult to extract. This paper describes a time series based approach that can be used to extract this know-how, which can subsequently be used for operator training. The method is illustrated with an example taken from a smelting process.IntroductionThe emergence of Total Quality Management and Statistical Process Control philosophies in the North American industry (Brown, 1991 ), as well as the increasing popularity of the IS0- 900x standard accreditation, provide a strong incentive to review the training procedures for plant operators. The importance of employees training is acknowledged in two of the 14 points of Deming's road to quality (Deming, 1982) as well as by control engineering consultants (Hendricks et al., 1994). However guidelines for the operation of industrial processes are often transferred from an operator to another one with allowance for flexible interpretation of many operating rules. Such transfer is typical of complex industrial processes to which particulate processes of ore dressing and extractive metallurgy belong.Capturing the knowledge of experienced operators is often a difficult task. Cognition engineers are able to extract this knowledge and transfer it into rules that ultimately are implemented into expert systems. Cognition engineers are however not always available and plant supervisors may prefer to keep training as an internal liability. Another difficulty through interviewing of the process operator is its lack of willingness to transfer what he has learned by hard working to other people, whose ultimate goals are to develop computerized applications that may replace him in a near future, or reduce its importance in the production line.This paper shows that time series can be advantageously used to capture the average process understanding of plant operators and to help reviewing training procedures. The method s illustrated with data collected during normal operation of electric arc furnaces and it is shown how the approach has led to a fruitful discussion between process engineers and operators that ultimately provided savings of more that $200,000 per year per furnace. There are 9 furnaces in the studied foundry. Such savings are not the results of application of advanced control strategies nor of ""miracle press button software», but of an improve awareness by engineers and operators of the importance of properly defining the operating procedures to control a process."