Evaluation of Improvements Resulting From the Automatic Control of Mineral Processing Operations

Introduction The ability to evaluate the improvement resulting from the implementation of an automatic control system is essential to the Justification of its cost. This evaluation may occur after anew control system has been commissioned. In other cases an existing plant which is operated manually may be automated only In part in order to determine In ether the met of it should be. Finally. in some instances dynamic isolation my be used prior to installation of the control system to determine the strategy to be used or to evaluate the improvements In process performance expected from automatic control. This paper describes the development of a methodology for making these evaluations. In order to obtain an accurate assessment of Improvements In a short time, statistical methods can be employed. Mineral processing plant, are subjected to many types of both random and deterministic disturbances of various duration. The statistical method, available for process control at analysis are designed to counter the obscuring effects that random disturbances have on such as evaluation. The basis for evaluation and comparison is some freely chosen measures of process performance termed the performance index or objective function. The experimental data used to calculate the performance index should, if possible, be collected by alternately running a circuit Or plant with the two or more type, or control being tested ("serial testing”) or more preferably, by running two circuits with the two type, of control side by side ("parallel testing") . The performance index is a random variable so that when comparing two types of control, statistical hypothesis testing must be used. By testing s hypothesis statistically one accept, or reject, a (null) hypothesis [ ] concerning the mean or variance of one or more performance indices. For example, a hypothesis Ho might be that the mean of a performance index with stabilizing control is larger than with manual control. Defining a Performance Index The performance index, denoted by [ ] , is a measure by which alternative strategies or situations my be cam peered on either s technical or an economic basis (see Table 1). Technical measures are concerned with the effects of control on process variables or outputs or on the deviations of controlled variable, from their setpoints. For commotion circuit, important technical measures of performance are throughput, product fineness and the consumption of liners, media and electrical energy. For flotation circuits important technical .satires are recovery, grade and reagent consumption. Economic measure such a, cost and profit n usually determined from a combination of technical measures and are usually the final factor upon which a decision may be made, although a control system my achieve higher values of technical measure , the control effort expended my be of such magnitude that the economic measures, such as cost, are poor. Running Comparisons Comparison, between automatic and manual control are normally done with data obtained before and after control system Installation, unless ore characteristics remain the same during the period of data collection the results will be biased. By running the test circuit with the control system alternately turned on and curled off until all possible ore type, have been run through the circuit, unbiased result, can be obtained. This mode of testing is termed “serial testing.” An alternative and potentially more powerful and quicker made of testing is "parallel testing" in which two circuits, one with the type of control being tested and one without, arm run aide by side. In this type of testing both circuits process the same ore so that valid results occur sooner than with aerial testing. The data may be averaged before testing or the differences in the performance indices may first be calculated before averaging. However, the two circuits must be nearly identical in their mechanical characteristics otherwise differences other than the degree of control will bias the results. It mechanical differences exist than it my be necessary to adjust one or the circuits prior to testing. Sometimes it is possible to correct far mechanical differences based on Inherent efficiency calculations made with a model. Hypothesis Testing Once the data is collected, hypothesis fencing can be carried out to determine if the automatic control system has resulted in a statistically significant improvement. One can construct a histogram of the distribution or performance indices with am without control such as Figure 1, or the