No abstract

In creating an automatic control system for a complex technological process, we choose a target control function (i.e., a criterion by which we can assess the effectiveness of the process), compile a mathematical description of the process,write algorithms for the laws obtained, and evaluate and analyze the accuracy of the resultant control algorithms.In studying the stages of a beneficiation process, we can choose a target control function for each stage. For instance, for a crushing plant, the control target may be the maximum possible quantity and a given content of final-product class in the grader (hydrocyelone) output. The drawback of this approach is that operation of a plant unit, which is optimal from the viewpoint of the target control function of this unit, may not be optimal for the target control function of the process as a whole. The operation of a crushing unit which is optimal according to the above target conuol function may involve marked variations in the density and volume flow rate of the grader output, which will have an adverse effect on the operation of the flotation unit. However, in the absence of automatic control sensors for a number of technological parameters, it is practically impossible to determine the optimum conditions for a technological process as a whole. The estabRshment of a system of centralized control and the development of new sensors should greatly facilitate the solution of this problem.Any actions on a technological process can be divided into disturbing and control groups. Disturbing actions include the properties of the ore (mineralogical composition, content of useful component, nature of dissemination, par~cle size, textural and structural properties, etc.), and also some very slowly-varying characteristics of the equipment (changes in the lining of the grinding mill, impeller wear, etc.). Control actions include the quantity of ore fed to the mill, the number of balls, the water conditions of the mill and grader, and the reagent conditions. In addition, we can distinguish the planned output characteristics of the process -the extraction of metal into the concentrate, the content of metal in the concentrate, metal losses in the tailings, output of finished concentrate, etc.The goal of process control in a beneficiation plato can be formulated as follows.According to the measured values of the disturbing actions we must determine the values of the control actions so as to optimize one of the planned output parameters (or a group of output parameters), while keeping the remaining output parameters within certain limits.Let us examine in more detail the problem of control and the possibilities of realizing a static model, taking a crushing plant as our example. In this ease, the aim of control is to obtain the maximum possible output of the useful class (?.o~og)while keeping the percentage content of useful class within given limits (/31 -~/~ -~Bz). In addition, if we remember that the output parameters of the crushing plant are the input parameters for the subsequ...

In organizing the control of an engineering process at an enrichment plant, it is necessary to select tile frequency of sample withdrawal to secure a given accuracy for the mean value of any particular parameter in a given time interval T (e.g., the mean hourly, shift, or daily value). Lokonov [1] suggested a formula for the minimum necessary number of single (partial) samples N, I/2 (1) N=kIx where k is the coefficient of guarantee of a given accuracy of sampling, V is the standard deviation, and P is the permitted error of sampling in relation to the mean value of the parameter (in ~ This is merely one possible expression for the number of statistical samples required to find the expected value of a random variate to a given accuracy. In selection of the sampling frequency,(1) does not give reliable results, because it does not include the parameters characterizing the timewise variation of the controlled variate.In fact, we can cite random variates with the same expected value mx and dispersion Dx (and, hence, the same standard variation Vx), but with different timewise variation. Furthermore, it is obvious that the same number of measurements will give a high accuracy for determining the mean over a short time interval T, but a lower accuracy over a long one. This fact is also ignored in (1).Below we will discuss the determination of the root-mean-square error of calculation of the mean value of a controlled parameter x(t) in a given time interval T from N equally-spaced discrete measurements. It is given by to+T XT--== T1 ,I x (t) dt . (2) To find the mean value in the same time interval from the results of N equally-spaced discrete measurements we can use the expression N --.