Computer Simulation at Wheal

Mathematical modelling methods were used at Wheal Jane. to investigate the sensitivity of the Stokes hydrosizer to changes in the operating variables, with the view to improving the process efficiency. The major conclusion was that the hydrosizer was overloaded. The results indicated that throughput and spigot densities were the most sensitive variables. The model predictions demonstrated that a significant improvement in classification efficiency (both in sharpness of separation and in fines misplacement) could be achieved by (i) increasing hydrosizer capacity and (ii) increasing spigot 1 density to its practical limit. Based on these predictions, a decision was made to install a second hydrosizer in parallel with the first. Subsequent test work on the new circuit validated the predictions and confirmed that an improved efficiency had been achieved. From monitoring plant performance for 4 month periods before and after the change, it was established that a significant increase in overall tin concentrate grade had resulted. Further improvements are now expected from optimizing the hydrosizer/primary tabling circuit (as an integral unit), that is by tuning the table operations to more fully exploit the now better classified feeds. Recommendations, in this direction, will be provided by computer simulation methods. This paper reports on the computer simulation techniques used in this study: their basis and their application to optimizing both unit processes and larger blocks of the process flowsheet. The mathematical models of the Stokes hydrosizer and the shaking table are discussed briefly and their scope and limitations are ex~mined within the context of this investigation. Model predictions and sensitivity analysis results are presented and are compared with the measured data. The technical improvements are discussed in detail and the improved production statistics are analysed in terms of the financial benefit gained by Wheal Jane. Finally, the advantages of using computer simulation and other computer-based techniques, in promoting enhanced plant performances, are examined. Introduction Warren Spring Laboratory is currently working with industrial partners, Carnon Consolidated (Cornwall UK) and Beralt Tin and Wolfram (Portugal), on an EC sponsored project to develop software for the minerals industry aimed toward enhancing plant performance. The bulk of the software produced has been in the form of mathematical models of individual unit processes (eg, spirals, hydrosizers, shaking tables etc,) and a f10wsheet simulator. The simulator allows these individual units to be linked together forming flowsheets, paralleling what happens on a processing plant. Simulation and modelling techniques offer many potential benefits to both the plant engineer and plant designer. The conventional approach to optimizing or modifying an operating plant is that planned changes are based upon available process data or, perhaps, on comparisons with similar plants. Often, the information available to the engineer will be incomplete. So the decisions made will avoid major risks and, where ever pos~iQle. are aimed at minimizing disruption to the plant. Also. because of • Warren Spring Lahoratory. [)epartment of Trade and Industry t Carnon Consolidaled w, MINI:RAI. I'IH1Cl·.SSING IN 1'111' UK COMPUTER SIMULATION AT WlilcAI. JANE. CORNWALl. 97 ~ limitations on the time that can be allocatcd to the solution to any given prohlcm, the rangc of alternatives that can hc cxplorcd is often severcly restricted andthp ?ptimum solution IS rarely achieved III practlcc. SlIllulatlon and modelling methods enable the cngineer and/or designer to assess the potcntial merits of several operational strategies before implementing any real change on the plant. The techniques must be seen as lools intended to enhance the decision-making, enabling more cost-effective decisions to be made. The techniques are intended to supplement rather than replace professional cxpertise. The software developed in this collaborative project is now installed at the plants of the industrial partners and this paper describes the financial benefits already gained by Carnon Consolidated at Wheal Jane tin processing plant in Cornwall. UK. Gravity concentration plays an important part in the Wheal Jane flowsheet, (see Wells'). A single Stokes hydrosizer is used to classify the feed for the shaking table operations. However, for some time metallurgists had not been fully happy with the performance of the hydrosizer. Fines entrainment was particularly acute. The reason had not been fully established although several contributory causes had been identified, one being that the hydrosizer was treating a high tonnage of relatively fine feed that contained a significant fraction of very fine material. It was decided to use the hydrosizer model to address this problem aiming to optimize, if possible, its performance. The results of this simulation exercise, detailed later, indicated that improved hydrosizer performance could be best obtained by reducing the hydrosizer throughput. This gave Wheal Jane the impetus to change its flowsheet, and install a second identical hydrosizer running in parallel to the existing one. After monitoring the new flowsheet for four months it was evident that the hydrosizer performance had dramatically improved giving Wheal Jane real financial benefits of around £120,000 extra revenue per year. However, this is not the end of the story. Improved hydrosizer performance has caused a change in the loadings of the primary tables. Were these now running below their best? The solution will be provided through an optimization of the hydrosizer/primary tables as an integral unit. This study, which forms the next part of the simulation exercise, is now being carried out at Warren Spring Laboratory. Details of the work at Warren Spring Laboratory with respect to modelling and simulation of mineral processing plant have been published in previous papers, Tucker et ai, I. 2 Mackie et aP and Manser.4 Therefore, it is not proposed to discuss this work in great detail here. However, a brief outline of the basic gravity model (GMODEL), the simulator (GSIM) and two of the models (hydrosizer and shaking table) incorporated in the package are included for completeness. Modelling and Simulation GMODEL is a general purpose physical separation model designed initially for gravity separation devices. The software has a modular structure. The main block or model skeleton contains the input/output routines, the optimization mathematics and the standard mineral processing calculations. All the device specific information, that is essentially the model equations, are contained within a device module. These modules can be linked in with the model skeleton as required. Obviously a different device module is required for each individual separating device. Separation performance within the model is characterized by material transfer coefficients (T), which describe the probability of material being transferred from the feed to an individual output stream. These coefficients are formulated in terms of material properties such as size and specific gravity. Separation performance is represented by: A;i TK;; = BK;j where Aq is the mass flow of the ith density fraction of the jth size fraction in the process feed. TKij is the material transfer coefficient describing the partition to the Kth output stream (BKij). A single functional form (Fp) is used to describe separation performance over the normal range of operating conditions. TKij = Fp (i, j, Vl, P) main model equation and P = function (V2) auxiliary model equation. V 1 are machine operating variables dealt with explicitly in the main model. Their effect on performance is known sufficiently well to set up precise mathematical rela'tionships. V2 refer to other operating variables whose effects on performance are less well defined. P represents a set of model parameters. To model an existing plant, performance of that plant must first be measured. Model ca1ibration, that is definition of the parameter set P, is achieved by minimizing the differences between the measured performance data and model transfer coefficients through optimization of the parameter values. After the main model is calibrated in this way, the determined parameter set will define the exact relationship in the auxiliary model. Predictions for a new set of operating conditions can then be made directly from the model equations. GSIM is a process simulator which parallels on a computer what happens on a plant. The individual device modules in GMODEL can be linked, individually or as banks, together to form complete flowsheets or blocks of a more complex flowsheet. In order to use the simulator the flowsheet under examination must first be described in terms of the actual devices used and how they are linked together. Secondly, the feed or feeds to the circuit must also be described. This is done in the same way as for GMODEL, that is in terms of their size and specific gravity distributions, massflow rates and pulp densities. Lastly, the problem needs definition. In its simplest form GSIM can be used to predict the performance of an existing plant under a given set of conditions. However, GSIM can also be used as a design tool to optimize the circuit operating conditions in order to best achieve a pre-set goal. This goal is normally one of two options, either a mineral recovery can be optimized whilst maintaining a minimum acceptable grade or mineral grade can be optimized whilst maintaining a minimum acceptable recovery. In addition other constraints may be imposed such as target massflows and pulp densities of the feed per device or block of devices. The mathematics within GSIM are complex and are described in detail in Mackie and Tucker. 5 A model of the Stokes Hydrosizer r' Within the hydrosizer model the physical laws of settling are combined with a probabilistic approach to give the probability of particles in each size/specific gravity category reporting to a particular product. For the purposes of modelling the hydrosizer the best method of expressing the results was as depletion coefficients (T*). The T* for a spigot is the probability of particles that have not reported to a preceding spigot reporting to that spigot. Within the model T* is expressed as the sum of two functions: T* = Fl + F2 where Fl models the transfer due to settling and F2 models entrainment of fines.