State-of-the-art production accounting system for mineral sands mining operations

Size: px
Start display at page:

Download "State-of-the-art production accounting system for mineral sands mining operations"

Transcription

1 PRETORIUS, A.M.E., CROUS, G., and RADEMAN, J.A.M. State-of-the-art production accounting system for mineral sands mining operations. The 7th International Heavy Minerals Conference What next, The Southern African Institute of Mining and Metallurgy, State-of-the-art production accounting system for mineral sands mining operations A.M.E. PRETORIUS*, G. CROUS, and J.A.M. RADEMAN * *CSense Systems Anubis cc Production environments have only become harsher over time. In order to remain competitive all players has to know their strengths, as well as their weaknesses. The ambiguity in the industry s production spreadsheets and the demand for transparent production figures along with established integrity force players to up their game. Not only does competition demand more accurate and reliable information about stock and material movement, but legislation such as the Sarbanes- Oxley Act of 2002 also enforces transparency in the production environment. CSense Systems has developed a production accounting solution for the mineral sands mining project of QIT Madagascar Minerals (QMM). Mining was kicked-off in the Fort Dauphin region of Madagascar at the end of Production accounting entails the application of basic mass balance principles. The whole production site is divided into different plant areas also called blocks. The data from active streams are validated in accordance with statistical validation rules, establishing the integrity of the data used. The stock levels of the respective compounds, minerals and the total dry mass for each block (plant area) are adjusted in accordance with the daily material movement. The resulting validated data and inventory levels are finally adjusted and signed off by supervisory personnel on a typical monthly basis, ensuring a completely transparent hierarchy of information. The CSense production accounting system thus empowers production personnel to make strategic decisions based on a superior information hierarchy. This enables QMM to be a prominent competitor in the heavy mineral industry. Introduction Current status of the industry Production environments and the economic playing field have only become harsher over time. In order to remain competitive all players has to know their strengths, as well as their weaknesses. In the production environment knowledge about process capability and asset availability empowers decision-makers, while the lack thereof can be viewed as a weakness. The majority of production sites currently keep track of their valuables by means of manually updating spreadsheets. These updates on average have a three-day delay due to the availability of laboratory results. The resulting disconnected planning leads to uncertainty in process capability and asset availability, resulting in wasted resources, inventory surplus, and lost business opportunities. The lack of timely information in a unified, contextualized production perspective leads to poor problem resolution and decision making. The lack of process knowledge or inaccurate models leads to poor asset utilization. The inability to readily gather and distribute process knowledge is a weakness. Production transparency However, the biggest problem is data reliability given the fact that various users have the ability to tweak figures to display their reality, unaudited. Not only does competition demand more accurate and reliable information about stock and material movement, but legislation such as the Sarbanes-Oxley Act of 2002 also enforces transparency in the production environment. The AMIRA P754 practice code acknowledges these requirements and it is noted on the AMIRA website: Corporate governance is increasingly focused on transparency between the technicalities of sampling, assays and reconciliation with financial performance. Apart from the competition in the mineral industry, legislation nowadays forces major players to compete transparently. Data integrity Levels of information distribution Keeping track of the valuables throughout the production process forms an integral part of the control and the planning of the production process. Knowledge empowers personnel to make strategic, yet objective decisions. It is essential that the right information is provided to the right people in real time. Figure 1 displays the information technology layers that exist within the production environment. STATE-OF-THE-ART PRODUCTION ACCOUNTING SYSTEM FOR MINERAL SANDS MINING OPERATIONS 177

2 Various people have proposed decision rules for detecting unnatural patterns on control charts. Among these are four decision rules suggested by Western Electric (1956), four rules by Grant and Leavenworth (Grant, et al.,1988) and eight rules by Nelson (Nelson, 1984). The statistical validation rules used in the production accounting system are based on these proposed rule sets. The above-mentioned rule sets are used to identify outliers, which are observations that deviate considerably from the majority of observations. Outliers may be caused by sensor noise, process disturbances, instrument degradation, and/or human-related errors, as emphasized by Liu et al. Figure 1. Information technology architecture It is clear from the information technology architecture that the provision of correct information to the production decision-makers is dependent on proper data collection and data distribution practices. Raw plant data are transformed to information and consequently to knowledge. Plant data require a measure of integrity in order to prevent skewed knowledge and wrong decisions. The reality is that many management information systems use erroneous plant data. Principle areas that contribute to error Contrary to popular belief, sampling variances are not selfcompensating, but additive. The original work by P. Gy (Gy, 1973) finds the following four principle areas adding to total sample error: Material variation variation in source composition, e.g. raw materials, stock piles Process variation variation in process performance and/or equipment degradation or instrument drift, etc. Sampling variation variation in sampling equipment performance (maintenance) and/or changes to sampling procedures Analytical variation variation in laboratory equipment performance (maintenance) and/or variation in assay procedures (different assay methods). The additive nature of variation in samples confirms the care with which all sampling practice should be approached. The need for data integrity Establishing the integrity of data is of utmost importance in view of: The ambiguity and disconnected nature of current production spreadsheets The production transparency required by current legislation and standards The industry requirement of management information systems that is dependent on various information layers The definite existence of sampling errors. Establishing data integrity Data variation Over years of investigating variation in data statistical techniques have been employed to monitor and manage process performance. It is known that as you travel three standard deviations from the population mean, the probability of occurrence beyond that point starts converging to a very low number. This tendency is irrespective of the distribution shape. Data validation techniques Data validation techniques can be categorized into the following levels of maturity: Static validation (univariate) Fixed validation parameters are specified against which the data points of a single variable can be compared. This is the most basic form of data validation and involves a set boundary like a slurry density that should not be less than kg/m 3. Dynamic statistical validation (univariate) Statistical validation rule sets are used to validate and compare the most recent data point in accordance with statistical parameters (variance, average, etc.) over a moving time window. Rules-based validation (uni- or multivariate) Rules-based validation can be performed on a single process variable as well as multiple variables. Process rules such as known relationships between process variables can be employed in the validation process. An example of a validation process rule is that the flow rate of one process stream may never exceed half of that of another process stream based on process configuration and mass balance principles. Model-based validation (multivariate) Data points are compared against data from a multivariate model (soft sensor) for validity. Model-based validation requires the development of a model of the process variable to be validated as well as the identification and preparation of the input variables for the model. In the event of using this validation technique test data and plant conditions need to have achieved a level of maturity in order to ensure their relevance and stability. Production accounting CSense Systems has developed a production accounting solution for the mineral sands mining project of QIT Madagascar Minerals (QMM). Dredge mining was kickedoff in the Fort Dauphin region of Madagascar at the end of The mineral separation plant of QMM produces an Ilmenite concentrate along with a zircon-rich material. Process model for production accounting Production accounting entails the application of basic process mass balance principles. As shown in Figure 2, the whole production site of QMM was divided into different plant areas also called blocks. A hierarchy of blocks was also introduced, allowing larger blocks to house smaller blocks. 178 HEAVY MINERALS 2009

3 Figure 2. (a) Blocks and streams (b) Block hierarchy Validation index (VI) and confidence indicator (CI) A validation index (VI) and confidence indicator (CI) were assigned to each measurement validated by the system. The VI is based on the rules in the validation rule-set (Table I). The VI values are binary values and their association with the respective validation rule is summarized in Table II. Table I Statistical validation rule set The feed and product streams representing material movement to and from the respective blocks were identified. In order to quantify and qualify the material movement flow, measuring devices and sampling points were identified associated with each stream. A typical process block with its feed stream and product streams are displayed in Figure 3. The whole production accounting process will be discussed in accordance with this figure. Validation Figure 3. Typical process mass balance block Statistical validation rule set In order to establish the integrity of the field measurements and the laboratory results, a dynamic statistical validation technique was employed. Table I contains the statistical validation rule set applied over a moving sample window of 100 data points. Validation rule Rule 1 Rule 2 Rule 3 Rule 4 Rule 5 Rule 6 Rule 7 Rule 8 Rule 9 Validation rule Description None of the validation rules failed Failed activity condition to determine the activity of a process stream. The operation of an associated pump or conveyor belt was indicative of an active process stream Each data point is compared with the specified absolute maximum value associated with the measurement (Engineering limit known as high high limit, HHL) Each data point is compared with the specified absolute minimum value associated with the measurement (Engineering limit known as low low limit, LLL) Validate if a single data point lies beyond 3 times the value of the standard deviation (sigma) from the data sample average (same side) Validate if two out of three data points lie beyond twice the standard deviation from the sample average (same side) Validate if four out of five data points lie beyond one standard deviation from the sample average (same side) Validate if six sequential data points are not all increasing or decreasing Validate if seven or more data points are not on one side of the sample average Table II Validation Index values Rule 1 0 Rule 2 1 Rule 3 2 Rule 4 4 Rule 5 8 Rule 6 16 Rule 7 32 Rule 8 64 Rule VI STATE-OF-THE-ART PRODUCTION ACCOUNTING SYSTEM FOR MINERAL SANDS MINING OPERATIONS 179

4 The CI is a percentage value which gives a measure of reliability of the plant or laboratory measurements. A CI of 100% implies that none of the validation rules was breached. The breach of each validation rule had a negative influence on the CI, which in turn caused a lower confidence level in the measurements. The more validation rules were breached, the less the CI. Value types Each data point entering the production accounting system had an associated value type, estimated, CSense Adjusted, Measured and User Adjusted, described in Table III. This practice ensured an auditable production accounting system as the source accountable for the data point was recorded. The production accounting process The production accounting process is diagrammatically represented in Figure 4 and the whole process will be discussed accordingly. A user-friendly configuration tool was developed in order to simplify the configuration of the accounting system. External inputs The measurements from the field instruments and laboratory analysis are usually made available by means of a database (LIMS, Weigh Bridge and/or custom databases). A third party database was used to make the raw data available to the production accounting system. Table III Value types Data Value Description classification type Estimated 1 A temporarily unavailable measurement (e.g lab assay) is filled with an estimated value CSense adjusted 2 A measurement exceeding the limits of Rule 3 and 4 is clipped to a specified value Measured 3 An online measurement or laboratory assay User Adjusted 4 A user with the appropriate authorization adjusts values Production accounting database (MS SQL) The production accounting system can be described as a calculation layer that provides results to the production accounting database. Subsequent to each stage of value addition the data is stored in the database. This enables a measure of drill-down functionality into the results as well as transparency. Get data The production accounting system systematically recovers raw data from the external database and stores these raw values in the production accounting database. This ensures the highest level of transparency as none of the original plant data are lost. Figure 5 displays the various options for the acquisition of raw data: Database mapping Constant value is employed A formula is entered to derive the value Calculated using mass balance principles The value of another stream is used. Validate data Each raw data point was validated according to the statistical validation rule set. A validation index, confidence indicator and value type were attributed to each validated data point. Figure 6 shows the density measurements for stream MIN_001_0001. The associated VI, CI and value type for this raw measurement are displayed in Figure 7, Figure 8 and Figure 9 respectively. A VI with a value of 64 refers to statistical validation rule 8. Six sequential data points that are all increasing may be indicative of instrument drift. It is clearly indicated that the CI decreased as a statistical validation rule was breached, while the value type classification remained measured. Figure 5. Data acquisition options Figure 4. Production accounting process Figure 6. MIN_001_0001_DEN density measurement for stream MIN_001_ HEAVY MINERALS 2009

5 Figure 7. Validation index of density measurement for stream MIN_001_0001 Figure 8. Confidence indicator of density measurement for stream MIN_001_0001 Calculate stream characteristics Upon validation the stream characteristics are calculated and stored in hourly buckets. This implies that the dry mass of the material movement is calculated from the flow and density measurements. The configuration of these calculations is specified by means of the configuration tool, shown in Figure 10. Laboratory assays are used to calculate the tonnages transferred for the respective mineral and chemical components. During the interim (starting 9:00 a.m.) while the laboratory assay is unavailable and the mass assay (Figure 11a) and the laboratory assay (Figure 11b) for the process stream is estimated, the value type is classified accordingly in Figure 11c. The data used for the illustration does not represent QMM production figures. As soon as the laboratory result valid for the period starting 09:00 a.m. is made available, all the associated estimated values are replaced with the actual measured values. Figure 12a shows the mass assay for stream MIN_002_0003 derived from the actual laboratory result (Figure 12b) and the value type classification changed to Measured (Figure 12c). Calculate stock The material movement per process stream is aggregated over periods of shifts, days and months. This allows the user to track chemical components throughout the process. Figure 13 shows the dry mass distribution of the compound TiO 2 surrounding a surge bin, block MIN_002, the unit of measure is tonnages per day. The opening stock and closing stock as well as the total feed streams and product streams for that specific day are also recorded, only illustrative data is used. It is noteworthy that some process streams were not equipped with the appropriate measuring devices for mass flow. The flow of dry mass of stream MIN_002_0004, which is Trommel losses, was assumed to be 1% of the total dry mass fed to the surge bin, MIN_002. This and similar configuration exceptions were also accommodated. Stock validation and mass balance The stock levels can be validated according to physical constraints such as the actual capacity of process units and mass balance principles. Figure 9. Value type of density measurement for stream MIN_001_0001 Human validation Production personnel or management with the appropriate authorization are required to sign off and lock the production figures provided by the production accounting system. Corrections or changes up to a minutely resolution Figure 10. Stream configuration calculation of actual dry mass STATE-OF-THE-ART PRODUCTION ACCOUNTING SYSTEM FOR MINERAL SANDS MINING OPERATIONS 181

6 Figure 11. Estimation of unavailable laboratory results (a) estimated mass assay (tonnages per component; (b) estimated TiO 2 wt%; (c) value type during estimation are allowed while the reason for change and the identity of the authorized person is saved along with the raw data. This ensures transparency and production figures with integrity. It is noteworthy that values classified as user adjusted supersede all other value types which enforce personnel to take ownership of the production figures. Final validated data The final locked figures are finally stored in the production accounting (MS SQL) database. Figure 12. Replacement with laboratory results (a) actual mass assay (tonnages per component; (b) measured TiO 2 wt%; (c) value type with actual measurement External reporting As displayed in the production accounting process in Figure 4 a third party reporting system was employed to make the results from the database available to the production team. The Figures display the conceptual design layout of production accounting reports, according to the data that would be available from the production accounting system. As described earlier production personnel would be allowed to view summarized production figures (feed and product streams) and stock levels for the respective blocks. 182 HEAVY MINERALS 2009

7 Figure 13. Component balance (tonnages) Figure 14. Monthly block report Figure 15. Hourly report Figure 14 displays the block summary on a monthly level. The dry mass column represents the total dry tonnages while the respective mineral columns represent the dry tonnages per mineral component. The additional value types stored with each value allows colour-coded values. In the illustrative report below the user adjusted values are red, the estimated values are pink and the actual measurements are black. As all the data of the respective calculation layers are stored, a drill-down functionality on the basis of a month, day, shift and hour would be feasible. In Figure 15 the hourly summation of the online measurements for a stream is shown. Faulty field measurements can be rectified by a user with the appropriate authority. The shift average of the online density measurement is corrected on a shift resolution, STATE-OF-THE-ART PRODUCTION ACCOUNTING SYSTEM FOR MINERAL SANDS MINING OPERATIONS 183

8 Figure 16. Report editing Figure 17. Monthly sign-off indicated with a red colour, which imposes that all the applicable data of a higher resolution is locked as displayed in Figure 16. Finally the monthly data per stream could be signed off by the appropriate authorities at month end. Conclusion The production environment requires proper production accounting systems which use data with established integrity and are simple to configure. The answer to the question of an auditable production accounting system with data integrity is executing at QMM. CSense Systems developed a production accounting system which is efficient and provides transparent production figures to personnel on various levels of the company s organizational structure. Key functionalities of the system are: A user-friendly configuration tool exists The system is totally configurable Production figures are available on a basis of shifts, days and months Production figures with attributed validation indexes and confidence indicators are provided Temporarily unavailable laboratory results and values derived from it are estimated for the interim Changes made to data by personnel are completely traceable Authorized personnel are required to sign off and lock production data for complete transparency. The CSense production accounting system thus empowers QMM production personnel to make strategic decisions based on a superior information hierarchy. This enables QMM to be a prominent competitor in the heavy mineral industry. References AVERY, A. ABB Sharpens Focus on CPM with Common Platform Product Offerings and Powerful Connectivity and Visualization Tools. ARC Insights, pp CARGILL, R., FREEMAN, N., and GILBERTSON, R. Metal balance of a large integrated operation using a standard industrial software package. FISKE, T. Manufacturers Making CPM Top Priority. ARC Insights, pp GRANT, E.L. and LEAVENWORTH, R.S. Statistical Quality Control. Sixth ed. McGraw-Hill, New York GÜLBAY, M. and KAHRAMAN, C. Development of fuzzy process control charts and fuzzy unnatural pattern analyses. Computational Statistics & Data Analysis vol. 51, pp HEAVY MINERALS 2009

9 GY, P.M. The sampling of broken ores A review of principles and practice. The Institution of Mining and Metallurgy. London, Geological, Mining and Meallurgical Sampling: pp , and discussion pp JIANG, W., LIU, H., and SHAH, S. On-line outlier detection and data cleaning. Computers and Chemical Engineering, vol. 28, pp MINNITT, R.C.A., RICE, P.M., and SPANGENBERG, C. Part 1: Understanding the components of the fundamental sampling error: a key to good sampling practice. The Journal of Southern American Institute of Mining and Metallurgy, vol. 107, pp NELSON, L.S. The Shewhart control chart-tests for special causes. J.Qual. Technol. vol. 16, pp RADEMAN, J.A.M. Is smart adaptive technology a requirement for metal accounting? MMS Mag November pp RADEMAN, J.A.M. The need for data integrity in the information hierarchy. MMS Mag April pp WESTERN ELECTRIC. STATISTICAL QUALITY CONTROL HANDBOOK. Western Electric, New York A.M.E. Pretorius Process Engineer, CSense Systems I graduated at the end of 2007 as a chemical engineer from the North West University, South Africa. My first and current employer is CSense Systems. Through my position at CSense Systems I have gained extensive experience in applying their process performance enhancement products in various industries. These industries include: base metal refining, aluminium anode production and heavy mineral sands mining, concentration and smelting. I was responsible for the process interpretation and the configuration of the Production Accounting solution at QMM. STATE-OF-THE-ART PRODUCTION ACCOUNTING SYSTEM FOR MINERAL SANDS MINING OPERATIONS 185

10 186 HEAVY MINERALS 2009