The importance of context

How mass balancing and a structured data management approach can add greater meaning to laboratory and plant data

In the mining sector, it’s essential that plant engineers and managers have confidence in the accuracy of the information they receive, especially from the plant laboratory. Recent improvements in technology and a greater understanding of the ways that data can be used means that technical staff are starting to get more out of their laboratory information; however, at many plants it can be very difficult to determine the level of accuracy and reliability of individual assay results, and therefore how useful the data is – let alone if it can be used at all.

Many sources of uncertainty

The variable nature of ore entering a plant – and its behaviour during processing – means that relatively large changes in data patterns could represent genuine scenarios. Yet, without having full confidence in the results, plant decision-makers are only able to quickly accept or discard results based on experience or very basic comparisons.

The key issue is that plant data is typically collected, aggregated and analysed manually – in some instances using fairly basic spreadsheets. As well as the risk of human error affecting results, these manual methods make it difficult to collect enough information to conduct a broad investigation, especially in a timely fashion.

In some instances, plants may utilise a basic plant information system. These systems, however, are often fairly limited in their scope and level of flexibility, which again makes it difficult to analyse results quickly or with adequate depth.

The value of metallurgical data is greatly increased when it is considered alongside other relevant information, as it is given a situational context. Without this, results cannot be effectively evaluated in terms of their accuracy and utility.

When measuring plant performance, for instance, how can a plant engineer determine if a specific tailings grade is a particularly good or bad result? Is a high value indicative of a single outlier, or a broader trend? Is a particular plant target value based on last year’s average, a simplified model or a budgeted financial number?

To truly understand assay results, we need to know where that value came from in the first place and how it relates to everything around it. Is the sample point located before the pond, or after? Is it collected from the top of the tank, or the bottom? The effect of large volumes on mixing or particle size distribution within a tank should also not be ignored.

Without examining its contextual significance, there is no way to identify if an unexpected result is an accurate representation of the plant, or if it is due to an issue with sampling, analysis or administration.

There are many steps in the sampling and analysis process that could lead to inaccuracies. A poor result could be caused by suboptimal sample collection, handling or preparation. For instance, how long was a sample held in a bottle prior to being washed? Was the sample washed?

There could also be uncontrolled instrument drift that affects the analysis, a mistake made during data entry or an error in an additional calculation. This multitude of factors can make it difficult to identify inefficiencies in the plant or to isolate issues that may impact product quality. It means that opportunities can be missed, or costs incurred that could potentially be avoided.

Creating context – structured databases and dynamic mass balancing

The solution to this context conundrum lies in having a robust plant information system that is able to take all influential factors into account and output the collated results quickly and accurately.

With the right system in place, it can be possible for engineers to cross-check results against a second comparative data set that is generated from a dynamic plant mass balance. The dynamic mass balance takes the most important information generated by the plant, such as production and ore feed rates, and uses this as the main inputs for a model of the plant. It also takes into account a broad range of other inputs – including multiple sample assay results and plant instrument data.

All of this information is used to create a second set of data, where flows, inventory changes and even chemical reactions are accounted for. The dynamic model is also able to capture volumes and residence times, which is particularly important in hydrometallurgical plants.

Automated data transfer components also allow information to be accessed, collected, evaluated and reported without any significant delays or user intervention. This automation also helps to ensure data integrity, as any corrections occur outside of the automated process and therefore can be logged.

All of this means that decision-makers can quickly determine if a laboratory result is relatively good or relatively bad. It is also possible to see the elemental mass flow between vessels, or even to determine plant inventory.

Consider the following scenario in Table 1, which uses actual plant data.

Table 1 shows a series of copper concentration laboratory assay results recorded for a leach thickener overflow stream. These numbers don’t look particularly interesting at first glance, as most appear to be relatively similar.

Table 1. Leach thickener overflow Cu concentration assay results (real plant data).

However, when these results are presented graphically in Figure 1, it becomes more obvious that there could be an issue related to this data. Is the large jump in concentration accurate? Should all of these results be used in the calculation of plant performance metrics, or by other decision-making tools?