Interlate on the Benefits of Analytical Modeling

By Darius Okle, Senior Metallurgist at Interlate.

We are always looking for models. In particular, the type of modeling done by the factory puts on a quiet Thursday afternoon, the type of modeling I used to do in Microsoft Excel with the analysis tools pack. I was looking for proof of the relationships I knew existed and I was looking for confirmation that I understood the plant better than everyone else.

To become technical, it is empirical modelling. Using R and Python for analysis, tableau for visualization and a team of data scientists, empirical modeling is an offer we have embraced. From a personal point of view, this is something that both SMEs (subject matter experts) and data scientists are passionate about. From an industry perspective, there is clearly untapped value to be delivered.

This brings us to the three modeling applications we currently offer that bring the most to mining operations:

Free white paper

Multi-Concentrate Grade Optimization – Hard Rock Mining and CHPP Optimization

By Interlate

Enter your details to receive the free paper:

Please check your email to download the white paper.

  • Software sensors
  • Comparative analysis
  • Optimization

Softsensors for precise sensor measurements

Soft sensors or sensors that produce inferences or measurements in the plant from other available sensors vary in their accuracy and functionality. The best ones are usually built from underlying engineering or physical phenomena. Such as a torque sensor used on a rotating drum can give an indication of feed density and therefore feed quality. Its accuracy depends on the number of other variables that affect it and are measured. Its usefulness depends on the accuracy and reliability of the measurement and the kinds of decisions it allows us to make.

The calibration of the instrument or soft sensor is what makes it accurate. This predictability and the assurance of its accuracy make the difference between a “sensor” and a “soft sensor”. Keeping this in mind, developing soft sensors is something the team takes great pride in doing with input from SMEs, data scientists and our grumpy statistician to deliver to our customers.

Performance Benchmarking

When it comes to benchmarking or benchmarking, we try to do more than predict. Along with the forecast there is the mine estimate for tonnes and grades, followed by a throughput estimate and an aggregate loaded on equipment availability. For benchmarking, you most often look at the day before. More advanced apps can watch the measurements as they come in and give a live reference.

The objective is however the same for both to separate the impact of variables over which you have no control from the impact of variables which you can influence. Give focused, data-driven attention to areas of the plant that have the most opportunity for improvement. As much a tool for the plant metallurgist to know what to focus on as a communication and management tool.


Optimization can be difficult to discuss without first addressing the logical conclusion of motor questions. Driving questions are what we ask our customers to ask, in a way that helps drive our descriptive analyzes and guides the work in a way that provides information they can walk away with and apply to the factory immediately after our closing presentations. These questions are usually: “Which bit rate gives the best performance?” “Which grade gives me the best recovery?” “What combinations of float parameters should we use to process West54 type ore?” These are great questions to define what KPI is, but they are complicated to answer and extremely difficult to answer with techniques or applications that do not include modeling. So the solution is to overlay the models on top of each other taking care of all the convolutions, combinations and permutations you can get with later simulations to find an efficient solution.

For optimization, the empirical model is constructed to answer the constructed question. For example, let’s look at a polymetallic, lead/zinc/silver float. Mass produced a lead concentrate, then a zinc concentrate both with paying units of silver. The question would be, “What should my quality goals be for my products to produce the highest yield?” This is complicated by the different relationships that exist in the plant for different types of food. As operators, we know this is due to geometallurgy and how these relationships change for different flow rates, p80s, circuit configurations and mixes.

So the model is built to take all of this into account, followed by a mass balance, feeding into payment terms and shipping parameters. Next, we constrain the output for technical and operational limits. Finally, after its construction, we are ready for optimization. We used to try to interrogate our results with heatmaps and iterations, but now, thanks to the efficient work of our data scientists, we can do it with mathematics and machine learning. This approach can be bundled into an app so you can ask different types of questions, and if you’re ready, integrate emissions data and start uncovering the true hidden costs of your operation.


Hopefully this gives a good overview of how models can be used to help operations make decisions and improve plant performance. It’s something proven, it can work well, it makes sense and often the return on investment couldn’t be faster. Our long term plan is to work towards a combination of optimization and benchmarking to provide dynamic benchmarking or target KPI for plant operation. This would involve the benchmarking and optimization techniques we’ve talked about combined with systems used in matchmaking systems seen in competitive sports and video games to estimate expected performance and adjust ranking or, in this case, target KPIs for change.

If you would like to work with us to develop and refine our dynamic benchmarking or would like to take a closer look at some of our mature products, please contact us.

Related companies