Smoothing without thinking could be trouble. Smoothing is a concept for re-solving a mine scheduling problem by adjusting conditions in an attempt to achieve period-to-period normalized results. It is done in a variety of ways and for a variety of reasons. While I generally support this concept, I think in some cases we can lose track of what the not-so-smooth solution may be telling us in the greater planning process.

This monologue investigates some of the reasons why mine planners smooth schedules in a strategic (or long term) planning context. I offer it in the spirit of fostering contemplation and possibly introspection among my fellow planning colleagues for the greater good of improving mine scheduling practices. So this is me stirring the pot…

Why Do We Do It?

Here are a few common examples of the reasons why folks like to smooth schedules:

Accounting Expectations

We’ve all been there. You spend days/weeks/months developing a mine schedule which meets all the physical and quality constraints of your operation, only to be sent back to the shop to do a bit more ‘refinement’ work to thread this needle. Perhaps your period-to-period deviation on tonnes processed or grade was too great, and a re-work was required in order to produce exactly the numbers that are expected, regardless of what the model valuations are telling you.

Feeding the Beast

Another situation is the “fill-the-process-plant-to-exactly-its-theoretical-capacity-in-every-period-at-all-costs-or-we’re-taking-away-the-coffee-machine” routine. While this may be true in a textbook case, it is hard to support in every instance and every period  in real situations. The reason for this, as you know very well, is that in mining we are dealing with a spatially and temporally variable situation. (We are not scheduling cans of beans on a conveyor belt in a controlled facility.) The honest fact is that in some cases, based on the actual scheduled conditions and availability of material, it may make sense to deliver just slightly fewer tonnes in a period or two to the plant, provided the situation has been modelled accurately. (I can already hear rumblings of protest and discontent…)

Generalizing

Often what lies at the very core of the reasons people engage in an early, heavy handed smoothing practice is this: there is a known limitation in the accuracy of the model or data being used to generate the mine schedule. This limitation could be imparted by a lack of confidence in inputs, by simplifications in approach, or by misunderstanding the behavior of the processes being modelled. After all, it is a complicated problem we are dealing with here, so let’s cut ourselves some slack. (Remember, it’s not beans.) Regardless, we deliver a solution which meets general expectations in lieu of a solution that is calculated as optimal, because the optimal solution does not completely represent the actual mechanics of the problem, and it looks bad to an accountant to boot.

When It Gets Really Bad

Digging a little deeper into this, I’ve found that in many cases, known model accuracy limitations which force us into generalization and heavy smoothing are not all created equal. You could argue that some are indeed necessary, as there is just not enough confidence in the actual data to model the process accurately. In other words, attempting to model something comprehensively with dubious data support becomes strictly an academic exercise, so why bother?

However, I’ve also seen other situations where data support was there, however modelling approximations were intentionally fabricated strictly due to restrictions (or shall I say sanctions?) of the tools used to generate the schedule. This could be through front-ending decisions, fixing material destinations, or scheduling on a flat tonnage basis when we are actually controlled by haulage hours. The list goes on, my friends. Whatever the case, if we do not have the flexibility or capability to model the real value-driving decisions or interactions between the critical elements of our operation because the tool we have does not support it, we basically have no choice but to generalize.

What Can We Do About It?

Never fear. There is a balanced way of approaching smoothing with regards to strategic scheduling. There is no doubt that some smoothing, done at the right times and in the right instances, is good practice. Firstly, we need to ensure we have confidence in our data. Next, we need to use the best technology and tools available to model the integrated mining situation as best as possible. Then, after we get that first scenario run, it is always worth taking a step back and having a good look at that rough solution to understand what it is telling you before pulling out your hatchets.

Maybe the variability you see in your physicals is related to a misalignment in your phasing design? Perhaps that erraticism in truck hours is based on a limitation in your waste routing? Every situation will be different, and every planner will have their own style of interpreting and acting on schedule results, however the fundamental concept I offer here is to approach smoothing in a systematic, purpose-driven way; and to take a moment to consider what that nasty-looking first pass may really be trying to tell you about your schedule. Well? Stirred up? Let me know what you think about The Great Smoothing Debate.