Stochastic Mine Planning: Common Question I Always Hear

"Do you generate a separate sequence for each orebody simulation?" I hear this very often. And the short answer is: 𝗡𝗼. 𝗡𝗼. 𝗔𝗻𝗱 𝗮𝗯𝘀𝗼𝗹𝘂𝘁𝗲𝗹𝘆 𝗻𝗼. Think about it: If you had a different extraction sequence for every simulation, which one would you follow? Which one would the medium and short-term planners inherit? Which one would operations execute in practice? It would be 𝗰𝗵𝗮𝗼𝘀. Planning needs clear and actionable guidance, not 100 sequencing versions. Let me break it down simply: When you're solving a LoM plan optimization problem using block-based scheduling approach, there are three major decisions to be made: • 𝗘𝘅𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻 𝘀𝗲𝗾𝘂𝗲𝗻𝗰𝗲 (which blocks to mine and when) • 𝗖𝘂𝘁-𝗼𝗳𝗳/𝗱𝗲𝘀𝘁𝗶𝗻𝗮𝘁𝗶𝗼𝗻 𝗽𝗼𝗹𝗶𝗰𝘆 (where the material goes: mill, stockpile, leach pad, waste, etc.) • 𝗣𝗿𝗼𝗰𝗲𝘀𝘀𝗶𝗻𝗴 𝘀𝘁𝗿𝗲𝗮𝗺 𝗱𝗲𝗰𝗶𝘀𝗶𝗼𝗻𝘀 (material flow after the block's first destination is determined) Now, here’s the important part: The extraction sequence is what’s called a 𝗳𝗶𝗿𝘀𝘁-𝘀𝘁𝗮𝗴𝗲 decision (in two-stage stochastic optimization terminology). Which means: It must be defined 𝗿𝗶𝗴𝗵𝘁 𝗻𝗼𝘄 before any additional information is revealed. It is also called a 𝘀𝗰𝗲𝗻𝗮𝗿𝗶𝗼-𝗶𝗻𝗱𝗲𝗽𝗲𝗻𝗱𝗲𝗻𝘁 decision. You need a sequence that planning teams can actually work with. "First-stage" decisions are commitments you make before knowing exactly what Mother Nature has prepared for you. They're the backbone of the mine plan. ✅ So what sequence does stochastic mine planning actually give you? • A 𝘀𝗶𝗻𝗴𝗹𝗲 𝗼𝗽𝘁𝗶𝗺𝗶𝘇𝗲𝗱 𝗲𝘅𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻 𝘀𝗲𝗾𝘂𝗲𝗻𝗰𝗲. • Built by looking at all orebody simulations 𝘀𝗶𝗺𝘂𝗹𝘁𝗮𝗻𝗲𝗼𝘂𝘀𝗹𝘆. • Designed to 𝗺𝗮𝘅𝗶𝗺𝗶𝘇𝗲 𝗡𝗣𝗩 while trying to minimize deviations from targets across uncertainties. In other words: You don't optimize for one future. 𝗬𝗼𝘂 𝗼𝗽𝘁𝗶𝗺𝗶𝘇𝗲 𝗮𝗰𝗿𝗼𝘀𝘀 𝗺𝗮𝗻𝘆 𝗳𝘂𝘁𝘂𝗿𝗲𝘀, and find the path that, on average, gives you the best, and 𝗺𝗼𝘀𝘁 𝗿𝗶𝘀𝗸-𝗿𝗲𝘀𝗶𝗹𝗶𝗲𝗻𝘁 𝗼𝘂𝘁𝗰𝗼𝗺𝗲. That’s why the extraction sequence is scenario-independent. That’s why stochastic planning works in the real world and major mining companies are starting to integrate it into their planning workflow. So next time someone asks: "Do you generate a sequence for every simulation?" you can just smile and say: "𝗡𝗼, 𝘄𝗲 𝗴𝗲𝗻𝗲𝗿𝗮𝘁𝗲 𝗼𝗻𝗲 𝘀𝗲𝗾𝘂𝗲𝗻𝗰𝗲 𝘁𝗵𝗮𝘁'𝘀 𝘀𝗺𝗮𝗿𝘁 𝗲𝗻𝗼𝘂𝗴𝗵 𝘁𝗼 𝘀𝘂𝗿𝘃𝗶𝘃𝗲 𝘁𝗵𝗲𝗺 𝗮𝗹𝗹." 𝘐𝘯 𝘧𝘶𝘵𝘶𝘳𝘦 𝘱𝘰𝘴𝘵𝘴, 𝘐 𝘸𝘪𝘭𝘭 𝘵𝘢𝘭𝘬 𝘢𝘣𝘰𝘶𝘵 𝘵𝘩𝘦 𝘰𝘵𝘩𝘦𝘳 𝘥𝘦𝘤𝘪𝘴𝘪𝘰𝘯 𝘵𝘺𝘱𝘦𝘴 𝘢𝘯𝘥 𝘩𝘰𝘸 𝘸𝘦 𝘨𝘦𝘵 𝘱𝘳𝘰𝘣𝘢𝘣𝘪𝘭𝘪𝘴𝘵𝘪𝘤 𝘧𝘰𝘳𝘦𝘤𝘢𝘴𝘵𝘴 𝘰𝘶𝘵 𝘰𝘧 𝘢 𝘴𝘪𝘯𝘨𝘭𝘦 𝘳𝘪𝘴𝘬–𝘳𝘦𝘴𝘪𝘭𝘪𝘦𝘯𝘵 𝘴𝘦𝘲𝘶𝘦𝘯𝘤𝘦. 𝘚𝘵𝘢𝘺 𝘵𝘶𝘯𝘦𝘥!