Today we took the grandkids to the beach, Huntington Beach, and I snoozed in the sand as they played in the waves and built sand-castles.
Then we went for ice-cream and bought the grandson a new Trek bicycle. Before that I finished a book by Norman Fenton and Martin Neil Risk Assessment and Decision Analysis with Bayesian Networks. It opened my eyes to ways of doing risk assessments and making decisions in mining situations. I have not hitherto seen the use of Bayesian Networks in a mining context, and truth to tell, still have not.
The Wikipedia article on Bayesian networks is scholarly and dense with mathematics. A Google search of Bayesian Mining, yields hundreds of articles on data mining using Bayesian methods. This is the only thing I found that remotely connects Bayesian and mining as an extractive industry.
Is it that the Bayesian method is too complex for ordinary people in the mining industry to understand or use? It cannot be that the methods are not powerful enough to deal with the difficult situations that arise in mining.
One professor of mining whom I told that I was reading the book remarked: “I have always meant to get down to the topic, but never found time.”
So right out here is my challenge to the mining industry, it practitioners, consultants, and academics: read about and learn the Bayesian method and show us how to apply it to solve mining problems.
The book is easy to read, albeit it there is much mathematics you have to deal with. But it is fascinating and fun to read. The story it tells is how to use Bayesian logic (and mathematics) to update estimates of risk so that you can make informed decisions.
I had always thought of risk as the simple product of probability and consequence. So do all the other mining people I know who profess to use risk assessment methods. But they are simplistic as this book rapidly demonstrates.
They may tart up their talk with reference to robust & resilient systems, black swans, fragile & anti-fragile systems, and other fashionable things like MAA or FMEA. Once you have read this book you cannot but reject this talk as acronym heaven sans virgins of understanding.
I submit, as a blogger, that the mining industry is way behind current learning when it blathers on about acronyms of yore and fails to use Bayesian logic. As this book explains it is easy to do and powerful.
When you buy the book you get to download a computer code that make use of the Bayesian method easy, even if you do not fully understand the mathematics—which is actually very easy, being no more than a series of equations of the form, the probability of x is equal to the probability of z divided by the joint probability of X and Y. And if you know P(y) you can back calculate a posterior probability. OK: I am not sure that is correct, so go see the book for details.
When I get back to the company computer in Vancouver I will play with the code we downloaded and tell you how I succeed. The first thing I plan to do is back analyze the probability of failure of the Bafokeng slime dam given what I subsequently learnt when three more similar tailings facilities in the area failed. (No record of these subsequent failures. Only me and Gordon McPhail know the details.)
Another challenge: can those folk presenting at the InfoMine Seminar on Cold Region Covers go beyond FMEAs and consider Bayesian methods in talking about the long-term performance of soil and vegetation covers in cold places.
I and others have written a new EduMine course on Risk Assessment, Decision Making, and Management of Mine GeoWaste Facilities. With luck it may be live sometime in 2014. I will update the draft to be loaded with more on the use of Bayesian methods. I am thinking of examples as I ride the boardwalk in the sun of Southern California. And as I here appeal to you to help me with examples.
This article appears courtesy of I Think Mining. To read more musings about mining click here.