Catastrophic failures forensic analyses are, for the most part, completed by experienced engineers. At least, that’s the case for tailings dams failures and other catastrophes like major infrastructural accidents and aviation. Are we sure that is the best way?
Normalization of deviance, management and decision-making are often ingredients of the catastrophic failure, not just technical faults. Because of this, we believe independent panels should include social scientists, as they are the most qualified to study social and organizational dimensions of failure causality.
The idea is not new. Indeed, in the 1970s, British sociologist Barry Turner investigated a series of disasters and noted that warnings or prior near-misses were overlooked. Had anyone, particularly the skilled professionals involved in those cases, picked up on these signs, catastrophic accidents could have been averted.
Why do Failures Happen?
So why does this happen? We published a paper in 2016 (Oboni, F., Oboni, C., A systemic look at tailings dams failure process, Tailings and Mine Waste 2016, Keystone, Colorado, USA) that showed that a simple model looking beyond engineering hazards into factors like overconfidence, insufficient effort, and mistakes was sufficient to model the rate of failure of the worldwide portfolio.
After the last few investigations by independent panels around the world, we think our 2016 model should be extended to cover “soft issues” such as poor perception of warning and near-misses, complacency toward normalization of deviance, unawareness of risk and finally misplaced risk appetite.
Why Aren’t you Talking About Failure Modes?
Failure modes explain how a failure occurs, not why it occurs. When we consider “soft issues,” there is no need to use failure modes to answer the question of why. As Terry Eldridge (Golder) said in his keynote lecture at Tailings and Mine Waste 2019, engineers focus on failure modes when they design. Risk analysis is another job, uses different skills, and must focus on the “whys”. Perhaps even more importantly, risk analysis should begin on a project from inception and should be able to detect emerging issues that can lead to entirely different risks than the ones engineers identified.
Failure modes also come from a heritage of siloing. By boxing complex reality in failure modes, it is easy to forget that failure arises because of a multitude of unfortunate small choices, mistakes, or inattention, and that it is generally not the result of one cause alone.
Catastrophic Failures Forensic Analyses by Sociologists
A sociologist from Yale, Charles Perrow, wrote a book called Normal Accidents. Perrow stated that apparently trivial events and non-critical incidents sometimes interact in ways that preliminary engineering analyses could not predict. He called those failures “normal accidents.”
Tailings dams generally function with plenty of trivial events and non-critical accidents such as localized erosion, seepage, settlements, and poorly maintained diversion channels, weirs and water management ancillary structures. Human error like defective monitoring, poor understanding of the geology, and overly shallow investigations can also be a contributing factor.
Engineers cannot eliminate all of these, of course, especially not at inception. It is the job of the risk analyst to identify these risks and evaluate their compounding effect on the probability of failure. Again, that cannot be a boxing exercise based on failure modes. In addition, we know where working in silos leads: the rate of failure of tailings dams around the world has been known, with obvious uncertainties, since 2013.
Of course, no matter how detailed a risk analysis is, it will never be able to identify and compound all possible deviances. However, in ORE2_Tailings™ we have spent years to finally distill a list of KPIs that deliver an understanding of the causality of failures.
Are Tailings Dam Failures True Normal Accidents?
Normal accidents are trivial or non-critical incidents that unexpectedly compound into critical incidents. Tailings dam failures will continue to be normal accidents if we keep boxing reality like with failure modes approaches. If we develop risk assessments that avoid boxing, then we have a chance of decreasing blind spots. Additionally, if the risk analyses reveal that some risk elements in the portfolio are strategic and require change to the system, then we are better armed to do so.
Normal accidents tend to occur in complex systems. Dams are certainly a complex system: they are investigated, designed, built, operated, monitored and managed by interacting teams over long timespans. Dams can also be interdependent with other dams and structures. IoT and automation add a layer to the complexity. Risk assessments must be on par with such complex systems to be useful.
Normal accidents also generally do not challenge common engineering understandings and theories, likely because the deviances that generate them are rarely intrinsically surprising. Because dams are “unique structures,” lessons learned are difficult to assimilate—a dam can only fail once!
Closing Remarks
Normal accidents are not the only sociological theory on accident causes. Other approaches are oriented towards “technological innovation”−linked catastrophes. As we have shown, characterizing dam failures as normal accidents is quite fitting. However, the industry is improving as we speak, and we are confident that with a few more appropriate changes we will be ready to challenge present public perception on the risk associated with tailings dams.
