Is data indigestion a risk?
Jun 11th, 2019
In a world that used to be data-starved, is data indigestion a risk?
Data starvation
It is rare that we review a project where the monitoring system really works.
Instruments fail, are destroyed by neighboring construction, by contractors and animals. Winter conditions, soft soils and tailings or other factors forbid access, timely replacement or maintenance.
And then, piezometer do not measure the levels they should because of some geotechnical oddity or construction problem, inclinometers are too short, etc.
In some cases the most sophisticated monitoring systems turn out to be very expensive thermometers and nothing more than that.
Even in well equipped sites, the geotechnical world is in perpetual data-starvation mode.
The solution
In a recent article entitled The internet of tailings we read that our friend Chad LePoudre of SNC-Lavalin predicts a data-drenched future. In that world there will be, he adds, integrated management systems allowing real-time reevaluation of slope stability predictions, potentially even using artificial intelligence.
We share that vision. Actually we are sure it is going to happen, it is happening right now. Indeed all that technology is enticing and bears big promises. The key will be advancements in monitoring using sensors and devices capable of broadcasting their data via gateways, directly to digital dashboards. And that too is already happening.
The idea is not new, but modern technology allows to lower the prices of such an installation making it very competitive with respect to classic manual readings. And, again, modern technology makes these systems less needy of energy and maintenance. Reportedly some systems can work for ten years without battery replacement!
But there is more! As we have written in this blog (and in a specific chapter of our forthcoming book), drones and satellites offer a valid complement to classic monitoring techniques.
Nowadays optical observation and InSAR offer geometry surveying tools, ways of checking ponds, humidity, vegetation stress. Thanks to Space observation, large areas, that are impractical even for fixed wing drones, can be monitored independently of weather conditions, predatory birds (ever seen a condor attacking a drone?), etc.
Where is the catch?
There are numerous catches in this rosy scenario future. Unless we perform a well thought-out holistic risk assessment and a cost-benefit analysis we cannot decide which path is the best.
Devices will keep breaking, misreading what they should read, due to a number of issues. And then let’s not forget possible cyberthreats on the monitoring networks and data communication.
Data driven artificial intelligence will be lacking of experience. That is because fortunately failures are not that common. Historic failures occurred in a data-starved world, remember? And, perhaps more importantly, artificial intelligence will never understand which “original sins” the structures may hide. For example:
- boreholes were too short to recognize that weak layer in the foundation,
- engineers were too audacious,
- factor of safety was used to “predict” safety and finally,
- no one considered seismic conditions for “minor” events,
Indeed, let’s not forget that elements leading to an increase of the probability of failure of a dam may exist before anything can be measured. Furthermore measures, even if in real time, come oftentimes too late to avoid catastrophic failures.
Is data indigestion a risk?
At Riskope we believe it is. The hazard arises from two fronts.
The first one is that in order to cope with the high volume data flow we will rely more and more on artificial intelligence. Thus we will simply stop thinking. We are at a point where independent review panels acting in the aftermath of recent catastrophes have spelled out where the problems lie. Numerous voices, including ours (forthcoming book) have risen, and solutions which allow a paradigm shift exist (ORE2_Tailings link)
In order to give appropriate answers to society, lenders, regulators we believe monitoring devices are not sufficient. Because an indigestion of real-time, machine interpreted data is not sufficient or appropriate.
We do not need a bazooka to kill a fly. Similarly, we do not need to go from data-starvation to data-indigestion. What we need is to look at the proper data. Use “simple” solutions that condense decades years of experience we have gathered as geotechnical engineers.
Am I talking like an “old man”? Perhaps, but I am pretty sure that more technology is not really going to substantially reduce the likelihood of future failures.
The mistakes of the past cannot be corrected by accumulating data in the future. After all, if you ate and drank too much last night, taking your blood pressure, monitor you heart and checking your temperature every two seconds will not predict when… well, you know what I mean, right?
Tagged with: data indigestion, likelihood of future failures, ORE2 tailings
Category: Probabilities, Risk analysis, Risk management
Right on