This is a QFDI newsletter from April 2011, discussing the danger of using ordinal scale math in FMEA, namely for computing risk priority number (RPN) for assessing black swan events. The topic is too important that we thought to share it again for those who missed it.
"The role of Quality in Fukushima nuclear crisis"
1. Centralized consensus vs. triage leadership in disaster preparedness and decision making.
One of the tenets of quality management is "Plan-Do-Check-Act." We find that when the planning has been done properly and consensus built among constituents, most processes will fulfill requirements, and the Check-Act serves to fine tune the process. In Japan, this consensus building is called "ne-mawashi" or going around the roots of a tree before transplanting it to make sure everything is ok.
While TQM experts praise consensus as good for planning, there is a downside that Dr. Deming warned about in chapter 6 of his book The New Economics. That is — "with shared responsibility, no one is responsible." Thus, ne-mawashi can lead to finger pointing and blame instead of collaboration, as well as increased murkiness in accountability and delay in critical actions.
2. This raises these quality questions:
(a) In a disaster, do we go back to Plan or do we go directly to Do-Check-Act (sometimes called Do-Redo) at the local level?
'Planning' may require subject matter experts who may not be optimally located since the exact location of the disaster may be unknown until after it occurs, and time which may be limited by threat to life or subsequent failures in other systems.
Also, in terms of 'planning' resources, are the same resources being competed for various emergency operations (such as fire, police, medical), or should different resources be planned? From a time perspective, should the priority be given to allocating the resources to take care of those who are still alive and need immediate assistance, or should the resources be expedited first to cooling nuclear fuel to address the medium term risk to the life and livelihood of survivors?
In the case of Japan, were certain needs more urgent than others? Such as the need to verify the emergency level vs. the need to issue a quick evacuation order; the need to determine resources for disaster relief vs. the need to add resources to prevent a nuclear event, etc. And how should those priorities be made, by whom, and when? Should such priorities have changed the way the leaders approach the 'planning,' 'doing,' and 'checking'?
(b) In disaster preparedness, how has the extent of the disaster be predicted?
If the disaster falls within the predicted parameters, the planned response may be sufficient. If the disaster rises to unanticipated levels, however, as is the case in Japan, the response plan can easily become insufficient.
"Beyond expectation" was how virtually everyone — from Tokyo Electric Power Company (the operator of Fukushima power plant) to the government nuclear power regulators and safety commission— described the March 11 earthquake and tsunami in Tohoku region, although retrospective review of historic data begins to hint otherwise.
The probability of a nuclear fatality was set in 2003 by the Japanese Nuclear Commission (JNC) to not exceed 1 × 10-6 per year or about 1 in a million years. On the Japanese nuclear event, Nassim Nicholas Taleb, author of The Black Swan, cautions, that model error causes underestimation of small probabilities and their contribution (see his web site). This highly improbable event with massive consequences is what Taleb calls a "Black Swan."
(c) Is standard FMEA practice adequate for for a Black Swan event?
In FMEA (Failure Modes and Effects Analysis) we try to account for this Black Swan by looking at not only frequency of occurrence, but also impact and detection. Assuming JNC's probability estimate for a nuclear fatality of 1 × 10-6, the likelihood of a M9.0 earthquake at less than 1 per 100 years or 1 × 10-2 (worst case prediction), and the likelihood of a 20 meter tsunami at less than 1 per 100 years or 1 × 10-2 (worst case prediction), the probability of all three occurring simultaneously would be 1 × 10-10, or 1 in 10,000,000,000 (one in ten billion).
Showing posts with label black swan events. Show all posts
Showing posts with label black swan events. Show all posts
Subscribe to:
Posts (Atom)