Page 440 - Design for Six Sigma a Roadmap for Product Development

402 Chapter Eleven

a. Transferring the risk of failure to other systems outside the
project scope
b. Preventing failure altogether [e.g., design poka-yoke (error-
proofing)]
c. Mitigating risk of failure by
(1) Reducing “severity” (altering or changing the DPs)
(2) Reducing “occurrence” (decreasing complexity)
(3) Increasing the “detection” capability (e.g., brainstorming
sessions, concurrently, using top-down failure analysis such
as FTA)
10. Review analysis, document, and update the DFMEA. The DFMEA
is a living document and should be reviewed and managed on an
ongoing basis. Steps 1 to 9 should be documented in the appropriate
business publication media.
The potential failure modes at any level can be brainstormed by
leveraging existing knowledge such as engineering or architecture
analysis, historical failure databases of similar design, possible
designed-in errors, and physics of failures. For comprehensiveness and
as a good practice, the black belt should instruct the DFSS team mem-
bers to always maintain and update their specific list of failure modes.
The understanding of safety-related and catastrophic failures can be
enhanced by a fault-tree analysis (FTA), a top-down approach. FTA,
like FMEA, helps the DFSS team answer the “What if?” questions.
These tools deepen the understanding of the design team to their cre-
ation by identifying where and how failures may occur. In essence,
FTA can be viewed as a mathematical model that graphically uses
deductive logic gates (AND, OR, etc.) to combine events that can pro-
duce the failure or the fault of interest. The objective is to emphasize
the lower-level faults that directly or indirectly contribute to high-
level failures in the DFSS project structures. Facilitated by the struc-
ture’s development, FTA needs to be performed as early as possible, in
particular to safety-related failures as well as Design for Reliability
(Chap. 10).

11.3.1 FTA example
In this example, the FTA will be applied to a vehicle headlamp. The
electric circuit is very simple and includes the battery, the switch, the
lamp itself, and the wire harness (Fig. 11.4). For simplicity, we will
assume that the latter is reliable enough to be excluded from our
study. We will also assume certain failure probabilities for some com-
ponents. For a given time period, the probability of failure is the DPMO
or the unreliability for the assigned distribution of failures (not neces-
sarily normal). Such probabilities can be estimated from warranty and