Automated Design Analysis and Generation of Diagnostics
Submitting Institution
Aberystwyth UniversityUnit of Assessment
Computer Science and InformaticsSummary Impact Type
EconomicResearch Subject Area(s)
Information and Computing Sciences: Artificial Intelligence and Image Processing, Computation Theory and Mathematics, Information Systems
Summary of the impact
Automotive design analysis software based on qualitative reasoning
research in the Advanced Reasoning Group at Aberystwyth is deployed at
more than 200 automotive and aeronautic OEMs and Tier 1 suppliers
world-wide. The software necessitates companies changing their process for
performing design analysis, and companies are willing to do this because
of the attendant benefits.
The major benefits of use of the software are early feedback on potential
problems with the design of automotive systems, and improved safety of
automotive designs. Related benefits are improved product time to market,
and cost savings. A representative example of production savings of $2.5
million has been given for use of the software on a single product design,
as well as Ford Motor Company's estimate of $20M per year saved in just
their company.
Underpinning research
The Advanced Reasoning Group at Aberystwyth have been a leader in
Qualitative Reasoning Research for more than 15 years. For a good deal of
that period, it was funded to coordinate qualitative reasoning in Europe
as lead node of the European network on Model-based and Qualitative
Reasoning (MONET — 1995-2001; MONET2 — 2001-2006). During this time, one
significant strand of its research was aimed at automating design analysis
and at producing diagnostics for automotive systems. The design analysis
automation research was funded by EPSRC and the EC, and carried out in
collaboration with companies including Ford, Jaguar, the Motor Industry
Research Association, Daimler-Benz, Fiat, Volvo. Research in this area has
continued during the REF period with funding from the Technology Strategy
Board, BAE Systems, Flight Refuelling and Sumitomo Electrical Wiring
Systems.
The research resulted in several key developments between 1995 and 2005.
Aberystwyth developed novel techniques for qualitative reasoning of
electrical systems [3.1,3.2], and for deciding on the effects of component
failure in a system and abstracting the results to an appropriate level
for presentation to users [3.3]. The use of qualitative reasoning to
decide current flow in electrical networks [3.1] coupled with a higher
level state-based component representation [3.2] was pioneered in the
group, and proved effective for simulation of automotive electrical
circuits. Such representations were easier to build from reusable
libraries than the equivalent numerical simulation, and covered a wider
range of possibilities, including failure situations. The use of
functional labels to interpret system level states and provide natural
language analysis reports [3.3] also originated in the group, and this
innovation made it possible to produce focused results in language
appropriate for engineers.
In addition, because the qualitative models could be adapted for use in a
wider range of tasks than was possible with numerical simulation, methods
could be developed for combining these features together in order to
achieve different kinds of task-based reasoning, and the group produced
further novel design analysis tools that were of commercial interest
[3.4,3.5,3.6].
The ease of model building for the engineers, the wide applicability of
the models constructed, and the usefulness of the results achieved, were
attractive to automotive companies, and Ford Motor Company funded the
initial commercial development of tools based on the research prototypes
built by the group on EPSRC grants.
Research in this area has continued at Aberystwyth since the formation of
the spin-out company, FirstEarth, continuing to improve the modelling
techniques, and to develop novel diagnostic and design tools using these
techniques, specifically widening the application area from electrical
systems to mixed domain systems, and pioneering new kinds of analysis
(diagnosability and automated generation of on-line diagnostics). One
example of this on-going work is given in [3.6].
Key Researchers:
Mark Lee, Professor
Chris Price, Professor
Neal Snooke, Lecturer (01/09/1997)
References to the research
[3.1] M. H. Lee. Qualitative circuit models in failure analysis
reasoning. Artificial Intelligence 111 (1-2), pp239-276, 1999.
DOI: 10.1016/S0004-3702(99)00032-6
[3.2] N. A. Snooke, Simulating electrical devices with complex behaviour,
AI Communications 12 (1,2) pp45-58, 1999.
[3.3] C. J. Price, Function-directed electrical design analysis, Artificial
Intelligence in Engineering 12 (4) pp445-456, 1998. DOI:
10.1016/S0954-1810(97)10013-9
[3.4] C. J. Price, N. A. Snooke, S. D. Lewis. A layered approach to
automated electrical safety analysis in automotive environments. Computers
in Industry 57, pp451-461, 2006. DOI: 10.1016/j.compind.2006.02.001
[3.5] C. J. Price and N. S. Taylor, Automated Multiple Failure FMEA, Reliability
Engineering and System Safety 76(1), pp1-10, 2002. DOI:
10.1016/S0951-8320(01)00136-3
[3.6] N. Snooke, C. J. Price. Automated FMEA-based diagnostic symptom
generation. Advanced Engineering Informatics2028 26(4), pp870-888,
2012. DOI: 10.1016/j.aei.2012.07.001 REF2 submitted.
Details of grants
[3.7] Grant: An Automated FMEA Assistant. Jan 1993 — December 1995, £217K
from EPSRC and £21K from Ford Motor Company.
[3.8} Grant: AQUAVIT: Advancing Qualitative Analysis for Verification,
Interaction analysis and Testing. April 1996 — March 1999, £215K from
EPSRC(CDP) Grant GR/L20542 and £55K from Ford Motor Company.
[3.9] Direct funding from Ford Motor Company to develop AutoSteve. Oct
1995-March 1997. £203K.
[3.10] Grant: VMBD: Vehicle Model Based Diagnosis. £144K from European
Commission DG XIII (total project size £4M) .
[3.11] Grant: Whole Vehicle, Whole Lifecycle Electrical Design Analysis.
January 2000 — May 2003. £245K from EPSRC (Foresight Vehicle) Grant
GR/N06052 and £219K from Ford Motor Company.
[3.12] Grant: ASTRAEA: Safety Cases for Unmanned Aerial Vehicles.
(£218K). This was a wide ranging project led by Technology Strategy Board
with a total value of £32M. Aberystwyth received funding in automated
detection of on-board problems using qualitative models.
Details of the impact
Automotive design analysis tools based on qualitative reasoning research
at Aberystwyth are deployed at more than 200 automotive and aeronautic
OEMs and Tier 1 suppliers world-wide [5.1]. In order to use the tools,
engineers in the companies need to change their design processes, but this
is seen as being worthwhile, because of the improved safety and the
potential for cost savings that results.
Safety is improved, because safety design analysis can be carried out
early in the design process and then repeated painlessly as the design
evolves [3.4,3.5]. Cost savings result both from the reduced engineer
effort needed to do the safety analysis, and from the fact that the
analysis can been carried out early in the design process — fixing design
problems early is much cheaper than fixing them late in the design
process.
The qualitative reasoning based automated FMEA research described in the
papers in section 3 led to the first ever commercial automated FMEA tool,
called AutoSteve, developed by Aberystwyth Computer Science in
collaboration with Ford Motor Company. This resulted in a spin-out
company, FirstEarth, that supported and further developed AutoSteve,
marketing and selling it to other automotive companies. FirstEarth was
well supported by the Department over a five year period (1998-2003), and
further results of our research were embodied in its design analysis
tools. Researchers, academic members of staff, graduate and undergraduate
students were all seconded to the company at some point. At its height,
the company employed 12 staff, and of those, nine had been trained by the
Department. Of those nine, six chose to work for Mentor Graphics when it
bought FirstEarth in 2003, and four of those (all originally RAs on
related research projects) are still key technical staff for Mentor
Graphics in their Newbury research and development centre [5.1].
FirstEarth Limited and Aberystwyth University developed further
commercial automated design analysis tools based on the continuing
research at Aberystwyth. These include automated sneak circuit analysis,
fusing strategy design analysis, and simple design verification. The
underlying qualitative reasoning meant that all these tools were able to
work on a single library of system models built by automotive engineers,
and that useful analysis could be carried out much earlier in the
automotive design lifecycle than had previously been possible. A language
for describing the function of systems that had been developed as part of
the research, along with ways of using that language to interpret system
behaviour, and this meant that the tools could generate consistent reports
in the kind of language used by engineers and managers — another
significant advantage over previous tools.
When Mentor Graphics took over FirstEarth Limited, they incorporated the
tools into their "Capital" harness design toolkit, and renamed AutoSteve
as part of the toolkit. They have split AutoSteve into four tools that
support Capital Harness Design, all built on the Aberystwyth research
described above [5.2]. The subtext after each tool name is the description
from their website of what that tool does for the engineer:
- Capital SimCertify: Bullet proof design using FMEA
- Capital SimProve: Eliminate sneak faults
- Capital SimStress: Size components for worst-case configurations
- Capital SimTransient: Validate transient behaviour
In order to estimate the reach of the impact of this research, in 2012 we
asked the Integrated Electrical Systems Division (IESD) of Mentor Graphics
about the usage of this technology over the past few years. They informed
us that Mentor Graphics presently has design tools based on technology
developed at Aberystwyth installed at over 200 vehicle producers and Tier
1 suppliers. These tools are having a significant effect on improving
vehicle safety and on the efficient production of vehicle designs.
Potential design problems are identified early.
The tools are also saving significant design costs. The rapid automated
safety analysis of designs means that many more design options can be
explored cheaply. A recent example of this is provided by one of the
world's largest producer of automotive harnesses. In 2009, the company
used the tools to identify how to save 50 cents per harness produced. When
you are producing 5 million harnesses, that gives savings of $2.5 million
for a single product.
In 1999, the Ford team working with FirstEarth and Aberystwyth University
on AutoSteve was awarded a Ford Europe Technology prize for the impact of
AutoSteve. The business case produced by Ford in submission for the prize
estimated that AutoSteve was then saving Ford around $20M per year, but
that the key benefit was the improved analysis of design safety [5.3].
The supporting impact letter from Mentor Graphics says "Mentor Graphics
presently has design tools which include technology developed at
Aberystwyth deployed at more than 200 vehicle producers and Tier 1
suppliers.", and highlights both "improved design quality" and "usage can
also result in significant cost savings". Given increases in costs since
Ford did their study in 1999, and the wide deployment of these tools, we
conclude that the overall economic impact is very significant indeed.
Sources to corroborate the impact
[5.1] Confirmation of the present state and usage of the tools can be
obtained from the Product Line Director at Mentor Graphics Inc.
[5.2] Mentor's web site also has details of Capital Harness Software
(http://www.mentor.com/products/electrical-design-software/capital/)
and of the Capital design analysis products that were originally AutoSteve
(http://www.mentor.com/products/electrical-design-software/capital/additional-automation/electrical-analysis).
[5.3] Details of the Ford Technology Prize for the AutoSteve work are
given on page 32 of an article in AI Magazine: P. Struss and C. J. Price,
Model-based systems in the automotive industry, AI Magazine special issue
on Qualitative Reasoning, pp17-34, vol. 24(4), Winter 2003.