Docstoc

Proceedings of

Document Sample
Proceedings of Powered By Docstoc
					                 Proceedings of ASME 2010 International Mechanical Engineering Congress & Exposition
                                                                                         IMECE 2010
                                          November 12-18, 2010, Vancouver, British Columbia, Canada



                                                                                       IMECE2010-39259


                  DRAFT: VERIFYING THE USABILITY OF FAILURE-BASED
                          COMPUTATIONAL DESIGN METHODS

                        Sarah Oman                                          Michael Koch
             Complex Engineering System Design                  Complex Engineering System Design
                         Laboratory                                          Laboratory
             School of Mechanical, Industrial, and               School of Mechanical, Industrial, and
                 Manufacturing Engineering                           Manufacturing Engineering
                  Oregon State University                              Oregon State University
                     Corvallis, OR 97330                                Corvallis, OR, 97330
                Email: omans@onid.orst.edu                      Email: michael.david.koch@gmail.com

                     Matt Bohm, Ph.D.                                    Irem Y. Tumer, Ph.D.
               Design Engineering Laboratory                     Complex Engineering System Design
             School of Mechanical, Industrial, and                            Laboratory
                 Manufacturing Engineering                       School of Mechanical, Industrial, and
                  Oregon State University                            Manufacturing Engineering
                     Corvallis, OR 97330                               Oregon State University
                Email: bohmm@onid.orst.edu                                Corvallis, OR, 97330
                                                                 Email: irem.tumer@oregonstate.edu

ABSTRACT                                                        INTRODUCTION
In the early stages of the design process, there is a need to   As outlined in the literature [1],[2], there is a need to
provide designers with tools to assess risk and failure so      provide designers with tools to assess risk and failure in
as to avoid costly redesigns, comply with established           the early stages of design for multiple reasons. An
safety measures and to promote innovation throughout            important reason is avoiding redesign in the late stages of
the process. Over the last several decades, various             development, as it can be costly to make changes when
methods have been proposed and researched that                  physical production has occurred. Another prominent
accomplish such tasks, including Risk in Early Design           reason to design for risk and failure is safety. As such,
(RED) and the Function Failure Design Method (FFDM).            the need to assess such risks or failures in the conceptual
The following paper looks to extend prior research by           design phase is paramount to ensuring that products and
looking at the validity of such failure-based                   designs are ready to move into the prototyping or
computational design methods. This was accomplished             manufacturing stages.
by analyzing products with a history of failure,
decomposing these products into functional models and           Function-Failure Design Method (FFDM)
running both RED and FFDM analyses on these models              The Function-Failure Design Method (FFDM) [3] is a
to see how closely such methods are able to identify real-      methodology that utilizes matrix multiplication to tie
world failures. The end goal was to assess whether both         functions to failures. The basic premise of FFDM works
RED and FFDM can correctly predict the failures                 by taking Function-Component and Component-Failure
documented by consumers as well as which failures are           matrices and multiplying them to create a Function-
most correctly predicted.                                       Failure matrix. Failure data is drawn from information on


                                                                1
the Bell 206 Rotorcraft, NTSB reports, maintenance
manuals and engineering judgment publications [4].

In order to utilize FFDM, a web-based application has
been developed that requires the user to input an
adjacency matrix that shows the correlation between
every function-flow state of a product. The output of the
analysis can either be a list of failures for each function-
flow state or a MorphMatrix of possible solutions to each
function-flow state with the failures associated with each
                                                                            Figure 1 – FFDM Output in List Form
artifact solution.
                                                               With the Morphological Matrix failure output, the
The format for FFDM input is a Text (Tab delimited) file
                                                               program suggests artifacts from the database that satisfy
in the form of an adjacency matrix, similar to Table 1
                                                               the function and flow, and shows the associated failures
below, that provides data on how the function-flow pairs
                                                               that the artifact has documented, as shown in Figure 2.
are connected to each other. The data has to be
specifically formatted and worded in order for FFDM to
run correctly. If this is not followed, an error will appear
prompting the user to check their data format.
                                 transfer electrical energy




                                 position human material
                                 actuate electrical energy
                                  import electrical energy




                                   import human material
                                    guide electrical energy




                                    import human energy




                                                                        Figure 2 – FFDM Output in Morph Matrix Form




                                                               Risk in Early Design (RED)
     import electrical energy    0   1   0   0     0   0   0   The Risk in Early Design method has been developed to
                                                               list failures as well as calculate the likelihood and
    transfer electrical energy   0   0   1   0     0   0   0
                                                               consequence of those failures. As such, it is an extension
       guide electrical energy   0   0   0   1     0   0   0   of FFDM, and utilizes this information to describe the
    actuate electrical energy    0   0   0   0     0   0   0   severity of such failures [5].
      import human material      0   0   0   0     0   1   0
    position human material      0   0   0   0     0   0   1   RED is implemented through a computer software
                                                               program that runs through the MacOS Terminal Window
       import human energy       0   0   0   1     0   0   0   and provides users with a list of potential risks and
              Table 1 - Example Adjacency Matrix               failures for all the function-flow pairs of a proposed
                                                               design. The user inputs the list of function-flow pairs
For the output, each failure lists a percentage indicating     into the program along with several needed procedural
how often that failure happens for a particular function-      text files that represent heuristic data for component
flow state relative to the entire list, along with the         failure and function-component pairs.
documented number of times that failure occurred within
the database. Sample output is depicted in Figure 1.           The process of analyzing each product begins with the
                                                               creation of a functional model in order to identify every
                                                               function-flow pair of the device. Using the model, the
                                                               function-flow pairs are then formatted in a Text (Tab




                                                               2
delimited) file as a list in a very specific manner, such as                      Table 2 – Risk Chart Output
below:

    Import Human
    Import Electrical Energy                                    PROBLEM DEFINITION
    Import Solid                                                In order to test the usefulness of both RED and FFDM,
    Guide Human                                                 products were taken from the Consumer Products Safety
    Import Human Energy                                         Commission website at CPSC.gov that were recalled or
    Guide Human Energy                                          had a significant number of complaints due to a particular
    Import Control Signal                                       failure [6]. The failures range from a trigger/switch stuck
                                                                in the “on” position to the device catching fire. These
Careful consideration must be taken in order to ensure          products must not already be found in the Design
that the program is able to analyze each function-flow          Repository as this could cause skewed results.
pair. If the wording or formatting is incorrect, RED will       Functional models were created for 14 products so that
not output any failures for that given function-flow. For       they can be evaluated with RED and FFDM to gather
example, it does not analyze “Import Liquid Material”,          information on the predicted failures and risks that both
but does analyze “Import Liquid.”                               methods propose. Table 3 lists the products analyzed for
                                                                this research.
The analysis can then be run for four different scenarios:
Human centric subsystem failures, Human centric system          The analysis for the consumer products in question were
failures, Unmanned subsystem failures and Unmanned              run in RED as Human Centric at both the System and
system failures. RED will output a text file containing         Subsystem levels. However, the outputs for all products
two data types. The first output is a list of every potential   were identical between the System and Subsystem levels,
failure for each function-flow pair inputted, along with a      most likely due to the limited functions of the products.
pair of values representing the likelihood and                  Thus, only the System level failures for RED are
consequence of such a failure, occurring on a scale from        presented in the research.
1-5 (i.e. (3,5) represents a likelihood of 3 and
consequence of 5). The output looks similar to the text
below:                                                                Product                  Manufacturer            Cite
                                                                    Circular Saw                     Ryobi              [7]
Channel energy fails due to high cycle fatigue, (5,5)
                                                                     Miter Saw               WMH Tool Group             [8]
Channel material fails due to high cycle fatigue, (3,5)
Connect material fails due to high cycle fatigue (5,4)          Steam Cleaning Mop                   Lysol              [9]
Support material fails due to high cycle fatigue (2,5)           Vegetable Chopper                 Lifetime            [10]
                                                                      Blender                        Haier             [11]
The second output is a Risk Chart created in the form of a         Coffee Grinder                 Starbucks            [12]
5x5 matrix populated from the number pairs in the first             Food Mixer                   Fit & Fresh           [13]
output, aggregating the paired-values for the likelihood             Crossbow                  Master Cutlery          [14]
and consequence of the failures. Table 2 shows the Risk            Lawn Trimmer                Black & Decker          [15]
Chart for a toy. In the Human-Subsystem analysis, nine               Nightlight                    Sylvania            [16]
failures occur at the highest consequence but lowest                Office Chair                  MooreCo              [17]
likelihood.                                                          Tea Kettle            Bristol, Martha Stewart     [18]
                                                                     Treestand                      Summit             [19]
                        Risk Chart                                     Toaster                    Wal-Mart             [20]
                     Human – Subsystem
                                                                           Table 3 - Products from CPSC Recalls List
                        0 0 0 1 0
                        0 0 0 0 1                               Once the RED and FFDM analysis was complete for each
                        0 0 0 1 0                               of the products, the results were analyzed to see how
                        0 0 0 0 0                               closely they related to the actual real-world failures and
                        0 0 14 12 9


                                                                3
risks. The following questions were considered during         the product given the information provided by the
this analysis:                                                website and the use of failure definitions.

       Were the correct real-world failures identified for                              Crossbow
        each product? If the correct failures were                Recorded Failure         Trigger loosens and discharges
        predicted, was the risk of that failure easy to           Function of Failure        Actuate Mechanical Energy

        analyze?                                                  RED                                  None
                                                                  List FFDM                 High Cycle Fatigue (17%,1)
       Did RED and FFDM miss important failures?
                                                                                              Abrasive wear (17%,1)
       Does RED or FFDM do a better job of                                                       Creep (33%,2)
        predicting failures on certain products?                  Morph FFDM                      Creep (67%,2)
       What improvements could be made to either                    Lever (30.0%)         Impact deformation (33%,1)
        method?                                                           Table 4 – Failure Data for the Crossbow

The goal of this research was to at least partially                                     Circular Saw
understand the strengths and weaknesses of RED and                Recorded Failure              Blade guard breaks
FFDM with hopes of finding areas that can be improved.            Function of Failure       Transfer Mechanical Energy
                                                                  RED                           Abrasive wear (5,1)
                                                                                              High Cycle Fatigue (5,4)
                                                                                               Impact Fracture (5,1)
ANALYSIS                                                                                           Yielding (4,2)
From inputting the functional models into RED and                                              Surface Fatigue (3,1)
FFDM, the result was three sets of failure data for each          List FFDM                  abrasive wear (14%, 30)
of the 14 products pulled from the Consumer Products                                        low cycle fatigue (6%, 12)
Safety Commission Recalls list. From the raw lists of                                         surface fatigue (4%, 9)
data, the key function-flow pairs that exhibited the              Morph FFDM                  abrasive wear (27%, 4)
                                                                     link (8.74%)           high cycle fatigue (20%, 3)
failures were isolated and analyzed. The condensed lists
of possible failures for each product are presented in the                                surface fatigue wear (13%, 2)

Appendix, while Tables 4 and 5 provide examples of the                                          yielding (13%, 2)
output for the Crossbow and Miter Saw. Note that for                                          ductile rupture (7%, 1)
the section outlining the MorphMatrix Failures, the                                          low cycle fatigue (7%, 1)
suggested artifact solution that matches the actual                       Table 5 – Failure Data for the Miter Saw
product is listed in the bottom left, along with the
percentage that product is used for the given function.       Initial results, however, are favorable that FFDM is
                                                              capable of predicting the possible correct failure of
Of the fourteen products analyzed, RED was unable to          consumer products, while RED is capable of correct
predict the failures for six, simply because the program      failure prediction approximately four out of seven times.
did not have any cataloged failures for the given function    Further studies will expand on the number of products
of failure for those products. The functions include          analyzed in order to provide more reliable numbers.
“change solid”, “actuate mechanical energy”, “position
human material”, “separate solid” (two products               OBSERVATIONS
exhibited failures in this function), and “store liquid.”     Throughout this process, several observations were made
                                                              that could help in improving the usefulness of FFDM and
Whether or not the product had output from RED, the           RED.
overall failure data proved hard to interpret. Data from
the CSPC website did not list exactly how the product         As the products were analyzed, our team was working
failed, whether the crossbow failed from the trigger          backward in the sense that products that had already
actually breaking, the return spring for the trigger          failed were decomposed and analyzed to see if the known
deforming, or any number of ways. Thus, the final lists       failures were identified. In some ways, this represents
of failures for each product are all possible failures for    the „hindsight is 20/20‟ fallacy. Because of this, we
                                                              observed that it might be difficult for a designer to utilize



                                                              4
the information presented by RED and FFDM when
moving forward with a design. While both methods can           Finally, purely from a usability standpoint, more effort
yield insight into potential failures, the designer can        into making the software user friendly and intuitive is a
become overwhelmed by the large amount of failures that        must for wider adoption of such methods. Our team often
are presented, which in turn makes it difficult to know        ran into errors with the software that was based solely off
which failure will be applicable. For many of the              formatting of text files. Aside from being rather
products analyzed, the real-world failures were rarely the     frustrating, it limited the desire to use such tools not only
most documented or highest percentage in the lists.            from our team, but by other students.

Furthermore, both methods only work for finding failures       CONCLUSIONS
of direct functions. This can be viewed two ways. First,       From the observations made while using RED and
both methods can‟t readily predict part connection             FFDM, there are several conclusions that can be drawn
failures, as in the case of the vegetable chopper where the    from this analysis.
blade falls in. Thus, some failures such as this example
can pose significant harm to humans but may be difficult       First, RED and FFDM do an adequate job of propagating
to identify by simply using the RED and FFDM methods.          large lists of failures that potentially contain the actual
Second, since both methods use functional models for           failure that occurred in the real-world. In many ways
what is supposed to occur as their basis, it can be hard to    this is a double-edge sword as it is great for guiding the
take into account for functions that were not intended.        designers though the design process and presenting
From our research, this arose with the example of the          problems they might not have considered. On the other
stroller which was amputating fingers when it was closed.      hand, since both programs work by identifying failures
Though the stroller was not built to cause human harm, it      for each function, the lists can quickly become lengthy
functionally did so. As such, if there was a way to add        and unmanageable even for simple products. For
the possibility of similar human harm to the failure           example, a product simplified down to just 10 functions
analysis, it would help greatly in identifying areas with      can output as many as 6 pages of listed failures. Running
such problems.                                                 such analysis on a complex system would most likely be
                                                               of little use, as it would be too hard to interpret the
In relation to RED, the failure data set used in this          multitude of data.
analysis method seemed to be incomplete and often gave
no relevant failure data. In general, this was a problem       Furthermore, the databases that both methods pull from
for both methods and is directly related to the strength of    are limited in cross-referencing capabilities, especially
the failure databases for both RED and FFDM. Similarly,        with the CPSC site. While both RED and FFDM use
for RED, the risk chart and list of failures seems to be the   standard function and failure taxonomies to run their data
same for each function-flow pair, suggesting that the          analysis, our team did not find the use of such
interaction between failures is not taken into account, but    taxonomies in adoption by agencies that report such data.
rather standardized across all products. This has obvious      Rather, CPSC uses non-standard methods for reporting on
problems as the complexity of each product is different        product recalls. While they are performing a service to
and the ability to analyze what could lead to bigger issues    the community by reporting such problems, they do not
that are important when trying to understand the risk of       allow for further analysis to be done by the engineering
failure in systems.                                            design community at large. As such, this makes the large
                                                               data repository of such organizations much less useful
It was also found that choosing different functions, even      then it could potentially be. To help with this, our team
when they might be very similar, can have significant          suggests better collaboration with groups like the CPSC
effects on the outcome of the failure analysis. This           to develop more organized and standardized product
brings into question the importance of which functions to      failure databases so that such information can be more
use, and the real meaning between each. This calls to          readily used in multiple areas.
question how experienced or versed the user has to be
with functional modeling and function taxonomy in order        Finally, since much of the failure data used by RED and
to differentiate between functions such as mix, change,        FFDM came from relatively complex systems, the failure
and separate.


                                                               5
database should be extended to ensure that the failure         included the word “material” and that FFDM could only
data in the system relates to the scale of the product.        analyze certain function-flow pairs.
Since much of the data currently in RED comes from
complex systems, the ability to use it on basic household      REFERENCES
products like this study might be extending it beyond its      [1]   K.G. Lough, R.B. Stone, and I. Tumer, “Prescribing
capabilities. However, as the previous conclusion states,            and implementing the Risk in Early Design (RED)
these issues could be remedied by pulling in regularly               method,” 2006 ASME International Design
updated and current failure information from groups like             Engineering Technical Conferences and Computers
the CPSC, thus, making the failure databases much more               and Information In Engineering Conference,
relevant and complete.                                               DETC2006, September 10, 2006 - September 13,
                                                                     2006, Philadelphia, PA, United states: American
Overall, it seemed like both RED and FFDM are steps in               Society of Mechanical Engineers, 2006.
the right direction. They provide useful information that      [2]   K.G. Lough, R.B. Stone, and I. Tumer, “The Risk in
can push the designer in a direction that avoids larger              Early Design (RED) method: Likelihood and
problems later on. However, to extend the usefulness of              consequence      formulations,”    2006     ASME
such methods, the key really lies in having failure                  International Design Engineering Technical
databases that are more comprehensive. This will allow               Conferences and Computers and Information In
increased completeness of the results, and give designers            Engineering Conference, DETC2006, September
more confidence in the methods themselves. Also,                     10, 2006 - September 13, 2006, Philadelphia, PA,
making the software more user-friendly and available is              United states: American Society of Mechanical
important for wider adoption. FFDM seems to be heading               Engineers, 2006.
in the right direction with the web-based interface, and it    [3]   R.B. Stone, I.Y. Tumer, and M. Van Wie, “The
would be very useful if RED had a similar                            function-failure design method,” Journal of
implementation.                                                      Mechanical Design, Transactions of the ASME,
                                                                     vol. 127, 2005, pp. 397-407.
                                                               [4]   K.G. Lough, R.B. Stone, and I. Tumer, “The Risk in
FUTURE WORK                                                          Early Design (RED) method: Likelihood and
Further research will soon include analysis of real world            consequence      formulations,”    2006     ASME
complex systems and whether RED and FFDM are able                    International Design Engineering Technical
to accurately predict those failures. This is similar to the         Conferences and Computers and Information In
structure of the study presented herein, but would include           Engineering Conference, DETC2006, September
products such as automobiles and heavy machinery.                    10, 2006 - September 13, 2006, Philadelphia, PA,
                                                                     United states: American Society of Mechanical
Further analysis should be done to include more                      Engineers, 2006.
consumer products in the product database in order to          [5]   K.G. Lough, R.B. Stone, and I. Tumer, “Prescribing
expand the results of which failures are predicted best.             and implementing the Risk in Early Design (RED)
The database of documented failures should also be                   method,” 2006 ASME International Design
expanded to contain as many products as possible,                    Engineering Technical Conferences and Computers
including more complex systems. The majority of                      and Information In Engineering Conference,
products within the database are limited in size and scope           DETC2006, September 10, 2006 - September 13,
currently, while the failures documented are mostly from             2006, Philadelphia, PA, United states: American
larger, complex systems.                                             Society of Mechanical Engineers, 2006.
                                                               [6]   “CPSC Home Page.”
Improvements to both programs should include a method          [7]   “Ryobi Corded Circular Saws Sold Exclusively at
of taxonomy and format checking or a debugger in order               Home Depot Recalled By One World Technologies
to minimize errors in the inputs. Initial analysis                   Inc. Due to Laceration Hazard.”
conducted for the research proved to be invalid when it        [8]   “Miter Saws Recalled by WMH Tool Group Due to
was discovered that RED did not recognize inputs that                Laceration Hazard.”
                                                               [9]   “Lysol Steam Cleaning Mop Recalled by Conair



                                                               6
     Corp. Due to Burn and Laceration Hazards.”                 Reannounce Recall of Trimmers/Edgers Due to
[10] “Lifetime Brands Recalls Fruit and Vegetable               Laceration and Burn Hazards.”
     Choppers Due to Laceration Hazard; Sold at Sam's    [16]   “OSRAM SYLVANIA Recalls Portable Nightlights
     Club.”                                                     Due to Electric Shock Hazard.”
[11] “Blenders Recalled by Haier America Due to          [17]   “MooreCo Recalls Ergonomic Office Chairs Due to
     Laceration Hazard.”                                        Fall Hazard.”
[12] “Laceration Hazard Prompts Recall by Starbucks of   [18]   “Bristol Model and Martha Stewart Collection
     Coffee Grinders; Made by Tsann Kuen.”                      Enameled Steel Tea Kettles Recalled Due to Burn
[13] “Fit & Fresh Mixers Recalled by MEDport LLC                Hazard.”
     Due to Laceration Hazard.”                          [19]   “Summit Treestands Recalls Hunting Tree Stand
[14] “Rifle Crossbow Recalled by Master Cutlery;                Brackets Due to Fall Hazard.”
     Crossbow Can Discharge Unexpectedly.”               [20]   “Wal-Mart Recalls General Electric Toasters Due to
[15] “Increasing Injuries Prompt Black & Decker to              Fire and Shock Hazards.”




                                                         7