 |
 |
 |
 |
 |
 |
 |
||
|
|
|||||||||
Faq 35
35) I
am very interested in EXAKT and how it fits with other maintenanance technology
and software on the market. These are the points that I would like you to
elaborate on:
Focused information regarding your CBM
solution:
A) Our CBM solution comes in two parts. First there is an analysis and
modeling engine that fuses information from the CMMS and from the CBM
databases. It proceeds to look for statistically significant relationships
between the monitored data and the failure, potential failure, and preventive
replacement events that took place during a time period. Once these
relationships are found, they are incorporated into an equation (called a
proportional hazard model or PHM). The PHM relates the hazard rate, working
age, and monitored data. “Hazard” (Hazard rate x small time interval)
is the short term probability of
failure in a small time inteval.
“Working age” is any measurement that is considered to be
proportionate to accumulated stress on the asset, for example widgets produced
or fuel consumed. “Monitored data” can include internal variables such as
temperature, pressure, vibration, or oil analysis data. It can also include
external variables such as environmental factors or fuel type. Each monitored
variable is weighted in the PHM according to its influence (as found by EXAKT)
on the hazard rate.
Next, a predictive algorithm (called a transition probability model) is
joined to the PHM. Its purpose is to project the future values of the
significant monitored variables over an observation interval. Often, as
a practical observation interval or decision window, we chose the CBM
monitoring interval.
Finally we add an economic model so that subsequent CBM decisions will account
for risk. We define risk as the
combination of the consequences of failure and the probability that failure
will occur within the next observation interval. Next we "optimize" the
risk. The CBM data interpretation policy (model) will address an objective or
to an “optimal” combination of objectives. Low average unit cost, high
availability, high survivability, or a desired compromise amongst these
objectives must be achieved (over the long run) by adhering to the proposed
optimal data interpretation “rule”.
The second part of the EXAKT CBM solution is an "intelligent decision
agent". The agent deploys the models developed in the analysis and
modeling module of EXAKT (called “EXAKT for Modeling” or EXAKTm). The agent
(called “EXAKT for Decisions”, or EXAKTd) operates silently on the network,
responding to new CBM and CMMS data as it appears in its respective databases.
EXAKTd issues a remaining useful life estimate (RULE) and an optimal decision
recomendation. It places these results into a database from where the CMMS,
EXAKTd's own UI, or any other software will access them.
Real time data acquisition techniques
A) The EXAKT decision support system uses data from any source. For
continuous monitoring, you usually need a data reduction technique and often a
dimension reduction technique. The former is an appropriate way to summarize
continuous data so that you do not react falsely to noise or transients. The
latter is a way to summarize multiple variables into a few variables for
decision modeling. This is a rather large topic. Dr. Daming Lin of OMDEC is an
expert in this area. He has written an interesting survey that you can find at http://www.omdec.com/articles/p_surveyCBM.html.
Rules basis and general mathematical
approaches for CBM
A) Here is a case study of a fleet of nitrogen compressors, which has an introduction, that describes the general mathematical approach. http://www.omdec.com/articles/p_recipN2Compressors.html. See next answer for a discussion of rules basis.
Are the rules easily extensible by end
users or are they open?
A) Open. But EXAKT is not a rule based expert system. It is a
complementary decision tool that can be incorporated with an expert system.
There are two possible ways for this to happen. An expert system may reduce the
dimensions of data (the number of variables), reveal one or more failure modes
in progress, and issue respective severity ratings. A severity rating may be
tested and used by EXAKT as a "significant covariate". An example of
this is described at the end of the article http://www.omdec.com/articles/p_CBMDecisionMakingwithExpertSystems.html.
The second way for EXAKT to work with an EXPERT system is for the expert system
to access EXAKTd's remaining useful life estimation and optimal decision and
include these results into its rule base. EXAKT, thereby, deepens and extends
the expert system's knowledge base.
If so can they be easily defined from an
existing FMEA/RCM assessment?
A) Rules should be based on (and continuously updated from) the FMEA/RCM living knowledge base. The EXAKT Work Order Processor (EWOP) unifies the CBM, CMMS, and RCM databases to make this happen. It provides convenient access to the RCM knowlege base directly from the work order. A precisely structured "failure code" on the work order matches the item-function-failure-failure mode in the RCM knowledge base. When no RCM record exists, that accurately describes the current situation, the EWOP will update the RCM database, via audit trailed revisions, for subsequent validation by an RCM facilitator.
Does the product provide defined rules
for common equipment classes and their associated failure modes?
A) Yes. Models are developed for each failure mode/component for which CBM data patterns are found to be significant to failure or to potential failure. We call this "marginal analysis". Reports specify the combined health state of the equipment as well as the health state of specific modeled components.
If so what is the engineering basis and
what validation techniques did you use?
A) During model building with EXAKTm there is a report on PHM Parameter Estimation. To validate variable significance, the report shows the estimated value, the estimated standard error, the Wald statistic, and p-value. If a monitored variable is not significant or influential (to the hazard rate), it may be safely excluded from consideration by the decision model. For each parameter, including the shape (age) parameter, a 95% confidence interval is reported.
Once a PHM is built, various goodness of fit tests are applied. One of the tests, the Kolmogorov-Smirnov test, reports a p-value and acceptance based on a 5% significance level. This means that the model relating hazard to the weighted combination of significant monitored variables (and working age) will be the appropriate one 19 times out of 20.
Finally the ultimate decision model may be tested using EXAKT's cost comparison function. The method is described in: http://www.omdec.com/articles/p_CBMbenefitsAnalysis.html
Who owns the IP?
A) OMDEC, the Ontario Center of Excellence, and the University of
Toronto. OMDEC, however, holds the sole commercial license to develop and
distribute the software.
Are they supported, maintained and
updated with improvements?
A) Yes. On the average every six months.
What are the common conditional inputs?
(process readings, failure experience, etc.)
A) EXAKT is a general purpose CBM software. It uses whichever inputs determined by the PHM to be significant risk variables. All of the inputs you mention have been used.
What are the typical outputs? (alerts, recommendations, maintenance
actions, etc.)
A)
General
technical architectural information
A) EXAKT is a Windows program that reads inputs from any database and writes its outputs to any database. Its internal data model is relatively simple and can be mapped to MIMOSA or any other schema.
Any further information that you may
have about how a software company could embed the CBM engine, etc.
A) There are no issues regarding integration of EXAKT. It can be accomplished in many ways. OMDEC has developed the EWOP (EXAKT Work Order Processor) to bridge the human-CMMS-CBM entities in a way required for reliability analysis.
|
