In most cases when a fault develops on a complex system, current monitoring systems will generate a string of error codes. It is then up to the engineer to interpret these alarms, error codes and messages and troubleshoot the problem. This can still take hours, as he has to access the manual, read the circuit diagrams, check control signals, perform electrical, hydraulic and mechanical tests all the while keeping track of what he is doing and slowly narrowing down the problem to the root cause.
Normal monitoring systems just monitor the status of on-board components, but don’t do any active diagnosis.
With iTech, you get a built-in expert. The entire logic and all the device information and configurations are stored in the diagnostic database. Therefore when a fault occurs the system goes into diagnostic mode.
If there are any alarms present the system loads them into its logic and immediately knows what they mean. From this it can identify the root alarm.
Once the root alarm is known, the system can identify a list of suspects devices, all those devices that could be causing that alarm or symptom.
iTech can now query the control program for any additional information on inputs and outputs (automated testing). This information is used to further eliminate certain suspect devices and reduce the list of remaining items to test.
Finally it can guide the engineer through a list of manual tests with detailed test instructions of the remaining suspect devices until the root cause of the failure is identified, isolated and rectified.
Model-based reasoning is used as the basis of all diagnostic applications built in iTech.
It uses the logical connections between components with forward and backward chaining and the optimization techniques described below to dynamically update and optimize its diagnostic strategy or test pattern.
The key to any Expert System is the correct calculation of probabilities (or, certainty factors). In iTech, probabilities are crucial since they are used to guide both the troubleshooter (using Information Theory) and ATE (using Heuristic Search).
In iTech, the probabilities are based upon a generalisation of classical Bayesian Probability Theory. This generalisation – called Dempster-Schafer Probability Theory – models the uncertainty, which is inescapable when modelling unknown problems (i.e. faults) in very complex systems (e.g. military electronics or automobiles).
iTech, for example, uses Dempster’s Rule (for combining multiple symptoms and test results) to model the possibility of multiple failures.
iTech can interactively guide a novice technician through each step of an actual troubleshooting session. iTech’s interactive advice is based upon Claude Shannon’s Information Theory; and combines: (a) the Cost(or time) of tests and setup procedures; and (b) the Probability of isolating the fault.
After deducing possible faults for test results and symptoms, iTech concludes automatic Expert Rule generation by inducing probabilities for these possible faults using a statistical Reliability Database (or, Failure Rate Database). iTech automatically builds and maintains this internal Reliability Database.
Each time that iTech is used, it remembers what has failed. It uses this to automatically re-adjust both its internal Reliability Database and its generated Expert Rules. This unique iTech learning algorithm is based upon the work by the famed mathematician Pierre-Simon Laplace.
Siemens and Thyssen Krupp developed and installed the first commercial MagLev train system
in Shanghai, China. The Transrapid started service in Shanghai, China in 2003 and utilizes
magnetic levitation and linear synchronous motors to propel it on its track at speeds of up to 500
km/h (312 mph).
Imagine this scenario A train is flying along at 450 km/h (280 mph). The cooling system for one of the converters that supply the power springs a leak, resulting in a loss of power and damage to expensive components if not fixed immediately. Before the passengers are aware of the situation, the realtime diagnostic system alerts the service department of the problem and within minutes the service engineers are on their way to the converters 30km from the maintenance area with the required components and tools and repair the problem immediately. How is this possible? iTech - Remy’s artificial intelligence diagnostic system - is connected to Siemens’ monitoring system through which it receives error messages from the BIT (built-in-test) for over 120,000 components in real-time. iTech uses its diagnostic logic to interpret and make sense of all these messages and arrive at a root cause or an ambiguity group. Normally an engineer would have to read through all the error messages and try to make sense of the specific combination of errors to arrive at the root cause of the failure. However iTech does this automatically and within seconds purely by understanding the system logic and dependencies. iTech also has a learning capability and as the number of failures increases, it becomes faster at diagnosing faults. iTech can also be used as an off-line diagnostic and does not necessarily have to be embedded or connected to a system. It can be used as a troubleshooting guide for a field engineer on a PC or even over the internet on a mobile device. iTech makes use of statistical theory, probabilistic reasoning, dependency modeling and other techniques to optimize diagnostics and troubleshooting approaches thereby resulting in the following proven benefits: • 70-80% reduction of downtime • MTTD(Mean-Time-To-Diagnose) reductions of up to 97%. • Improved troubleshooting accuracy • Improved learning curves for technicians