The ISIS project deals with the development of a multi-modal, multi-sensory instrumentation system with build-in intelligence, based on neural network methodology, for measurement performance improvement. Following this objective, the Greek team has been working on the areas of defining the system's structure, the distribution of system functions on it's various levels and the implementation of two of the system functional blocks which could be used as a workbench of the final structure. A brief presentation of the proposed solution is made in the following pages.
System description and requirements
The desirable properties of the ISIS are:
-
The ability of supporting
- Multiple, different sensors
- Expandable number of sensors
- Adaptability to different instrumentation procedures
- Improved measurement accuracy by
- Compensating of systematic, systematic drift and random errors and
- Dealing with missed data due to random (transient or intermittent) fault.
- Reliability by performing
- Data Validation: Detect and compensate/remove undesired or corrupted data.
- Self-testing: Monitoring and checking the performance of the entire system to diagnose failures of system devices.
Those accuracy and reliability properties require the execution of complicated intelligent tasks by a Central Computer with adequate computational power. The adaptability property implies that the processing parameters of the systems functions or even some function algorithms should be easily updated. Finally, the extendibility of the system can be achieved by interfacing all system units through a multi-point central Data Bus.
|