Next Page Prev Page Main Index Home

Real World Problem #2 "Risk Check"
Determine which parts are at risk.

This problem is aircraft-part tracking. It is to determine parts or assemblies that may be at risk after a specific part or assembly fails. As the parts move through the factory each develops a history. This includes many batch numbers one for each operation performed in manufacture. Any part not meeting the specifications for an operation will move to a new batch for rework. This creates a large and tangled web of interactions.

The problem query is to find all parts that have ever shared any single batch with the target part. The database must then allow questions to be asked about all parts sharing that batch, including where they are now and how many other shared manufacturing operations or components do they have.

This query must also allow qualifications to limit the search to likely candidates. This query must range from as narrow as parts that had threads cut with a particular die, to as wide as any part or assembly containing metal from a particular foundry.

When you look at all of the parts and assemblies in all of the possible aircraft, this is a monumental task. Such tasks are perfect for MPbase. The combination of internal characteristics found in MPbase make this a doable, if not an easy task.

First, the multidimensional hypercube nature of MPbase allows the relevant data to be logically located near each other. In this case only two linked cubes would be needed; one for the part or assembly and the other for the operation and batch.

Second, the content-addressable memory (CAM) nature of MPbase will cause all of the part records with similar histories to be stored together in the hypercube. Also, all of the similar operations will be stored together. This allows a relatively easy traversal of this complicated net.

Third, the massively parallel architecture allows for the sheer volume of data that such a database will need to store. This is helped significantly by the internal compression of the data. These two characteristics provide the needed performance.

Fourth, the search could be expanded to include all parts or assemblies that share similar operations or tools. This would allow the search of all parts that were ever touched by a specific die or tool, even if they did not share any common batch.

The natural information-based clustering found in MPbase makes short work of this type of very tricky data problem. It does so by turning it into a simple information-based problem.

In short, with MPbase this nasty real world problem is no problem.

Next Page Prev Page Main Index Home © 1998-2004 NPSI