Preparing the Aid System for Industrial Network Oriented Diagnostics - New Contributions in Information Systems and Technologies - page 277

Profibus. One aggregate has 9 AS-interface masters and each has the individual node
number in the Profibus network (all of them are the Profibus gateways). One of the
sub-networks responsible for palletizing the end product showed malfunction, and
immediately stopped the machine. After 2 days of unsuccessful investigation
Company “X” asked The University for help in service and maintenance of the
network.
Before the service visit, the maintenance engineer was asked to open the internet
link to the SUFINED system, login as a new user, create a new states feature vector
(SFV) and answer as many question as he could. On the grounds of this knowledge
(description of the environment) the authors of the paper could prepare well for the
service visit.
During the visit the gaps in the SFV were filled and the whole vector checked once
again. In accordance with the proposed scenario of the network-machine oriented
diagnostics, appropriate measurements, calculations and inference were carried out:

• At first, the reference measurement (with the use of the analyzer), in near-by

the area of the control cabinet. The number of detected modules was precise,
and also their types and assigned addresses in the AS-interface network. The
transmission factor was calculated ( W T = 5,98%) and the need of
maintenance indicated.

• The reasoning process based on the inputted data was completed.

• The environment of the module with the highest W T was checked.

• The mistakes and malfunctions were localized and eliminated.

• The final measurement for a final report carried out. The transmission factor

was calculated ( W T = 0,12%) and there was no indication of the need of
maintenance.

3.1 SUFINED System and Maintenance Time Reduction

During the service visit, the states feature vector (SFV) was created, basing on the
knowledge about the machine, network, and measured parameters such as
transmission factor W T . Next, the reasoning process based on SFV was completed. As
a result of the latter, the user was given the following information: 25 rules were
accepted for the reasoning process, in which 4 rules rendered a perfect match in both
reasons, rule 1 was half a match but the logical conjunction were “OR' type, so
finally, 5 conclusions were derived.
It is important to mention that the creation of a new states features vector of the
diagnosed system took about 15 minutes. The reasoning process was completed in 2
seconds, which can be omitted in the calculation time. Thus, such form of
nondestructive diagnostics is very fast (especially if compared with the two days
which the customer wasted looking for the reason of the malfunction).
The conclusions from the reasoning process showed possible reasons of problems,
such as excessive cables, and malfunction of the sensors connected to the module
with the highest transmission factor W T . Both conclusions were correct. In the control