NEMD :
Pc based diagnostic tool “Neural Engine for Machine Diagnostic‘s (NEMD)” is an analytical engine to learn patterns in data and classify them. The patterns can be in the form of a time sequence (example: vibration signals from a transducer, temperature and pressure signals from sensors etc.).NEMD is a user friendly and generalized neural network engine, which can be used to train any specific task among in a variety of applications. It can be used in every situation where an Artificial Intelligent Application is possible.
For example :
NEMD can be trained for a specific application say, given the engine vibrations it can be used to detect a certain fault, example slight unbalance. All that has to be done is feed several samples of signals containing unbalance along with samples of signals representing normal operating conditions then NEMD will train itself to recognize the pattern of signals containing such a fault subsequently NEMD will be able to correctly classify the fault in any future signal when fed to it, as an unbalance or as normal signal.
Example Use of NEMD for PUMP Diagnostics :
Application Problem: Machine problems in the joint.
Description: This problem consists of two machine patterns (normal condition, machine problems in the joint).each pattern will have six parameters related to amplitudes of vibrations at different frequencies etc. Given many such samples measured during the actual running of the pump, NEMD learns to identify the patterns and recognize the fault with an accuracy (success rate) between 95% to 97%.
Typical Front end of NEMD System:
NEMD acquires vibration signals and displays displacement, velocity or acceleration as a function of time or as function of frequency.
The data can come into 4 channels in real time from suitable transducers / pickups.
NEMD is not only a monitoring tool but it is also a diagnostic tool.
NEMD provides tools to select different neural networks and configuration tools to design any desired neural architecture. Useful graphical views to see “training error” graph and tools to train, verify and test the data patterns are available