Bombardier Transportation has developed a diagnostic system based on neural networks that aims to keep down the costs of maintenance of automated door systems on light railway vehicles.
The system, which was awarded US patent 6,636,814 today, monitors the condition of an automated door system to enable it to automatically let maintenance teams know exactly when maintenance should take place. Because the system ‘knows’ the normal and failure conditions of the door, it can determine whether the door is deviating from its normal mode of operation and calculate how soon it is likely to fail.
In use, data acquired from the door is fed to an A/D converter; once converted into a digital form, the data is processed before being sent to a CPU which analyses it using a neural network program.
Prior to its deployment, the neural network software has been ‘taught’ a variety of system conditions acquired when the system is running in both normal mode and when it is running into failure. Based on that knowledge, it is possible for the CPU to determine the rate of degradation of the system and schedule maintenance accordingly.
One benefit of the predictive diagnosis technique that the Bombardier engineers have developed is that it can predict door system failures far enough in advance so that the required repairs can be performed during scheduled maintenance periods.
This means that the automated door system will not be over-maintained, but only maintained when necessary, resulting in a reduction in maintenance costs, an increase in maintenance efficiency and an increase in system availability.