Research Output
Condition monitoring of hydraulic systems using neural network for data analysis
  Condition monitoring of engineering processes or equipment has become of paramount importance as there is a growing need for improved performance, reliability, safety and more efficient maintenance. Condition monitoring in railway industry as a whole covers a very wide field. To restrict the field we have confined ourselves to the non-intrusive monitoring of hydraulic systems. This thesis is mainly concerned with the investigation of the non-intrusive method based on ultrasonic concepts and neural networks for rapid condition monitoring and/or fault diagnosis of the hydraulic systems.

A comparison between diagnosing hydraulic systems and electric systems is made. The location of faults in hydraulic systems is more difficult. The key to fault finding in hydraulic systems is the location of pressure. The development of pressure measurement instruments is reviewed. In case of trouble-shooting hydraulic systems, pressure readings are often required to be taken at several temporary locations. Since the hydraulic system is fully sealed, the direct measurement instruments can not be practically utilised for this purpose unless they are built-in during the production stage of the system. Instead, the indirect pressure measurement systems can be very helpful for rapid diagnosis of hydraulic systems. The new approach is a combination of the acoustic effect of the fully sealed oil inside the pipe and the penetrating capability of the ultrasonic waves. The ultrasonic wave energy enters the interior of the hydraulic piping and passes through the contained fluid, of which the pressure is being measured.

Two modelling approaches for this non-intrusive pressure monitoring system have been presented based on FLNN and MLP respectively. They offer the ability to establish the direct and inverse models. For both methods the maximum relative error (%FS) achieved for either the direct model or the inverse model is well within 2 %FS in our case studies. However, compared to the MLP, the FLNN provides a reduced cost of computational complexity.

The novel non-intrusive measurement of hydraulic pressure based on ultrasonic concepts offers the capability of making pressure measurements for trouble-shooting without intruding into the pipe. It is specifically designed for rapid diagnosis of hydraulic equipment, where the conventional measurement instruments fail to make the necessary pressure readings within the sealed pipes. This has the advantage of not having an effect on the condition of the sealed hydraulic system and also of assisting rapid trouble-shooting to save time and cost. Testing the pipes with such a non-intrusive technique is of great interest to all metal pipe related industries for the provision of no disruption to pipe operations.

  • Type:

    Thesis

  • Date:

    30 September 2006

  • Publication Status:

    Unpublished

  • Library of Congress:

    TC Hydraulic engineering. Ocean engineering

  • Dewey Decimal Classification:

    627 Hydraulic engineering

  • Funders:

    Edinburgh Napier Funded

Citation

Yu, F. Condition monitoring of hydraulic systems using neural network for data analysis. (Thesis). Napier University. Retrieved from http://researchrepository.napier.ac.uk/Output/2254714

Keywords

condition monitoring; hydraulic systems; modelling approaches; MLP; FLNN; non-intrusive measurement; hydraulic pressure

Monthly Views:

Available Documents