Real time techniques and models to save energy in compressed air systems

  • Mohamad Thabet

Student thesis: Doctoral Thesis

Abstract

New systems were created to assist in improving the energy efficiency of Compressed Air Systems (CAS) to fill gaps which were identified in the research.
A new mathematical model was created that coupled supply and demand sides for the first time. The supply side produced, treated and stored compressed air, while the demand side delivered and consumed compressed air. Changes to pressure regulation and storage tank size were evaluated. Model Predictive Control (MPC) was compared to ProportionalIntegral (PI) control and MPC had a more stable and lower system pressure, and that would result in some energy savings.
A discrete wavelet transform extracted information about shapes in the supply side pressure signal that were associated with events. High frequency events related to tools were isolated from low frequency events associated with tank charging and discharging. A nearest neighbour classifier was created to detect patterns generated by different tools.
Patterns in a regulated line were also investigated and an algorithm for the automatic identification of tools was created. The algorithm segmented data into smaller sub-sections containing patterns of interest. Two methods for classifying patterns were investigated, a rule-based and a distance-based method.
Pneumatic tools were also identified from their sounds. Audio was divided into four categories: valve activation, cylinder activation, valve and cylinder simultaneous activation, and no active tools. Cumulative amplitudes within frequency sub-bands were generated as features using a Discrete Fourier Transform. A neural network was created to identify tools using the features.
A condition monitoring and fault diagnosis system was created that compared outcomes from the subsystems monitoring supply side pressure, demand side pressure, acoustics and the schedule of operations. That successfully demonstrated that faults in individual tools or system faults could be detected, identified and solutions suggested.
Date of AwardMar 2022
Original languageEnglish
Awarding Institution
  • University of Portsmouth
SupervisorDavid Sanders (Supervisor), Giles Tewkesbury (Supervisor) & Nils Bausch (Supervisor)

Cite this

'