Loading...

Develop a Monitoring Center Conceptual Framework for Chain Stores

Kishani Farahani, Masoud | 2023

55 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 55876 (46)
  4. University: Sharif University of Technology
  5. Department: Energy Engineering
  6. Advisor(s): Rajabi Ghahnaviyeh, Abbas
  7. Abstract:
  8. Among the equipment used in supermarkets, refrigeration systems should be the focus of energy efficiency initiatives; Because they are the biggest consumers of energy and refrigerant with significant maintenance costs. Fault detection and diagnosis (FDD) can provide considerable potential for energy savings as well as reduced maintenance costs. Although there have been numerous investigations of FDD for HVAC systems, there has been very little research on the application of FDD for supermarket refrigeration systems. Therefore, this thesis focuses on the application of FDD to these systems and helps to fill these research gaps. The studied system is a commercial refrigerator system on a laboratory scale, by which the evaporator fan failure is simulated and the required data is collected. The data is preprocessed, and using the random forest algorithm, an FDD model was developed and the importance of each sensor's feature in detecting and diagnosing the fault under study is obtained. The compressor discharge temperature sensor has the most important features compared to other sensors. Using the feature importance ranking method, new FDD models are developed using random forest and KNN algorithms with fewer sensors, and their performance is evaluated based on the F1_score criterion. To investigate the robustness of new FDD models, noise is injected into the most important sensor in three signal-to-noise ratios, and the robustness of each model is checked. Finally, among the new FDD models, the most optimal evaporator fan fault detection and diagnosis model is selected in terms of the best performance in the mode without noise injection, the highest robustness in the three signal-to-noise ratios, and the least number of sensors. The optimal FDD model using the random forest algorithm with only four temperature sensors and without using pressure sensors, with 100% performance in the mode without noise injection and a robustness higher than 90% in three signal-to-noise ratios, can detection and diagnose the fan evaporator fault of refrigeration systems
  9. Keywords:
  10. Machine Learning ; Fault Diagnosis ; Fault Detection ; Feature Selection ; Robustness Evaluation ; Supermarket Refrigeration Systems ; Evaporator Fan Failure ; Chain Stores

 Digital Object List

 Bookmark

No TOC