Proposal and experimental case study on building ventilating fan fault diagnosis based on cuckoo search algorithm optimized extreme learning machine

• Published in: Sustainable energy technologies and assessments Vol. 45; p. 100975
• Main Authors: Xu, Yingjie, Chen, Ning, Shen, Xi, Xu, Liangfeng, Pan, Zhongyu, Pan, Fan
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.06.2021
• Subjects: Building energy; Cuckoo search algorithm; Extreme learning machine; Fan; Fault diagnosis; Fault diagnosis; Fan; Cuckoo search algorithm; Extreme learning machine; Building energy
• ISSN: 2213-1388
• DOI: 10.1016/j.seta.2020.100975

•A CS-ELM based fault diagnosis method is proposed for building ventilation fan.•Fault diagnosis experiment of two different fans is carried out on established rig.•The recognition rate of the CS-ELM is raised by 21.49% at most, reaching about 93%.•The training time, testing time and stability are also improved by proposed method. As the necessary auxiliary equipment in the fields of building energy system, the fan's healthy condition is very important for energy saving and public safety. Machine-learning fault diagnosis can improve fan performance and bring benefits of energy, economy, and safety. However, the vibration signal of the fan is particularly susceptible to interference from other factors, which brings great trouble to the existing machine-learning fault diagnosis Therefore, in order to learn useful fault features more effectively, a fan fault diagnosis method based on cuckoo search (CS) algorithm optimized extreme learning machine (ELM) is proposed in this paper. Firstly, wavelet packet extracts the original signal features. Secondly, with the help of the powerful search ability of the CS algorithm, the ELM model parameters are optimized to fully learn the signal features. Finally, the optimized ELM is tested through two cases. The experimental fault data of two types of fan are collected and employed to verify the effectiveness of this method, and the superiority compared with other existing methods. The results show that the proposed method can increase total recognition rate by at least 2.25% and at most 21.49%, indicating good progress and application potential.