IPORF: A combined improved parrot optimizer algorithm and random forest for fault diagnosis in AUV

Autonomous Underwater Vehicle (AUV) has many applications in ocean exploration and underwater operations. However, AUV are susceptible to failures due to internal and external factors when operating in complex underwater environments, which seriously affects their mission execution and reliability. To timely and accurately diagnose the type and severity of AUV faults, this study proposes a method IPORF combining the improved parrot optimization (IPO) algorithm and random forest (RF) for AUV fault diagnosis. Experiments were conducted on a real AUV dataset, and the experimental results showed that the IPORF method was able to identify faulty and normal states with 99.59% accuracy; it was able to differentiate between five types, Normal, Add Weight, Pressure Gain Constant, Propeller Damage Slight and Propeller Damage Bad, with 98.78% accuracy. Compared with 17 advanced algorithms on the same dataset, the accuracy of IPORF is improved by a minimum of 0.42% and a maximum of 29.23%, the precision is enhanced by a minimum of 2.10% and a maximum of 88.36%, the recall improved by a minimum of 3.28% and a maximum of 34.83%, the F1-Score improved by a minimum of 0.64% and a maximum of 62.22%. The outstanding fault diagnosis capabilities demonstrated by the IPORF suggest that it offers a versatile and straightforward framework for diagnosing faults in AUV using various types of sensor time series data, making it a valuable tool for practical applications.