Purpose The use of various types of unmanned aerial vehicles (UAV) has expanded significantly due to their characteristics. The use of these systems in various fields requires their optimal control. Additionally, controlling these systems necessitates the detection and identification of the types of faults that may be encountered. Accordingly, in this article, bidirectional long short-term memory (BiLSTM) has been used to diagnose actuator faults in the quadrotor system. Also, the integral sliding mode control (ISMC) controller has been used as a fault-tolerant control (FTC) of this system. Design/methodology/approach In this study, the quadrotor system is first modeled and then the required data are obtained from the flight patterns. Next, the new (BiLSTM) method is used to identify the actuator fault of the quadrotor system. BiLSTM is a subset of deep learning and, unlike classical methods such as observer, due to its intelligent features, it can respond to new conditions and changes in system behavior with appropriate training. Findings BiLSTM as a subset of deep learning increased the accuracy of actuator fault detection in the quadrotor system. The use of BiLSTM allowed the severity and location of the fault to be separated and classified. ISMC was used for FTC. Finally, the accuracy of the network for fault detection and isolation reached 96.08% and the results of the ISMC control performance are also favorable. Originality/value One of the main aspects of BiLSTM is the use of adaptive learning to enhance fault recognition in various conditions. In this research, after using and training the BiLSTM network to detect and isolate the actuator fault, the ISMC is used as an active FTC. The main advantages of ISMC are eliminating steady-state errors, and it’s able to improve robustness.
Erfanian et al. (Tue,) studied this question.