Design and optimisation of an LSTM – MCPO control system for shaft-type construction hoists

Shaft-type construction hoists play a critical role in high-rise construction, where operations must be precise and safe under varying loads and unpredictable conditions. Conventional control methods often exhibit degraded performance due to nonlinear dynamics, load changes, and external disturbances. To address these limitations, this study proposes an intelligent hierarchical control framework integrating Long Short-Term Memory (LSTM) neural networks with a Multi-Constraint Parrot Optimisation (MCPO) algorithm. Adaptive control commands for hoist motors are generated using real-time sensor data, including cage position, velocity, and tracking error. The LSTM network models the temporal behaviour of the system to predict optimal control actions, while the MCPO algorithm optimises controller parameters by minimising performance metrics such as steady-state error, settling time, and overshoot. The architecture consists of a high-level intelligent controller that generates reference commands and a low-level proportional-integral controller that ensures rapid actuator response and disturbance rejection. Simulation results demonstrate the effectiveness of the proposed system: under no-load conditions, the steady-state error is 0.20% with a settling time of 1.5 s, and under full-load conditions, the steady-state error is 1.20% with a settling time of 3.5 s, with overshoot of 0.9% and −0.8%, respectively. These findings indicate that the intelligent control strategy enhances tracking accuracy, stability, and operational reliability, making it suitable for safe and efficient control of shaft-type hoists in high-rise construction applications.