Trust-Aware Models For Machine-Mediated Human–AI Interactions

Abstract

The rapid growth in industry resulted in the development of intelligent manufacturing systems that combine automation, Artificial intelligence (AI), and collaboration. Predictive Maintenance (PdM) systems available so far lack trust mechanisms and human-AI interaction capabilities. To solve this problem, this research introduced a trust-aware PdM approach by designing a hybrid Giant Trevally-optimized Intelligent Bidirectional Long Short-Term Memory (GT-IntBiLSTM) model. For this purpose, the Industrial Equipment Monitoring Dataset from Kaggle is used, which consists of multivariate data. Data preprocessing employs the Isolation Forest (IF) algorithm for removing outliers and a robust scaler for normalization. Independent Component Analysis (ICA) is used for feature extraction of latent features in industry data. The IntBiLSTM algorithm captures bidirectional dependencies within machine behavior and predicts faults and health status of machines. The GTO technique optimizes parameters used by learning models to improve learning capabilities. The proposed model provides trust-aware decision-making aids, helping people make informed choices about the maintenance of equipment and machines. The experimental findings prove that the suggested method outperforms the baseline methods regarding Root Mean Square Error (RMSE) of 30.178, Coefficient of Determination (R²) of 0.978, and Mean Absolute Error (MAE) of 20.034. The proposed model is implemented using Python-based Deep Learning (DL) tools which emphasizes the significance of incorporating trust-aware learning into advanced DL algorithms.