Hybrid AI Models for Energy-Efficient Thermal Management in Smart Mechanical Systems

Thermal management in smart mechanical systems is critical for ensuring operational efficiency, reliability, and energy conservation. Traditional control methods often struggle with dynamic thermal loads and complex system interactions. This paper presents a novel hybrid artificial intelligence (AI) model that integrates Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks with an attention mechanism for energy-efficient thermal management. The proposed model achieves 94.7% prediction accuracy while reducing energy consumption by 32.4% compared to conventional approaches. Real-world validation using industrial datasets from HVAC systems, data centers, and automotive applications demonstrates the model’s effectiveness in predicting temperature distributions and optimizing cooling strategies. Experimental results show inference times of 8.5-18.3 ms under varying load conditions, making the solution suitable for real-time applications. The hybrid architecture leverages spatial feature extraction through CNN layers and temporal dependency modeling via LSTM units, unified through an attention-based fusion mechanism. This research contributes to sustainable industrial automation by enabling intelligent, adaptive thermal management with minimal computational overhead.

Hali tarjima qilinmagan