The efficient operation of intelligent manufacturing systems relies on precise modelling and scheduling methods to handle complex production environments and address various operational challenges. This work focuses on the application of deep neural network models in intelligent manufacturing systems, particularly on the construction of network energy functions and the resolution of optimization problems. By treating deep neural networks as mappings of individual neurons, the energy function E can be expressed. The use of deep neural network-based modelling and scheduling methods is proven to enhance system stability and optimization while effectively tackling challenges stemming from system complexity and uncertainty. This research presents a mathematical model that integrates the principles of electromechanical systems to capture the dynamic characteristics of both the electrical and mechanical systems of computer numerical control machine tools (CNC machine tools). This method not only optimizes component configurations and performance parameters but also enhances machining precision and responsiveness of machines through model predictive control techniques. Furthermore, by integrating knowledge-driven and data-driven intelligent diagnostic technologies, it addresses the challenges encountered by expert systems in handling new faults due to the lack of a comprehensive knowledge base, thereby improving the accuracy and predictability of fault pattern recognition. Through data analysis, the evaluation features and their characteristic values of matter elements are identified, validating the effectiveness of deep learning methods in system modelling and scheduling.