Industrial Wireless Sensor-Actuator Networks (WSANs) are significant in modern industrial automation for real-time observation, communication, and management of industrial procedures. However, dynamic network conditions, congestion, limited available energy, variable quality of channels, and dissimilar requirements of various applications render scheduling tasks and resources a challenging issue. This paper introduces a smart scheduling system that combines Northern Goshawk Optimization (NGO) and Deep Reinforcement Learning (DRL) for efficient network resource management. The proposed system collects industrial data such as temperature, pressure, vibration and machine status as well as monitors network indicators including remaining energy, queue length, quality of channels, load of traffic, and the importance of packets. NGO is utilized for producing schedules, fitness evaluation, increase resource allocation and scheduling. The scheduling operations received the improvement from the DRL module that deals with observing the state of environment, estimating the reward, implementing the policy and selecting adaptive decision. Subsequently, an intelligent dynamic scheduling engine is employed and implements avoidance of congestion, transmission based on priorities, energy-efficient scheduling and real-time resource allocation. The framework provides scheduling actuators, sensors and emergency alerts in the process of industrial data transmission and controlling operations. OMNeT++ 5.6 is used for simulating the proposed system to test its efficacy in various environments related to industrial networks and various parameters are used to evaluate the proposed system. The DRL-NGO framework achieves a scheduling ratio of 95%, a network life of 1.9 s, and an average time of transmission of 30 ms, thus showing improvement in the scheduling capabilities in dynamic industrial WSAN environments.