Cyber-Physical Systems increasingly demand seamless coordination between human operators and autonomous processes, which increases the complexity. High cognitive workload in those environments amounts to a degradation of performance, decision fatigue, and increased susceptibility to system failure and cyber threats. To address these challenges, we propose a Neuro-inspired Cognitive Workload Optimizer (NCO), a novel machine-learning-based model for the monitoring, prediction, and optimization of cognitive workload for CPS performance improvement. The NCO framework employs neuro-inspired deep learning techniques, with LSTM networks coupled with an attention mechanism for assessing workload patterns dynamically in time. The adaptive operation of the system depends on executing a contextual analysis of system data and operator interaction metrics, whereby NCO recognizes fluctuations in workload and adjusts the operations of the system in real-time to maintain an optimal state for cognitive functioning. Thus, the model implements an adaptive feedback loop that prioritizes task distribution, resource allocation, and security management based on cognitive load estimations. In this way, CPS environments are hereby enabled to proactively mitigate operator overloads, minimize latencies, and enhance accuracy in decision-making, all while ensuring this is happening under dynamic conditions ensuring robust system performance. Experimental results on simulated CPS datasets indicate that NCO can reduce workloads peaks by 35%, improve system throughput by 28%, and provide better anomaly detection performance in conditions of high stress. The NeuroCPS-Optimizer thus opens up a new paradigm for cognitive-aware CPS management, ensuring that human and machine components are kept within safe and efficient bounds. This research thereby advances the creation of resilient and intelligent CPS that can self-adjust and sustain performance levels in complex and demanding environments.