This study introduces a novel method for exploring manufacturing parameters through digital twin simulation, utilizing three comparative neural network architectures: CNN, LSTM, and a hybrid CNN-LSTM model. Utilizing the AI4I 2020 predictive maintenance dataset, we illustrate the efficacy of digital twin technology in simulating industrial equipment behavior and offering clear visualizations of parameter relationships. Each model architecture captures distinct aspects of equipment behavior: CNNs are proficient in identifying spatial patterns across parameters, LSTMs are expert at capturing temporal sequences, and the hybrid model integrates these strengths. The trained AI models build a simulation system in a virtual environment where abstract parameters are turned into intuitive representations. This allows operators to investigate manufacturing parameters within a risk-free virtual environment. Experimental results indicate that the hybrid CNN-LSTM model attained superior performance, achieving an mAP50 score of 63.8%. In contrast, the LSTM-only model scored 57.2%, while the CNN-only model reached 51.9%. This underscores the efficacy of integrating spatial and temporal feature extraction capabilities.