Digital Infrastructure Optimization: Aligning System Architecture With Scalable Performance

Abstract

The digital infrastructure becomes crucial for smart manufacturing, which uses massive amounts of data generated by industrial IoT (Internet of Things). However, the current manufacturing infrastructures face challenges in processing large volumes of heterogeneous data and are unable to provide intelligent high-performance analytics by incorporating the DE-Trans WLSTM. Data is collected using industrial sensors that have been installed inside the machines and monitor different parameters like the temperature, vibration, pressure, and state of the machine. Standardisation using z-score normalization is used as the Preprocessing method in order to regulate the heterogeneous input data and avoid sensitivity to noise. Feature Extraction is done using the discrete wavelet transform which has the ability to decompose the time-series signal into various resolution levels to identify the transient variations as well as steady-state conditions of machine performance. The proposed DE-Trans WLSTM uses differential evolution to tune the parameters, Transformers for context-based learning, and the weighted LSTM for modelling temporal dependencies. Therefore, with this model, predictions on machine failures and workload are performed accurately. Microservices, containerisation, and intelligent resource management have been incorporated into the system to enable scalability, fault tolerance, and low-latency performance. Simulated results of the proposed method significantly overcomes the traditional models, achieving 98.55% prediction accuracy in the Python. The integration of advanced deep learning with optimised infrastructure provides a robust solution for Industry 4.0 environments, facilitating reliable, scalable, and data-driven manufacturing operations.