An Edge-Based Cyber-Physical Architecture for Resilient IoT pH Sensing: Mitigating Non-Linear Thermal Hysteresis via a Lightweight Hybrid Residual Soft Sensor

Integrating digital technology into the industrial sector requires robust cyber-physical systems supported by scalable edge-computing infrastructures. Real-time monitoring using traditional pH sensors is susceptible to the effects of non-linear thermal hysteresis. This study proposes an IoT-enabled hybrid soft sensor that combines a physical sensor model with a machine learning-based residual error estimator model at the network edge. Utilizing ESP32 and Raspberry Pi hardware, pH sensor signals are pre-processed using Hampel and Kalman filters before being fed into an XGBoost model to determine the estimated residual error. Once optimized, the system achieves R² score of 0.999976, and RMSE of 0.010927 pH. These results demonstrate how smart technologies can enhance the measurement quality of physical hardware in industrial environments.