Lightweight Adaptive Data Rate Adjustment for Cost-Aware Industrial IoT Monitoring

Industrial IoT (IIoT) deployments demand intelligent transmission schemes that balance monitoring fidelity with limited energy and bandwidth. This paper presents a lightweight adaptive framework that predicts a utility–cost slope via machine learning and refines it through Bayesian fusion with historical priors for anomaly-aware rate adjustment. Signal utility is derived from spectral and statistical features such as entropy and the Hurst exponent, while cost reflects link quality, anomaly likelihood, and SIM data usage. Deployed on a 342-day water-purification testbed, the method achieves up to 60% data reduction with negligible loss in reconstruction accuracy, demonstrating its practicality for resource-constrained IIoT edge devices.