Full-Process Temperature Prediction in Multi-Layer Robotic Grinding of High-Manganese Steel Under Limited Online Sensing

Thermal accumulation is a critical constraint in robotic grinding of ZGMn13 high-manganese steel, whereas the variables that can be prescribed or monitored reliably online are often limited to the normal load Fz, spindle speed n, and feed speed νw. Most existing studies focus on single-pass conditions or scalar thermal indicators, while full-process near-surface transient temperature histories in multi-layer robotic grinding remains insufficiently addressed. This study presents a full-process near-surface transient temperature histories framework for multi-layer robotic grinding under fixed wheel-workpiece conditions and limited online sensing. Multi-channel near-surface thermal measurements were first reorganized into layer-resolved time-series data. A process-driven thermal surrogate was then constructed from the deployable inputs Fz,n,νw, and a recursive temperature-evolution model was developed by incorporating intra-layer thermal retention and interlayer residual-heat inheritance. The proposed formulation predicts the near-surface transient temperature history over successive grinding layers. Experimental results showed clear layer-wise transience and progressive thermal accumulation during multi-layer grinding. Under representative conditions, the proposed framework reproduced the dominant transient structure of the measured full-process near-surface temperature histories, and grouped validation further showed that the recursive formulation preserved more useful history-level information than the reduced baselines within the tested domain. Within the tested operating domain, the predicted histories were further reduced to derived thermal indicators and planning-oriented peak-temperature maps.