Accurate gas mass flow allocation is important for multi-component gas mixing systems and fuel-cell test benches, especially when a wide total-flow range is covered by several parallel mass flow controllers (MFCs). This study evaluates the coordinated allocation of three parallel MFCs using traceable third-party calibration data and a reproducible numerical simulation workflow. The calibration curves define a constrained flow-allocation problem in which several MFC setpoint combinations can satisfy the same total target flow, while producing different full-scale error levels. A modified particle swarm optimization (PSO) strategy is compared with a rule-based allocation strategy and standard PSO. The numerical evaluation was extended with 500-run reliability statistics, ablation analysis, parameter sensitivity analysis, expanded switching-region evaluation, numerical perturbation checks, and benchmark optimizer comparisons. Across the main 12 operating points, the probability of obtaining a solution within 0.10 %FS (percentage of full scale (%FS)) increased from 0.207 with standard PSO to 0.310 with the improved PSO, and the probability within 0.05 %FS increased from 0.135 to 0.193. The positive-improvement rate relative to the rule baseline increased from 0.384 to 0.547. The results support the use of calibration-curve-based optimization to reduce switching-region error peaks and to improve repeated-run low-error solution probability in parallel-MFC allocation.

Electric dynamometer test benches generate heterogeneous multi-channel logs whose structural completeness, physical consistency, and operating-condition support vary across files. We propose Coupling-Gated Data Quality Index (CG-DQI) sample weighting, a fold-local preparation protocol for 60 s next-window temperature-change prediction. Deployable weights use training-side structural and physical quality signals, while a target-magnitude coupling gate screens candidate association with the absolute target. The dataset contains 30,093 valid windows from 134 files and 30 date-defined campaigns. File-level leave-one-file-out validation reduced Ridge MAE from 0.7035 to 0.6932 °C and the mean MAE across three fixed Random Forest initializations from 0.4556 to 0.4489 °C per window. Strict nested leave-one-campaign-out analysis retained sparse CG-DQI in all 30 folds and yielded a campaign-level Ridge MAE difference of 0.0152 °C with a 95% bootstrap confidence interval (CI) of 0.0054–0.0309, while the 99th percentile absolute error (P99) increased by 0.0710 °C. Fixed-parameter XGBoost preserved a positive mean MAE direction, although its campaign interval crossed zero. CG-DQI, therefore, provides a traceable MAE-oriented preparation option within the present bench and campaign range; tail-sensitive and cross-device use requires an explicit constraint or recalibration.