Intelligent flux monitoring system using fuzzy logic: Flow-fault detection and time-to-empty validity assessment

This research aims to develop an intelligent flux monitoring system using fuzzy logic to support continuous flux discharge and assess the operational validity of time-to-empty (TTE) estimation during abnormal flow conditions.The problem addressed is the reduced reliability of conventional physics-based TTE estimation when the monitored system operates under faulty conditions, including supply-tube blockage, nozzle error and unstable flow.The study proposes a hybrid fuzzy-deterministic framework in which fuzzy logic functions as a diagnostic layer that complements the conventional mass-flow relationship.The framework uses flux tank weight and flux flow rate as measured inputs, whereas supply-tube-blockage and nozzle-error indicators are treated only as external reference labels for validation.The experimental record contains 500 observations sampled at 60-s intervals over 8 h 19 min, comprising 454 normal observations, 30 supply-tube-blockage observations and 16 nozzle-error observations.The Mamdani fuzzy inference system uses explicitly defined membership functions, a 16-rule base, minimum AND/implication, maximum aggregation and centroid defuzzification.All labelled fault observations have zero recorded flow, whereas normal observations range from 3.652 to 6.539 g/s; consequently, the fuzzy classifier and a fixed-flow threshold both achieve 100% sample-level accuracy on the available record. Deterministic TTE is defined as 1000 W/(60 Q) min for Q>0 and is treated as invalid when the fuzzy diagnostic index indicates abnormal flow.Using an empirical within-run depletion horizon derived from the tank-weight trajectory, normal-condition deterministic TTE gives MAE = 215.48min, RMSE = 248.85min and MAPE = 86.49%.These values are within-run consistency diagnostics rather than independent predictiveaccuracy estimates because the reference is derived from the same experiment and the recorded flow channel has an unresolved scale discrepancy.The results support the use of fuzzy logic as an interpretable flow-fault and TTE-validity gate, while independent TTE accuracy, physical sensor-noise robustness, fault-type isolation and PCB-quality improvement require calibrated and repeated experiments.