Data-Driven Evaluation of a Mamdani Fuzzy Duty-Cycle Controller for Solar Charging of a Lithium 18650 Battery-Powered Patient Monitor Prototype

Purpose: Portable patient-monitor prototypes based on single-board computers require a stable energy source, whereas photovoltaic charging varies throughout the day. This study evaluates, from field measurements, how a Mamdani fuzzy inference system maps photovoltaic conditions to a duty-cycle charging command, how this decision layer compares with alternative decision methods, and what the available solar energy implies for a lithium 18650 battery-powered patient monitor prototype. Methods: The open-circuit voltage (Voc) and short-circuit current (Isc) of a 10 W photovoltaic module were recorded on three days (74 records) through a 0.1 Ω shunt, and the shunt voltage was used to cross-check the current readings. Maximum available power was estimated with the datasheet fill factor. A two-input Mamdani controller with a complete nine-rule base (min-max inference, centroid defuzzification) generated a 0-100% duty-cycle command offline; the command was not applied to hardware as a PWM signal. The controller was benchmarked against a zero-order Sugeno system, a crisp rule-based system with the same rules, a proportional power map, and a fixed duty cycle using rule coverage, command variation, correlation with available power, robustness to simulated sensor error, rule-base completeness, and computation time. Result: The Mamdani controller produced a decision for all records and over the whole operating envelope, whereas the crisp version decided only 50.0% of records and 71.5% of the envelope. Its command followed the estimated power (Spearman ρ = 0.971) and was smoother and less sensitive to sensor error (RMSE 4.58 percentage points) than the Sugeno (5.47) and crisp (9.27) variants. Estimated daily energy was 56.5-66.5 Wh, about five 18650 cells, but estimated power never reached the 17.116 W estimated load. One 3200 mAh cell provided an estimated 37.4 min battery-only runtime. Novelty: The study provides an exploratory, data-driven evaluation that quantifies the quality of an interpretable Mamdani decision layer, namely its coverage, smoothness, robustness, and agreement with available power, against alternative decision methods, and places it within an energy evaluation of a patient-monitor prototype. It establishes a baseline for closed-loop hardware validation rather than evidence of charger performance or medical-device readiness.