Machine self-learning techniques are employed to devise programmable logic controller (PLC) programs, utilizing exemplary manual sampling data to train the machine algorithms. These algorithms then compute PLC parameters, which are promptly adjusted. As machine learning deepens, the discrepancy between measurements obtained from an online water sediment content monitoring system and those from manual sampling diminishes, thereby enhancing the system’s measurement accuracy. In this study, the sensor, under the guidance of truth value samples, effectively approximates the true value. The objective is not to achieve an exact match between a specific measurement and the truth value but to maximize the likelihood of achieving the smallest possible gap between the truth value and subsequent measurements. This artificial intelligence-based machine learning algorithm significantly broadens the scope of intelligent sensor applications, offering versatile and practical deployment scenarios.