Design of a Novel Predictive Technique to Estimate Liquid Level and Concentration Using Multi-Sensor Data Fusion

Abstract

Level is one of the important parameters to be measured in many of the industrial applications, in which Capacitance Level Sensors (CLS) play an important role in measuring liquid level because of its ruggedness. But CLS is inept at measuring levels irrespective of liquid types hence this work proposes a novel technique to overcome the demerits of conventional measurement of CLS. This paper aims to develop a novel predictive technique to measure the liquid level along with its concentration. The predictive technique comprises a Level Predictive Model (LPM) and a Concentration Predictive Model (CPM). LPM predicts an accurate level for the change in liquid type and CPM predicts the change in sugar concentration of a liquid. An experimental setup was established for the model development. LPM was developed by infusing the data fusion technique from the data obtained by three different sensors CLS, Pressure Level Sensor (PLS) and Ultrasonic Level Sensor (ULS) for adaptive level measurement. CPM was developed using a Support Vector Machine (SVM) model for predicting the changes in sugar concentration in a liquid. The work was validated by implementing the developed LPM and CPM on a real-time system. System validation results showed that LPM detected the level with an error of −6 to 0.0001 cm, and CPM gave an accurate result for predicting the sugar concentration in liquid with an error between +/−0.8%. The obtained results of models catered to the objectives of the work with enhanced accuracy and avoided the re-calibration of CLS-based Liquid Level Measurement Systems (LLMS).

Keywords

• Calibration and fitting methods
• concentration measurement
• data analysis
• liquid level measurement
• multi-sensor data fusion