Application of mathematical models and genetic algorithm to simulate the response characteristics of an ion selective electrode array for system recalibration

With proper training procedures, including the use of a sufficiently large training sample set, ISE array with Back-Propagation Neural Networks (BPNNs) can be used to simultaneously determine the concentrations for multiple ions in environmental water. However, it is not practical for end-users to re-measure the whole training sample set - dozens of samples - to re-build the BPNN model every time after a long storage or measurement period, since both the sensitivity and selectivity of ISEs would be gradually degraded over time. In this study, instead of using BPNNs, we developed and patented some mathematical models to describe the based characteristics of the ISE array's response, including the sensitivity and selectivity. The developed mathematical equations can be used to simulate the response values of ISE array, and to determine the analytes' concentrations from unknown samples. By measuring only a couple of standard solutions, which are practical to the end users, the ISE array can be automatically recalibrated using genetic algorithm (GA). This research is based on an array of Prussian blue modified Carbon Paste Electrodes (PB-CPEs), and the interferences between nitrate (NO3) and chloride (Cl) was used as a case study. Refereed/Peer-reviewed