| This thesis presents the application of artificial neural networks (ANN) to the self-referent calibration of thickness-shear mode (TSM) acoustic wave chemical sensors. Spectrum analysis of impedance measurements affords complete characterisation of the TSM sensor, which includes the use of the Butterworth-Van Dyke (BVD) equivalent circuit to quantify the electrical responses. The multi-parametric nature of this method and a novel weight-adjustment procedure, applied to ANN calculations, were utilised to effect a method of calibration in the presence of interferents. A network is trained, using exemplary I/O data acquired for a potassium chloride (KCI) system, to predict unknown outputs i.e. concentration, given four sets of measured inputs i.e. series resonant frequency (FS), parallel resonant frequency (F P), motional capacitance (Cm) and motional resistance (R m). The trained and tested networks achieved high predictive efficiency with errors in the range of 2%--7%. An interferent, ethanethiol, is added to test the robustness of the trained network and was found to adversely affect the predictive ability of the network. The magnitudes of the weights, which are associated with the set of inputs deemed to be most affected by the interferent (FS), are adjusted to minimise this deterioration. The resultant network, calibrated for the interferent, achieved a similar predictive efficiency for adulterated samples as that achieved by the original network for unadulterated samples. This calibration procedure was extended to a dual-interferent system in which two input variables are affected. A significant decline in predictive ability was observed for the unadjusted network, render it practically unusable. A two-point weight adjustment process was performed on the affected weights with considerable success but the predictive ability did not come close to the original level. The manipulation of two out of four input parameters may have over-extended the resources of the network but may prove to be more amenable for a larger network.;A parallel project was effected in which a chemically-selective platform was immobilised on a TSM device following activation with bifunctional thiols. The optimal surface activation scheme was found to be the self assembly of a mixture of bifunctional thiols of different chain lengths. An amino-terminated 25-mer DNA single strand was successfully coupled to the distal carboxylic group via carbodiimide chemistry. The viability of the immobilised DNA receptor was demonstrated through a hybridisation experiment with a complementary DNA fragment. In addition, the multi-parametric nature of network analysis proved to be sensitive to the hybridisation kinetic and a subsequent, capacitive-based conformational change in the coupled strands. |
