| Summary: | The use of algorithms and automation to do chemistry and formulation have the potential to speed up discovery. In this thesis, a series of new approaches are explored. First, a genetic algorithm was employed as an evolutionary algorithm, with its mono-objective and multi-objective variant to optimise viscous formulations. An autonomous platform was designed to mix and stir multiple ingredients that are liquid and solid at room temperature. Viscosity, pH and opacity of mixtures are analysed and then used as input for the evolutionary algorithm, which designs the consecutive experiments according to the fitness gathered for each formulation. This procedure is repeated in a closed loop until the overall fitness of the optimisation reaches an optimum result and finds a compromise with respect to the targeted physical properties. Secondly, a multifunctional automated platform was built to perform a set of reactions involving the formation of molybdenum blue nanoclusters. The crystallisation and further analysis were automated. In addition, the rig was used to test the ability to keep a solution of polyoxometalates at a stable redox potential during the addition of various oxidants and reductants. The fuzzy logic controller was employed to stat the redox potential of the solution under equilibrium and non-equilibrium regimes. Finally, the platform was used to apply modified redox oscillations on the polyoxometalates with the aim of forming and breaking novel building blocks that are hard to isolate in a traditional batch reaction. Thirdly, the same fuzzy logic approach was used, which is far from equilibrium a system of organic macrocycles of different sizes. The hypothesis for this approach is to be able to track the emergence of novel structures like the stacking of the macrocycles and the formation of nanowires of different lengths. Experimentally, an oxidant and a reducing agent were added to dithiol solution enabling the formation and the destruction of organic macrocycles with different sizes. The reaction was followed using an online UV-Vis dip probe and online HPLC. The dynamics of this system under non-equilibrium conditions enabling the detection of emergent species (macrocycles with different sizes, beta sheets from the stacking of the macrocycles to form nanowires with different lengths) that are undetectable under an equilibrium. Finally, three nature-inspired metaheuristics were implemented in LabVIEW: The Fireflies-levy flight's algorithm, the Cuckoo search and the Tabu search were tested with multimodal multidimensional functions. The validation of the algorithms was performed on mainly two nonlinear chemical systems. These metaheuristics showed very good performance to detect the optimum solutions at high and low resolutions. The Brute Force was used as the control algorithm to ensure the detection of the best solution. The highest frequency of the Belousov-Zhabotinsky reaction was tracked through these optimisations.
|
|---|
