2022 Mickey Leland Research Associate Mychal Amoafo and Mentor Larry Shadle working in the lab
2022 Mickey Leland Research Associate Mychal Amoafo working in the lab with mentor Larry Shadle. Mychal is working on the Advanced Systems Integration Team and his project is Intelligent Sensors for Control. In this research, typical process data is analyzed in real time and used to infer other process factors: information from the associated chemistry and physics that can then be used to improve sensitivity, control, process stability. For example, a flame temperature sensor for the Hyper combustor based upon calculating the adiabatic flame temperature. Once this is calibrated against the process response, then we can determine the real-time equivalence ratios, flows, and heat fluxes. In the thermal energy storage model it will give us the driving force for the energy input and then the wall temperature can be used to characterize the inventory of stored energy. Approaches including PCA (Dynamic or sampled, unsupervised) and neural networks (supervised) will be applied for dimensionality reduction and fault detection. We can force neural networks to do dimensionality reduction by designing encoder (or even an autoencoder) and decoder network. The encoder will reduce the dimensionality and we can analyze this reduced dataset for insight into fault detection. The student may use python to obtain initial results and reproduce in MATLAB if time is still available.
2022 Mickey Leland Research Associate Mychal Amoafo and Mentor Larry Shadle working in the lab
2022 Mickey Leland Research Associate Mychal Amoafo working in the lab with mentor Larry Shadle. Mychal is working on the Advanced Systems Integration Team and his project is Intelligent Sensors for Control. In this research, typical process data is analyzed in real time and used to infer other process factors: information from the associated chemistry and physics that can then be used to improve sensitivity, control, process stability. For example, a flame temperature sensor for the Hyper combustor based upon calculating the adiabatic flame temperature. Once this is calibrated against the process response, then we can determine the real-time equivalence ratios, flows, and heat fluxes. In the thermal energy storage model it will give us the driving force for the energy input and then the wall temperature can be used to characterize the inventory of stored energy. Approaches including PCA (Dynamic or sampled, unsupervised) and neural networks (supervised) will be applied for dimensionality reduction and fault detection. We can force neural networks to do dimensionality reduction by designing encoder (or even an autoencoder) and decoder network. The encoder will reduce the dimensionality and we can analyze this reduced dataset for insight into fault detection. The student may use python to obtain initial results and reproduce in MATLAB if time is still available.