Departmental Research Areas

Bioprocess Engineering

Processing biological materials is important in the pharmaceutical, food, and biotechnology industries. Many pharmaceutical products are derivatives of plant or animal raw materials. Nearly all foods require some amount of post harvest processing to preserve the quality of the food, to make it palatable, and to keep it from spoiling before consumption. And microbial fermentations are increasingly being used by the growing biotech industry to produce, for example, enzymes, amino acids, hormones, vaccines, antibiotics, insecticides, fermented foods, and ethanol.

Bioprocess engineers design processes, equipment, and sometimes entire facilities. Knowledge of microbiology, biochemistry, transport phenomena, properties of biological materials, and engineering design is needed to be proficient as an engineer working in bioprocessing. Some expertise in sensors and process control is also valuable.

One of the greatest research challenges for the field of bioprocessing is that of developing sensors to monitor the product and process. Sensors are needed, for example, to determine the feeding requirements and the optimal harvest time for fermentations. Conventionally, parameters like temperature, pH, dissolved oxygen concentration, glucose concentration, and optical density of the fermentation broth have been monitored to indirectly monitor the state of the process. Recently, direct observation of the microbes using computer vision microscopy and image processing has been shown to be a fast and accurate way to determine the state and needs of the microbes. Since there is considerable natural diversity in biological materials and processes, it is important for bioprocess engineering researchers to develop more on-line sensors which can measure the properties of raw materials, intermediate products, and final products to economically control bioprocesses.

A machine vision system was also developed to detect fissures in the endosperm of corn kernels. These fissures, known as stress cracks, were determined by edge detection, preserved by an outline edge elimination operation, and represented by a feature vector. A classifier was built to distinguish stress-cracked kernels from non-stress cracked kernels by a predefined discrimination function which was learned by Learning Vector Quantization (LVQ) machine learning algorithm from training samples. The system achieved accuracies ranging from 83 to 98% with speeds of about two seconds per kernel.