| Author | Suwan Homhual |
| Call Number | AIT Diss. no.PH-00-1 |
| Subject(s) | Milk--Microbiology
|
| Note | A dissertation submitted in partial fulfillment of the requirements for the degree of Doctoral of Engineering. |
| Publisher | Asian Institute of Technology |
| Abstract | The hygienic quality of raw milk has a direct influence on the shelf-life and other attributes of the processed dairy products. Therefore, the techniques for determining the total initial becterial population in raw milk play an important role in monitoring its quality at the milk collecting centers and dairy plants before being processed. Presently available methods for assessing the microbiological quality of milk are either time consuming or require sophisticated and costly instrumentation. Therefore, this research was carried out to develop simple and inexpensive methods for rapid indirect qualitative assessment of the microbiological quality of raw milk. Development of such tecvhniques will be vital for maintining the quality of raw milk and improving the efficiency of the manufacturing process of dairy products in Thailand. The first phase of this study involved the modification of a conventional methylene blue dye reduction test (MBT) in which metabolic activity of viable bacteria leads to the depletion of dissolved oxygen (DO) and resulting change in color of raw milk. The normalized output voltage signals from a DO probe and a especially developed light sensing probe representing the color change in raw milk samples were used for developing the models to assess the microbiological quality in terms of standard plate count (SPC) and methylene blue reduction time (MBRT). Signal voltages at 5 minute uniform time intervals were selected as independent variables for the test duration of 40 minutes for DO monitoring and 50 minutes for light sensing probe. Comparison of experimental values of SPC and MBRT with estimated values from statistical models resulted in low R2 value ranging from 0.314 to 0.565 with the slope of regression line close to 1. However, the accuracy of estimation improved significantly when artificial neural networks (ANNs) were applied for the interprelation of signal patterns obtained from the probes. In case of the DO depletion, a two-stage ANNs in which Kohonen network was first applied for categorizing voltage patterns in Group I and II and then followed by a bavckpropagation network or general regression neural networks (GRNN) yielded the best results among all models. In case of the light sensing probe, single-stage ANNs (backpropagation or GRNN) indicated best performance. Theses results confirmed that the application of ANNs could lead to reliable and more accurate estimation of SPC (log CFU/ml) and MBRT in comparison with statistical models for qualitative grading of milk within less than 1 hour. The second phase of this research was concernted with the development of a method for rapid assessment of the quality of bulk raw milk from the routine measurements before being processed at the dairy plant. The routine measurements of physical and chemical properties of bulk raw milk included fat, solids-not-fat, specific gravity, acidity, transport time, temperature of milk in tanker, and MBRT. Results of statistical analysis for 240 samples of raw milk indicated a very poor correlation (R2 = 0.215) between MBRT and other routine measurements. Also the application of a backpropagation ANN showed only a slight improvement in the estimation of MBRT. However, a GRNN model showed considerable improvement over the statistical model for estimating MBRT with R2 value of about 0.763 between the predicted and experimental NBRT values. A two-stage procedure further improved the estimation of MBRT with R2 colse to 0.82. The GRNN models followed Kohonen self organizing network which initially classified the data into two groups. Thus qualitative assessment of the bulk raw milk quality was feasible from the routine measurements based on the application of ANNs. Results showed that both techniques offered potential for rapid simple or inexpensive assessment of the quality of bulk raw milk. Finally, a scheme was developed for on-line implementation of the developed ANN models in real time. It was possible to input the signal voltage from either a DO or light sensing probe through an A/D interface and display the SPC (CFU/ml) or MBRT predictions based on ANN models. |
| Year | 2000 |
| Type | Dissertation |
| School | School of Environment, Resources, and Development (SERD) |
| Department | Department of Food, Agriculture and Natural Resources (Former title: Department of Food Agriculture, and BioResources (DFAB)) |
| Academic Program/FoS | Postharvest and Food Process Engineering (PH) |
| Chairperson(s) | Jindal, Vinod Kumar; |
| Examination Committee(s) | Athapol Noomhorm;Nagarur, N. N.;anee Tantirungkij; |
| Scholarship Donor(s) | Ministry of University Affairs (MUA) Thailand; |
| Degree | Thesis (Ph.D.) - Asian Institute of Technology |