Abstract:
A fire in a milk powder plant, such as the incident which happened at the Te Rapa plant near Hamilton on Good Friday in 1993, could be both dangerous and costly. In order to continuously process millions of litres of milk into milk powder throughout the milking season, it is crucial that such incidences are minimised if not prevented. It is generally accepted that one of the main source of ignition is the milk powder deposits or lumps undergoing self-heating, with temperatures able to increase to in excess of 700°C. It is therefore important to be able to quantify the self-heating process in milk powders. To do so, the thermal ignition kinetics of the powders must be known. The first part of this study involved the measurement of the thermal ignition kinetics of a variety of milk powders. A new method of measurement, using the crossing-point temperature concept and based on the equation governing transient self-heating, was developed. This method was used in place of the traditional steady-state method. It was found to be accurate and results were obtained in much shorter time periods compared with the steady-state method. The materials tested were whole and skim milk powders, treated and untreated wood sawdusts, milk powders with various compositions (obtained by mixing different proportions of whole and skim milk powders), aged whole and skim milk powders (obtained by keeping the powders in the oven at constant temperatures for 24 hrs) and whey protein concentrate. Some whole and skim milk powder samples were tested in an enclosed container to determine the effect of air in the self-heating process. In addition, thermal gravimetric analyses on the fresh and aged whole and skim milk powders were carried out to examine the change in mass of the powders with temperature. A model was developed to simulate self-heating in reactive materials. The significant difference between the model developed in this study and the models of previous studies was the approach to the drying/wetting term used. This model was set up to solve simultaneously the heat and mass balances in the packed powder. Comparisons of the results were made with the experimental data recorded during the kinetic measurements. The model was able to predict qualitatively the self-heating of milk powders. More accurate thermal and physical properties would be required for more precise predictions. In the second part of the study, the characteristics of smouldering milk powders were examined. Milk powder samples were heated by hot air in a reactor. The sample
temperature, carbon monoxide and oxygen concentrations in the exhaust and smoke point (the sample temperature at which smoke was first detected) were monitored. The powders tested were whole and skim milk powders, 100°C aged whole milk powder and whole milk powder samples with increased moisture contents. In addition, the smoke generated during the combustion of whole milk powder was analysed using gas chromatography-mass spectrometry. The main volatiles formed were tentatively identified as hydrocarbons, alcohols and carboxylic acids. The results of the carbon monoxide measurements were used in the development of an early fire detection system for milk powder plants. The system was similar to the one developed at the Netherlands Institute for Dairy Research, and used a sensitive carbon monoxide analyser to measure the carbon monoxide concentration in the exhaust of drying devices. Trials were carried out at the Northland Dairy Kauri Site plant, and the system was able to detect the carbon monoxide sample injected into the plant. The cost of the system, excluding installation and testing, was estimated at about $32,000.