Projects

Projects: Projects for Investigator

In order to reduce the carbon in ash (CIA) levels arising from the application of advanced low NO_X technologies, it is necessary to improve the quality and consistency of the coal milling process. In many low NO_X retrofit applications, mill upgrades, including classifier upgrades, are required to achieve the improved milling performance. Unfortunately, plant space constraints often make it impossible to install classifiers of ideal geometries and the performance of non-ideal geometries is difficult to predict using existing design methods. In addition, low quality coals are increasingly being used, alone or in blends, to reduce plant operating costs. The grinding and classification behaviour of low quality coals and their blends has been found to differ from that of UK and world-traded bituminous coals. Consequently, classifier design rules that have been derived from the extensive experience of milling bituminous coals are less reliable when applied to lowquality coals. There is a clear requirement to improve and extend the range of applicability of classifier design methods so that they may be used to design classifiers of non-ideal geometries andfor coals outside the conventional range of experience.

A procedure for modelling classifier performance using the FLUENT CFD code has been developed. The method has been validated against physical model flow field and particleseparation behaviour measurements and plant performance data.

A 1/3 scale physical model of a typical E mill classifier design was designed, manufactured and tested to provide detailed flow field and particle separation behaviour data against which the CFD modelling procedures were developed and assessed.

A series of trials was carried out at a UK power station during February 2002 to provide performance data against which the CFD modelling procedures were further developed and validated.

The CFD simulation of the physical model included the selection of the kw turbulence model for use in subsequent modelling activities, based on accuracy of prediction and computational efficiency. Flow field prediction was shown to be largely grid independent, but it is recommended that classifier model meshes feature additional grid refinement around the vanes and vortex finder. The validity of using periodic slice models, to maximise computational efficiency while maintaining accuracy of prediction, was demonstrated. CFD modelling reproduced the physical model flow field with a reasonable degree of accuracy. Axial and radial velocities compared most favourably, with slightly poorer agreement in the prediction of tangential velocity and kinetic energy. The effects of vane angle and geometry changes on the flow field, pressure drop, particle collection efficiency and product particle size distribution were correctly modelled.

CFD models correctly predicted the measured trend of plant particle collection efficiencyand millp roduct fineness decreasing with increasing air flow rate.

The modelling procedure developed was applied to the investigation of the effects of modifications to classifier geometry on performance. The results dispelled several of the myths surrounding the effects of classifier geometric parameters on performance. They provide guidance regarding those parameters whose variation may be used to improve classifier performance.

The modelling procedure was also successfullyapplied tot he simulation of the primary classification that occurs in the mill body.