FIELDED MACHINE LEARNING SYSTEM FOR VOCATIONAL COUNSELLING

A fielded machine learning system for vocational counselling is presented in which learning is based on adjustments of weights on links in a network. The system exemplifies how a specific representation, consisting of weights assigned to associations between keywords, enables the effective use of machine learning algorithms for acquiring and continually refining domain knowledge. The representation is designed for coping with the types of knowledge that can be found in weak theory domains, that is, knowledge too difficult to formalize because it is incomplete or vague. Knowledge acquisition and knowledge refinement processes are automated in order to efficiently decrease the effect of unreliable knowledge caused by unknown or unspecified biases inherent in the knowledge sources. The knowledge acquisition algorithms presented here are capable of coping with incomplete and vaguely defined domain knowledge. The knowledge refinement algorithms are used on-line to enable a continuous refinement of ill-defined domain knowledge.     

APPARENT CONCENTRATION: A METHOD to PREDICT the FLOW PROPERTIES of VISCOUS FOODS FOR PROCESS APPLICATIONS

Compared with traditional methods, the prediction of flow properties of starch pastes is greatly improved with the new apparent concentration method proposed in this study. the new method considers the effects on the flow properties of minor variations in the preparation procedure and the composition of the product. the deviations with the new method range from ± 10% for high-viscosity pastes down to -30% for low-viscosity pastes. the deviations are mainly due to the time dependency of the shear stress, as these effects are considerable for starch pastes, but have not been considered in the modelling in this work. Successful use of the apparent concentration method requires a reference model with known parameter values, a simple rheological measurement on the actual product, and appropriate calculations. the simple measurement can be performed on-line or in the laboratory. A special residual analysis was developed to make it possible to split up the differences between observed and predicted values and attribute the parts to the following factors: measuring instrument, changes in the sample during the measuring process, model, parameter values, and the concentration used in the prediction. This residual analysis made it possible to identify the weakest link in the predictions.     

SCRAPED SURFACE HEAT EXCHANGERS

In this literature survey flow patterns, mixing effects, heat transfer and power required for rotation in scraped surface heat exchangers (SSHE) are thoroughly discussed, with the emphasis on assumptions and results, while the principal design of different SSHEs are only briefly discussed. The flow patterns control the desired radial mixing and the undesired axial mixing. the flow in a SSHE can be regarded as the sum of an axial flow and a rotational flow. the axial flow is laminar and the rotational flow is laminar or vortical. With laminar flow the radial mixing is poor, which causes poor heat transfer and allows the axial flow profile to control the residence time distribution. the precise onset of vortical flow in a SSHE is hard to predict. the vortical flow makes the radial mixing very efficient, giving good heat transfer and perhaps plug flow behavior. However, vortical flow also causes axial mixing which reduces the apparent heat transfer coefficient and increases the residence time distribution. The power required to rotate the shaft and blades is mainly determined by the design of the blades.     

MODELLING of VORTICAL HEAT TRANSFER IN SCRAPED SURFACE HEAT EXCHANGERS

Heat transfer experiments have been performed in pilot plant scraped surface heat exchangers, using starch pastes and water as model products. When the flow was vortical, the backmixing effects were considered using the plug flow and axial dispersion model; the true surface heat transfer coefficients and axial dispersion coefficients were determined from temperature measurements. These coefficients were modelled and the following variables were considered: rotational speed, viscosity, flow rate, number of blades, radius ratio and heat transfer direction. Using these models, the overall heat transfer coefficients and the axial temperature profiles can be predicted quite precisely for SSHEs over a wide range of operating conditions. the first blade improved the heat transfer coefficients while additional blades did not; and heat transfer was higher when the radius ratio was 0.5 than when it was 0.75. the differences between previously proposed models are explained. The most favorable flow pattern in SSHEs occurred when the flow was vortical, but close to the transition point to laminar flow.     

METHODS to DISTINGUISH BETWEEN LAMINAR and VORTICAL FLOW IN SCRAPED SURFACE HEAT EXCHANGERS

The transition between laminar and vortical flow has been investigated in scraped surface heat exchangers (SSHEs), using new methods based on viscosity and temperature measurements. The transition occurred close to the critical Reynolds number for annular flow of Newtonian fluids. Thus, the methods used to predict the viscosity can be successfully applied to SSHEs despite the very complex flow behaviour of the products, the complex geometry of the SSHEs, and the heat transfer in the SSHEs. In industrial applications, the transition can be detected using the new method based on temperature measurements at the outlet of the SSHEs together with some calculations. This is a very important result, since with this “sensor” the flow pattern can be controlled towards the optimal flow pattern in SSHEs, which in turn is a fundamental requirement when the taste of aseptic products containing particulates are to be improved by the introduction of continuous sterilization.     

Automated correction of linear deformation due to sectioning in serial micrographs

This paper describes an objective and automatic method for detection and correction of sectioning deformations in digitized micrographs, as well as an evaluation of the method applied to light and electron microscopic images of semi-thin and ultra-thin serial sections from brain cortex. The detection is based on matching of image subregions and the deformation model is bi-linear, i.e. two first-order polynomials are used for modelling compression/expansion in perpendicular directions. The procedure is applicable to prealigned serial two-dimensional sections and is primarily aimed at three-dimensional reconstruction of tissue samples consisting of a large number of cells with random distribution and morphology.     

CO2 Park

On lucky occasions, one gets the Big Idea—the concept that will guide and critique the entire design process. It is usually to that concept that the designer returns to test out decisions. But for us the designing of CO₂ Park worked a bit differently: We started with an inkling of an idea—a fascination with the natural phenomenon, Aurora Borealis—and worked our way through the project as if we were following a thin and barely discernible thread. One idea engendered the next. Often we had barely finished what we had set our minds to before we were moving off in another direction. When we seemed to be at a dead end, we did not look to the concept for answers; instead, we turned back to the last successful model, sketch, or thought to find our direction.     

MODELLING of LAMINAR HEAT TRANSFER IN SCRAPED SURFACE HEAT EXCHANGERS

Heat transfer experiments with starch pastes have been performed in pilot plant scraped surface heat exchangers (SSHE). Effects of the following variables were considered in the modelling of the laminar heat transfer: rotational speed; viscosity, flow rate, number of blades, radius ratio and heat transfer direction. The poor radial mixing restricted the heat transfer coefficients. When the radius ratio increased, the heat transfer rate increased. This was probably due to increasing mixing effects from the blades. the flow rate and the heat transfer directions did not affect the heat transfer coefficients. The scatter in the heat transfer coefficients was considerable, due to unstable flow conditions in the SSHEs. However, the scatter was reduced, compared with previous investigations, by the use of a static mixer after the SSHEs, since the radial temperature differences were eliminated. Better design of rotor and blades may improve the performance of SSHEs operating with laminar flow.     

Thermodynamic Model for Nonstoichiometric Ionic Phases — Application to CeO2–x

A two-sublattice model for reciprocal solid solutions is applied to an ionic phase with vacancies on the anion sublattice, assuming that there are some cations with a lower valency than normal. The model is applied to CeO_2–x, where x can take values up to 0.3. Experimental data on P o₂, as a function of temperature and composition can be well represented by this model, if interactions are described by two parameters, corresponding to a regular and a subregular solution parameter.