Classification techniques

Because membranes appHcable to diverse separation problems are often made by the same general techniques, classification by end use appHcation or preparation method is difficult. The first part of this section is, therefore, organized by membrane stmcture preparation methods are described for symmetrical membranes, asymmetric membranes, ceramic and metal membranes, and Hquid membranes. The production of hollow-fine fiber membranes and membrane modules is then covered. Symmetrical membranes have a uniform stmcture throughout such membranes can be either dense films or microporous. 

By selecting suitable descriptions for each aspect, an objective description of all feasible application methods (equipment and technique) is possible. This equipment and technique classification scheme outlined in Table 3.1 offers a taxonomy that would allow objective differentiation to be made between seemingly similar systems, based upon their measured performance judged against other comparable systems. 

For homogeneous NDT data and repeatable inspection conditions successful automated interpretation systems can relatively easily be developed. They usually use standard techniques from statistical classification or artificial intelligence. Design of successful automated interpretation systems for heterogeneous data coming form non-repeatable, small volume inspections with little a-priori information about the pieces or constructions to be inspected is far more difficult. This paper presents an approach which can be used to develop such systems. 

All the three techniques mentioned above may make use of fuzzy sets and fuzzy logic (for fuzzy classification, fuzzy rules or fuzzy matching) but this does not effect the discussion of the applicability to NDT problems in the next section. 

Hybrid systems. Depending on the problem to be solved, use can also be made of a combination of techniques leading to a hybrid system. For example, a rule-based system may use neural networks for solving classification subproblems (as is described in [Hopgood, 1993]), or a combination of a rule-based and a CBR system can be used as in the system for URS data interpretation described later in this paper. 

Neural network classifiers. The neural network or other statistical classifiers impose strong requirements on the data and the inspection, however, when these are fulfilled then good fully automatic classification systems can be developed within a short period of time. This is for example the case if the inspection is a part of a manufacturing process, where the inspected pieces and the possible defect mechanisms are well known and the whole NDT inspection is done in repeatable conditions. In such cases it is possible to collect (or manufacture) as set of defect pieces, which can be used to obtain a training set. There are some commercially available tools (like ICEPAK [Chan, et al., 1988]) which can construct classifiers without any a-priori information, based only on the training sets of data. One has, however, always to remember about the limitations of this technique, otherwise serious misclassifications may go unnoticed. 

The aim of this work which enter in a research project on NDT, is to conceive a system of aid for interpretation and taking decisions, on imperfections in metallic fusion welds, we have studied and tested several segmentation techniques based on the two approaches ( contour and regions ). A quantitative analysis will be applied to extract some relatives geometricals parameters. To the sight of these characteristics, a first classification will be possible. 

CLASSIFICATION METHODS In image processing, it often very interesting to built classes from the data structure. This technique can be partitioned into two categories ... 

For this purpose, a short overview will be given concerning some theoretical properties of the QCMD model (Sec. 2), This will allow for a suitable classification of the application problems. In the course of the following discussion, we will introduce two different classes of integration techniques ... 

Other methods consist of algorithms based on multivariate classification techniques or neural networks they are constructed for automatic recognition of structural properties from spectral data, or for simulation of spectra from structural properties [83]. Multivariate data analysis for spectrum interpretation is based on the characterization of spectra by a set of spectral features. A spectrum can be considered as a point in a multidimensional space with the coordinates defined by spectral features. Exploratory data analysis and cluster analysis are used to investigate the multidimensional space and to evaluate rules to distinguish structure classes. 

Spectral features and their corresponding molecular descriptors are then applied to mathematical techniques of multivariate data analysis, such as principal component analysis (PCA) for exploratory data analysis or multivariate classification for the development of spectral classifiers [84-87]. Principal component analysis results in a scatter plot that exhibits spectra-structure relationships by clustering similarities in spectral and/or structural features [88, 89]. 

Searching of one or more on-line databases is a technique increasingly used ia novelty studies. The use of such databases enables the searcher to combine indexing parameters, including national and international classifications natural language words ia the full text of patents, ia their claims, or ia abstracts suppHed by iaventor and by professional documentation services and indexing systems of various sorts. Because the various patent databases have strengths and weaknesses that complement each other, the use of multiple databases is thus pmdent, and is faciUtated by multifile and cross-file techniques provided by the various on-line hosts. 

Currendy, the Bauer-McNett classification and the QS test are the most widely used fiber classification techniques. Whereas there are quaUtative relationships between QS and BMN, there is no quantitative correspondence. It is readily understood that these standard tests do not provide accurate definition of the fiber lengths the classification also redects the hydrodynamic behavior (volumes) of the fibers, which, because of thek complex shapes, is not readily predictable. 

Other classification techniques have been developed which provide some insight on fiber lengths, typically the Ro-Tap test, the Suter-Webb Comb, and the Wash test. 

Glassification. Classification (2,12,26,28) or elutriation processes separate particles by the differences in how they settle in a Hquid or moving gas stream. Classification can be used to eliminate fine or coarse particles, or to produce a narrow particle size distribution powder. Classification by sedimentation iavolves particle settling in a Hquid for a predetermined time to achieve the desired particle size and size distribution or cut. Below - 10 fim, where interparticle forces can be significant, gravitational-induced separation becomes inefficient, and cyclone and centrifugation techniques must be used. Classification also separates particles by density and shape. Raw material separation by differential sedimentation is commonly used in mineral processiag. 

Supervised Learning. Supervised learning refers to a collection of techniques ia which a priori knowledge about the category membership of a set of samples is used to develop a classification rule. The purpose of the rule is usually to predict the category membership for new samples. Sometimes the objective is simply to test the classification hypothesis by evaluating the performance of the rule on the data set. 

In the discussion which follows, ciystaUization equipment has been classified according to the means of suspending the growing product. This technique redlices the number of major classifications and segregates those to which Eq. (18-31) apphes. 

The collection technique involves the removal of particles from the air stream. The two principal methods are filtration and impaction. Filtrahon consists of collecting particles on a filter surface by three processes—direct interception, inertial impaction, and diffusion (5). Filtration attempts to remove a very high percentage of the mass and number of particles by these three processes. Any size classification is done by a preclassifier, such as an impactor, before the particle stream reaches the surface of the filter. 

---