Home > User Area > CES Selector Documentation > Methodology > Screening

The Screening Step

The screening step can be performed very effectively using a computerised database containing the properties of the entities under consideration e.g. material properties or the attributes of standard components. The database, however, must have certain important characteristics.

Figure 1 The Screening Step

Comprehensiveness:

By comprehensive, we mean that the database contains all general classes of entities in the "kingdom" of interest. A materials database should therefore contain all classes of materials in the "materials kingdom" (metals, polymers, ceramics, natural materials and composites) [1]. A database of actuators should contain all classes of actuators (hydraulic, electrodynamic, electric motor, piezoelectric...), and so on. If it fails to do this, one or more important classes of entities may be overlooked completely, thereby ruling them out of the selection by default.
It is often tempting for engineers to assume that the best solution will be the one implied by previous experience and therefore to use a non-comprehensive data source for example, a catalogue containing only rolling element bearings. However this can inhibit innovation by inviting a non-optimal selection due to lack of familiarity/experience with alternative solutions. In this example, a plain bearing or a hydrostatic or magnetic bearing may be better suited to the particular application. Thus a non-comprehensive database, with only a limited set of classes, is poorly suited to the screening step.

Universal attributes:

For effective screening, the database must be tightly structured. It is essential to store the same set of information for all "records" (entities). This constrains the choice of "fields" (properties) to those which are common to all entities in the database. These properties are the so-called "universal attributes". Decisions about the "taxonomy" of the database, ie the hierarchy of records, are strongly linked to the choice of attributes.
For example, consider constructing a database of linear force actuators. It would be possible to include all of the main families of actuators (hydraulic, pneumatic, piezoelectric, electrodynamic, magnetostrictive,...) as records in a comprehensive "generic" database. The universal attribute set would consist of properties like "maximum force", "stroke", "resolution", "power consumption", "bandwidth" and "size". These attributes are relevant to all actuators, whatever the family and they can be used to select families, classes or sub-classes from a "generic" database of actuators [2]. However, such a database should not contain class-specific attributes such as "maximum fluid flow rate" or "servo-valve current" which are only relevant to one family of actuators (hydraulic). If a "generic" actuators database contained these "family-specific" attributes, and a selection was based on them, some families (e.g. electrodynamic or piezoelectric) would fail the selection by default, because they would not posses data for those attributes.

Two-stage screening process:

This requirement of common attributes generally leads to a two-stage screening process, whereby, in the first stage, a comprehensive database is used to select the main families and classes of entities in the kingdom, based on the universal attribute set maximum force, stroke etc in the actuator example. When one or more suitable candidates has been found, a "family-specific" database can be used to narrow down the selection to one or more particular entities. The family-specific database can contain attributes that are only relevant to the particular family of entities (e.g. maximum flow rate or valve current in a hydraulic actuators database) and these can be used in the second stage of the screening process, without eliminating any candidates by default. Note however that even class-specific databases should only contain attributes that are common to all of the entities they contain. Otherwise, members may be rejected because of an absence of data.

Completeness:

All of the records in the screening database must be complete, ie have no "holes" or "gaps" with missing data. Otherwise, the inevitable consequence is that some entities will fail selections because they have no data for a particular field.
There are two main reasons why data may not be available for a particular property. The first is that the property may not strictly apply for the particular entity. The second is that it may never have been measured or the information may not be in the public domain. In both cases it is important that no "hole" is left in the database, and so it is necessary to approximate or estimate the property. There are some reasonably accurate ways of doing this for materials data, using relationships between properties [3]. For example, the Young"s Modulus E of an isotropic material is related to its Bulk modulus K and Poisson's ratio n by the equation: E = 3(1 - 2n)K. So if E and n are known, K can be predicted confidently. Similarly, if the atomic volume Vm and the melting point Tm are known for a material, then the Young's modulus E can be estimated using E = CRTm/Vm, where R is the general gas constant and C is a constant that depends on the class of materials. More details are provided in [3].

Note that it is essential that such estimates in the database are flagged in some way, so that if the material is selected on the basis of one of these properties, the engineer knows to seek further information to check the estimated data.

Data quantity and precision:

The screening stage will in general consider a relatively small proportion of the total information available about a particular "kingdom" of entities. The remaining information will not be in a suitable format for the screening process and will generally be better suited to the "supporting information" step. The data will often be of low to medium precision, compared with the high precision data available later in the process. For example the price of a material or its strength stored in a general materials database may often be less precise than information provided directly by the supplier during the "supporting information" step. However, the less precise information at the screening stage can be used very effectively for ranking purposes.

Selection process:

In engineering selection, the screening process should normally be quantitative, and should be performed by linking the technical and economic requirements of the design with the attribute profiles stored in the database(s). In the case of materials selection, the links can be based on "material performance indices" which are combinations of material properties that characterise performance in a given application [1, 4]. Typical examples are the specific stiffness of a material E/r, or the specific strength sf/r (E is the Young"s modulus, sf is the failure strength and r is the density). These particular indices can be used to select the optimum material for a light, stiff tie rod or a light, strong tie rod respectively. Many material performance indices have been derived and tabulated for standard design cases in mechanical, structural, thermo-mechanical and electro-mechanical engineering [1, 5, 6].
Similar performance indices can be used profitably in other engineering selection problems, for example, the selection of force actuators [2], electrical energy sources (batteries, generators...) [7] or standard structural components (beams, columns...) [8].

Example:

The Cambridge Engineering Selector(CES) was developed by Granta Design to perform the screening step, using the approach described above. It has a database containing all of the main families of materials (metals, polymers, ceramics, natural materials and composites), with each family expandable (through a filter) to show the individual materials making up those families. For example, the Ferrous Metals filter has entries for nearly 500 cast irons, carbon steels, low alloy and stainless steels. Other family-specific filters contain data for light alloys, copper alloys, conductors, polymers, ceramics, foams and so forth. Where necessary these filters contain class-specific attributes. For example, the Polymers filters contains, among others, the additional properties: water absorption and oxygen index (a measure of flammability) which are not relevant for most other classes of materials.

The way in which the screening stage is performed in CES is illustrated in Figure 2, taken from a typical case study . It shows a materials selection chart of endurance limit plotted against density, generated by CES using the Materials database and the Generic filter. Each material is displayed as a bubble, which indicates its range of properties. This chart could be used (for example) to select materials for flywheels. In [5], it is shown that the best materials for a flywheel capable of storing maximum rotational kinetic energy are the ones with the highest value of M1 = sf/r, where sf is the failure strength (endurance limit). These materials lie above the "M1" line, towards the top left corner of the chart. They include carbon and glass fibre reinforced plastics. The best materials for low speed flywheels are the ones with a high value of M2 = r, towards the right side of the chart. These are heavy materials like cast iron and lead. See [5] for further details.

Figure 2 A chart of endurance limit, se, against density, r, showing the selection of materials for a low speed flywheel (maximise M2), or a high speed flywheel (maximise M1)

(from [5]).

The result of the screening step is a short list of candidate entities which satisfy the requirements of the design. However, because the data available in the screening database(s) is limited, it will invariably be necessary to seek additional information about the few likely candidates before the final choice can be made. This is the "supporting information" step in figure 1 on that page.