Home > User Area > CES Selector Documentation > Data Checking

Data Checking and Estimation

Data errors can be of several types. There is the risk of the operator making an error in entering a datum incorrectly. Standard procedures (multiple entry, for instance) can catch this. There is the risk that the datum is entered correctly, but it is wrong to start with ( perhaps in the wrong units). This can often be detected by 'range checks', described below. Finally there is the inaccuracy in measuring the datum in the first place. When this is large it can be caught by 'physical data checks', also described below.

Figure 1 below illustrates a hierarchical data checking scheme (such as that used by CES), in schematic form. All the data in the database are subjected to the checking procedure automatically.

Figure 1 The procedure for 'intelligent' data checking

For each material class, every property must lie in a well defined range. For example, the Young's Modulus of a metal must lie in the range bracketed by the values 5GPa (Lithium) and 560GPa (Osmium). If it does not, the value is wrong. The ranges for ceramics differ from those for metals, and both are distinct from those for polymers, for composites, for woods and so on. The range-check tests that each datum lies within an appropriate range.

The range can be as specific or general as necessary. For example, a data entry for a Low-Carbon Steel can be checked against the ranges for all metals, all Ferrous metals, all Carbon Steels or all Low-Carbon Steels.

It is possible to do considerably better than this. Material properties are not independent quantities. Metals with high melting points have high moduli. Polymers with high moduli have high yield strengths. Materials with low densities have high specific heats. The thermal expansion coefficient correlates with the melting temperature, and so on. There are sound physical reasons for these relationships, and they often exert tight constraints on material property data which result in the 'patterns' or correlations that are apparent on many of the material selection charts produced by CES.

For Example, Young's Modulus E, the atomic or molecular volume Vm and the melting point Tm are related by

EVm = CRTm

A data check based on these relationships can detect even small inconsistencies in two or more properties simultaneously and they can be used to check material property data to a high precision. The scheme is illustrated in the 4th column of Figure 1 above. In it are listed some correlations and the property-combinations which are involved.

The checking relationships have another equally useful function. There remain gaps in our knowledge of some material properties. The fracture-toughness of many materials has not yet been measured, nor has the electrical breakdown potential, nor many others. The absence of a datum for a material would falsely eliminate it from a selection which used that property, even though that material may be a perfectly viable candidate. This difficulty is avoided by using the range and correlation rules to estimate a value for the missing datum. Note however that such estimated data must be flagged as such, so that they can recognised as such later on.