Data encoding

Data encoding consists of converting historical failure records into formats that can then be used to calculate time-related and demand-related equipment failure rates. It is important that the encoding step be done so that plant information used to calculate failure rates can be traced through each step back to the raw data. The minimum set of information that must be recorded includes ... 

Figure 6.3 Example Data Encoding Form. From Science Applications International Corporation.

Due to such subtleties, the need to develop well-defined basic events, failure modes, and equipment boundaries prior to data encoding cannot be overemphasized. Familiarity with failure definitions and failure severities will be extremely helpful to the analyst. Figures 2.1 and 2.2, reprinted from IEEE Std. 500-19845, list a large number of failure modes by failure severity and may help encode failures. IPRDS also contains helpful information on failure encoding. Information on some equipment boundaries may be found in the Data Tables in Section 5.5. 

The primary value of proper and consistent data encoding is that it preserves the quality of input data. In addition, the use of accepted standard encoding schemes provides data that can be compared and combined with generic data. 

Data encoding The assignment of codes and identifiers to data extracted from plant records so that failure rates may be readily calculated. 

Von Neumann was able to construct a self-reproducing UTM embedded within a 29-state/5-cell neighborhood two-dimensional cellular automaton, composed of several tens of thousands of cells. It was, to say the least, an enormously complex machine . Its set of 29 states consist largely of various logical building blocks (AND and OR gates, for example), several types of transmission lines, data encoders and recorders, clocks, etc. Von Neumann was unfortunately unable to finish the proof that his machine was a UTM before his death, but the proof was later completed and published by Arthur Burks [vonN66]. 

Infrared spectroscopical data encode a lot of structural information and can be analyzed with the help of computational methods (vide supra) aiding in the identification of the observed species. Sometimes, two different electronic states may lie very close in energy and have similar geometries. In such cases (e.g., the quinonoid radicals to be described in Section II.B.), the predicted differences in the IR spectra are too small to allow an unambiguous assignment of the ground-state multiplicity. In this respect, ESR spectroscopy provides valuable comple-... 

The purpose of the following four chapters is to build a foundation of understanding of basic neural network principles. Subsequent chapters will address issues specific to choosing the neural network design (architecture) for particular applications and the preparation of data (data encoding) for use by neural networks. 

Figure 6.9 Converged total absorption spectra (N/16 = 64) achieved by Fade reconstruction. The time signals corresponding to benign (upper panel (i)) and cancerous (lower panel (ii)) ovarian cyst fluid were derived from data encoded in Ref. [23]. The reconstructed chemical shifts and concentrations of the 12 metabolites are shown above the corresponding peak.

Another complexity is that the compound associations are classified using a small number, for example the land 2 in the cid associations quoted above. These classifications might be called primary when the number is 1, secondary when 2, etc. Trying to apply a parsing rule for data encoded in a column is prone to error. There is no easy way to enforce the format of data in this column. 

The chemistry rules approach ended up being the basis for the original three-dimensional biomolecular structure file format, the PDB format from the Protein Data Bank at Brookhaven (Bernstein et al., 1977). These records, in general, lack complete bond information for biopolymers. The working assumption is that no residue dictionary is required for interpretation of data encoded by this approach, just a table of bond lengths and bond types for every conceivable pair of bonded atoms is required. 

We are often confronted with the problem of having to determine the correct sequence of words, phonemes or phones from a given waveform of speech. The general name given to this process is transcription. Transcription is involved in virtually every aspect of analysing real speech data, since we nearly always want to relate the data we have to the linguistic message that these data encode. 

NOTE 1 In the following discussion, computer manipulations of data encoded in this format are mentioned to facilitate the explanation. The programs to do so are available from one of the authors (MIK) but are not required to use this coding format.]... 

---