| 摘要 |
PURPOSE:To obtain the identification function of a state change pattern in time transition from measuring information by extracting a development coefficient at the time of developing pattern change to the combination of polynomials as the feature of pattern data, and modulating it to abstract expression. CONSTITUTION:At a feature extraction part 4, the state change pattern in the vicinity including the data of process measuring information at each time is divided into the appropriate combination of the polynomials decided in advance for the data. Therefore, the noise of a measuring value impossible to be expressed in the polynomial can be removed, and also, the feature of the data can be expressed by a coefficient in which the polynomials are combined. The change pattern in which the feature of time series data is expressed can be obtained from the feature extraction part 4 by processing the data that change in time series at each time sequentially. Then, an abstract expression part 5 converts the combination of appropriate feature change decided in advance to the abstract expression from the feature change pattern of obtained time series data. In such a manner, it is possible to obtain arbitrary abstract expression at every measuring data. |
