PARAMETRIC DESCRIPTION OF DYNAMIC BEHAVIOR FOR PLANT MATERIAL AND ITS APPLICATION TO QUALITY CONTROL
Nonphysical and parametric model was proposed for describing the dynamic behavior of the plant material excited by the random oscillating force. This was the autoregressive moving average(ARMA) model used in the time series analysis.
The ARMA parameters were estimated by the recursive least square(RLS) method formulated from the Kalman filtering theory.
The quality change of the plant material during storage was evaluated with these parameters. The Mahalanobis distance calculated from these ARMA parameters was used to evaluate the quality of the material quantitatively.
The frequency response function calculated from the ARMA(4,3) model which had less parameters than AR(10) model was closely resemble to the frequency response estimated from the FFT method.
The error rate of discriminant was high at the early stage of the storage, but at the final stage of the storage(40 days after) the error rate was almost zero. That is,it was able to distinguish the samples at the first and final stage of the storage.
The Mahalanobis distance calculated for the samples under the bad storage condition from the reference samples, that is, the samples at the first stage of storage, increased sooner than other samples. Then, the samples storaged under the good condition could be discriminated from those storaged under the bad condition.