Info

Variable No.

Figure 6.68. MPLS-based on-line scores and contribution plots of a faulty batch.

in a three-way array of size X(37 x 14 x 764). A subset of 37 batches is chosen from the original 41 batches. After unfolding by preserving the batch direction, size of the resulting matrix X becomes (37 x 10696)). AHPCA-based empirical modeling is performed on the unfolded array X with three principal components and the weighting factor d is chosen as 0.35 (for all the sampling intervals). Explained variability on X block by AHPCA model is summarized in Figure 6.71(b) and Table 6.13. The explained variability even with 3 PCs is higher than that of 4 PC MPCA model presented in Section 6.4.3 (Figure 6.45(e)).

Process monitoring stage: The adaptive model developed is used to monitor new batches on-line. The batch fault scenario with 10% step decrease in agitator power input between the 140th and 180th measurements (Figure 6.44 and Table 6.8) is used. New batch data are processed with

200 400 600

Measurements

Figure 6.69. MPLS-based on-line predictions of end-of-batch product quality. (•) represents the actual value of the end-of-batch product quality measurement.

200 400 600

Measurements

Figure 6.69. MPLS-based on-line predictions of end-of-batch product quality. (•) represents the actual value of the end-of-batch product quality measurement.

Figure 6.70. Adaptive hierarchical PCA scheme [496].

Figure 6.70. Adaptive hierarchical PCA scheme [496].

PC no.

X-block

Was this article helpful?

0 0

Post a comment