PREDICTING PULP QUALITY FROM PROCESS ACOUSTIC MEASUREMENTS ON A MEDIUM CONSISTENCY PULP STREAM

摘要

Sampling from a medium consistency, MC pulp stream is difficult whether performed manually or by an on-line device. Much would be gained if sampling could be avoided by using a non-invasive technique for measurement. Acoustic sensors are non-invasive and measure the total flow which means that no physical sampling is needed. Such a system should also be more robust and have less need for maintenance since the sensor will not be in contact with the process stream. In an attempt to probe the feasibility of such a system we have used vibration/acoustic measurements to get “fingerprints” of the pulp using Wavelet Transform Multi Resolution Spectra, WT-MRS as it moves through the pipe. Pulp samples were taken concurrently and analyzed in the laboratory. The samples were analyzed for standard parameters such as Canadian Standard Freeness, CSF (CSF is a measure of dewatering resistance for pulps) and mean fiber length. They were also used to form sheets subjected to physical testing. Models relating the acoustic fingerprints to pulp quality were developed using PLS. Good results were achieved for CSF and ISO Brightness using fully cross-validated models. If this approach were to be implemented in a mill, pulp quality values could be achieved at a maximum rate of one/sec.