On April 1, 2000 the Forintek project titled "Online Measurement of Resin Distribution" was established. Up until that time Forintek had investigated the effects of resin distribution on OSB properties and had developed an off-line instrument (GluScan) for measuring resin distribution. Ideally, however, a more practical instrument for mills would be one which provided continuous, online measurement. Recently a company named Process Sensors Corp. has been producing a commercial NIR sensor specially modified for measuring resin level in OSB, much the same way NIR moisture sensors are used today in OSB mills to measure moisture content. For the purposes of this project, Forintek agreed to evaluate the Process Sensors unit to determine the suitability of NIR technology for measurement of resin in OSB furnish.
Tests were done mainly at the Alberta Research Council (ARC) and Forintek in Vancouver. Some additional blending was done at the Borden Chemical lab facility in Vancouver. Under ideal conditions, test results demonstrated that the NIR sensor data provides good trending of resin level in furnish blended with either PF liquid, PF powder or MDI. For each of these resin types, the NIR sensor gave readings to within approximately ± 2% resin levels. Because of this accuracy range, the sensor would appear more practical for PF liquid applications which generally operate with higher resin levels than MDI or powder PF.
Further testing, however, showed that the NIR is sensitive to changes in furnish conditions, other than resin, which can interfere with measurements of resin level. These factors include furnish species, moisture content and fines content. Changes to these conditions can cause the NIR sensor to provide erroneous data. To help compensate for these variable conditions, tests showed that NIR data can be corrected by subtracting measurement values for control furnish (0% resin). This may be possible to accomplish in a mill by using two NIR sensors; one to measure control furnish before the blender and one after the blender or on the forming line to measure resin level of furnish.
Additional results showed that the NIR sensor was unable to detect any significant differences in resin level variability between the good and poor blends. Also the NIR sensor showed that it is not capable of measuring PF or MDI when both resins are used in the same furnish.
It is recommended that further pilot plant testing be continued with mill trials to determine whether NIR technology is a practical way to measure resin level online, and whether it can be used in conjunction with the GluScan offline system to measure resin distribution as a full mill quality control program to optimize resin usage.