James J. Filliben
Statistical Engineering Division, CAML
Polymers Division, MSEL
One of the most common methods used by the plastics industry today is injection molding (a special case of Liquid Composite Molding (LCM)-an industrially versatile and economically attractive technique for fabricating polymer composites of various sizes, shapes, and performance characteristics. The principle advantage of LCM is its ability to make large parts with complex shapes at a low cost. The LCM process consists of 4-steps: 1) it starts with a fiber preform being inserted into a pre-existing mold; 2) the mold is then closed, heated, and filled with a reactive prepolymer; 3) time is then allowed for polymerization; 4) when the composite part has gained sufficient strength (curing), it is demolded.
The entire process is temperature driven- it is critical characteristic of the polymer which drives both fill rate and curing rate. particular, the temperature at the core of the polymer. If the manufacturer could accurately monitor the internal temperature of the polymer at any given point in time, then this valuable piece of information may be judiciously used as feedback to make on-line adjustments to control both fill rate and curing rate. In-situ monitoring for on-line quality control of polymer composite processing is thus becoming widely recognized as an effective means of driving down cost by minimizing part-to-part variation. A number of in-situ sensing methods (ultrasonic, dialectric, optical, and spectroscopic) have been tried-unsuccessfully. More recently, Dara Woerdeman (and Richard S. Parnas) are carrying out work on a promising recent method which involves inserting a fiber optic sensor into the LCM preform and subjecting it to evanescent wave fluorescence. Woerdeman then examines the degree to which fluorescence wave-shift ("Stokes shift") serves as a predictor of monomer conversion (degree of resin curing).
Last year, the Parnas/Woerdeman paper given on their method at the Society of Plastics Engineers ANTEC '95 meeting won the "best technical paper" award, and the talk/paper is the "first reported development of an inexpensive sensor capable of making measurements throughout a complex part, rather than at a single point."
In regard to a model, a weighted quadratic spline
fit with a single knot turns out to be excellent
(with a 3 percent relative residual standard deviation),
with other system factors having no effect.
Figure 4: Injection Mold Modeling. Assessing the quality of quadratic spline fitting relating stokes shift as a function of time.
Date created: 7/20/2001