On-line crystallinity measurement using laser Raman spectrometer and neural network Celal Batur a, *, Mohamad Hanif Vhora a , Miko Cakmak b , Toprak Serhatkulu b a Department of Mechanical Engineering, University of Akron, Akron, OH 44325-3903, USA b Department of Polymer Engineering, University of Akron, Akron, OH 44325-0301, USA Abstract A neural network is con®gured and trained to measure the polymer crytallinity in real time and non-intrusive man- ner. After the training, input to the neural network becomes the laser Raman spectrum at selected ferquencies and the output from the network is the current crystallinity of polymer. In order to train the neural network, a training data set is constructed where the crystallinities corresponding to a given set of Raman spectra are pre-determined by the small angle light scattering (SALS) methodology. The technique is applied to measure the crystallinity of low-density thin polyethylene (LDPE) ®lm. A typical sampling period for the determination of the crystallinity is around 12 s. The technique is compared to the principal component analysis that uses the same input data for calibration. # 1999 Elsevier Science Ltd. All rights reserved. Keywords: Polymer crystallinity; Raman signal; Neural network 1. Introduction Real time, non-intrusive measurement of poly- mer crystallinity is an essential step towards an implementation of a structural control system that can regulate the structure development during polymer processing [1]. Crystallinity development follows a complex function of temperature history, which can be described by the generalized Avrami equation   1  e   t 0 KT  dr 1 where  is the crystallinity and the nonlinear function KT represents the temperature depen- dency. If cooling follows a ®xed gradient i.e. dT=d  C, the crystallinity can be written as a function of temperature and cooling rate as   1  e  1 dT=d  T 0 KTdT 2 A qualitative on-line observation of crystallinity via measurement of the birefringence is demon- strated by Bansal and Shambaugh [2]. In this study, we present an on-line and non-intrusive, ISA TRANSACTIONS 1 ISA Transactions 38 (1999) 139±148 0019-0578/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved. PII: S0019-0578(99)00012-9 * Corresponding author. Tel: +1-330-972-7367; fax: +1- 330-972-6027. E-mail addresses: batur@uakron.edu (Celal Batur); cakmak @uakron.edu (Miko Cakmak); toprak@uakron.edu (Toprak Serhatkulu)