Monitoring and control of crystallization processes is based on information of crystal size distribution (CSD). CSD monitoring, however, has been based on sampling and off-line analysis by different methods. However, the growing demand for improving the process controls. Along the last 20 years the evolution in process instrumentation has resulted in the development of techniques for in-line measurement of variable such as particle size distribution (PSD), which is in most cases the most important information in the monitoring of industrial crystallization processes. Laboratory experiments were carried out for measuring the chord length distribution (CLD) of different particle systems using a laser reflection sensor. Samples consisted of adipic acid particles of different size, ranging from 0 to 400m m. The experimental results, consisting of the particle number counting per chord length class, were used in fitting a neural network model for estimating the number-based size distribution. The results indicate the feasibility of using such a model as a software sensor in crystallization processes monitoring. In a review on the methodology and problems related with the use of FBRM, the CLD data is the primary information provided by the equipment. In order to become useful as a process sensor, the CLD data must be transformed into a number or volume based PSD. In present work, the method is based on fitting a neural network (NN) software sensor, which is proposed for transforming an FBRM sensor into a numberbased on line PSD estimated measure. In the chapter 2, I will present some basic knowledge about measurement and crystallization process. The principle methods for focus on using improved NN models will be introduced in chapter 3.The laboratory experimental manipulation about crystallization forms the contents of chapter 4. All sampling data will play the important role in NN learning model, in chapter 5, I will say some idea about how to carry out my design work of NN model and the chapter 6 presents the result of NN simulation. At last, chapter 7, it's conclusion.