This report summarizes the technologies that are emerging as potential on-line sensing tools for particle size discrimination and volume fraction assessment for feedback control in the chemical and pharmaceutical industries. Specifically, the physics and the inversion algorithms associated with ensemble measurements of turbidimetry, dffiaction or angular scatter, dynamic light scattering, Fiber optical dynamic light scattering, diffising wave spectroscopy, diffirse reflectance and/or transmittance, photon migration, acoustic spectroscopy and electroacoustic spectroscopy are described with examples of their deployment in industrial settings as found in the literature, In addition, non-ensemble techniques such as particle imagery and back scatter and reflection are also described, Since ensemble techniques must include particle interaction effects when interrogating size information in suspensions with volume fractions greater than ten percent by solids (or dispersed phase), a section of the review is devoted to “static” and “dynamic” structure and how these interactions can seriously impact size distributions 11 not properly accounted for, Since vendor information for particle sizers and analyzers is found elsewhere in the literature, this review does not survey this aspect of the particle size characterization.
There are two difficulties associated with the integration of measurements for control. The Jirst dtifficulty resides in the fact that the first principles models for the process are often unknown or not known accurately enough, thereby limiting the eflectiveness for measurement input, X,,,, into the inverse of the process model to predict in closed or open loop fashion, the corrective action, AY, for maintaining the desired process output as close to the setpoint as possible. Secondly, the ensemble measurement model which relates the process output, X, to the measurement, X, may also be inaccurate or unknown restricting the usefulness of the input into the inverse process model. In the review of “ensemble ” techniques, we attempted to carefully point out the assumptions of the measurement model which cause the measurement, X,, to be less than a realistic predictor of process output X. As described above, there has been no demonstration for the inversion of$rst principles model to obtain size distribution in the presence of particle interactions. while optical techniques promise the opportunity of Jirst principle modeling with simple Percus-Yevick theory for hard-sphere interactions, actual measurements and inversions to obtain f(x) and # remain to be performed to establish practicality of the approach. In so far as these ensemble measurement models can accurately reflect changes in the internal state of the process, the model mismatch in non-spherical, concentrated and interacting systems may not be serious. On the otherhand, it is imperative that process disturbances that result in off- specification product are sensitively detected by these ensemble measurements.
Owing to the uncertainties of process and ensemble measurement models, the error associated with the inversion of the ensemble measurement model to obtain X,,, (i.e., f(x) and 4), and the magnification of the error in X,,, upon inversion of the process model to obtain AY, it may be advantageous to feed measurement data directly to an empirical or semi-empirical process model. Thus, it is important to be able to employ raw ensemble measurement signals directly as measured process outputs, even though their physical significance has not been defined by a measurement model.