Glass-lined reactors are typically used in the pharmaceutical and specialty chemicals industries because of their material compatibility with most reactants and their cleanability. These reactors are typically equipped with a retreat-blade impeller placed close to the bottom of the tank a single baffle mounted from the top. The mixing performance of such reactors has not received significant attention in the literature, despite their ubiquitous presence in the pharmaceutical industry. In particular, mixing time, i.e., the time required by the system to achieve a predefined level of homogeneity, has not been studied to any significant degree.

In this work, the mixing time in a 60-L, scaled-down version of an actual industrial reactor having an industrial-type torispherical bottom and equipped with a retreat-blade impeller was experimentally determined for three different baffling configurations, i.e., unbaffled tank, partially baffled tank (in which the typical beavertail baffle was used), and fully baffled tank (i.e., with four full baffles). A conductivity method using sodium chloride (NaCl) as a tracer was used as one the main method to determine mixing time. Experiments in the unbaffled system were conducted by installing one two probes in the mixing vessel (at the wall and midway between the wall and the shaft, respectively). The presence of the conductivity probe(s) had a significant impact on mixing time. Results showed that different mixing times were obtained with the conductivity method depending on the location of the probe(s) and the number of probes. A separate colorimetric method coupled with image processing was additionally used to determine the mixing time. Both methods produced similar results when two probes were present, probably because of the baffling effects introduced by the probes themselves.

Results were also obtained for the partial baffled system and the fully baffled system, but only using the conductivity method. Experiments were performed when the agitation speed was varied in order to establish a correlation between mixing time and rotational speed for all systems. The mixing time in the partially baffled system was found to decrease with increasing impeller speeds and the predicted mixing time from regression agreed well with the experimental results. As for the fully baffled system, the mixing time varied in inverse proportion to the agitation speed, as in the typical baffled systems. The non-dimensional mixing number, θ₉₅N, was obtained within the speed range of 100-200 rpm for both the partially baffled and the baffled system. In the latter case, θ₉₅N was found to be independent of impeller speed, as expected.