(233aa) Real-Time Monitoring and Ejection of out-of-Spec Tablets Via NIR Chemical Imaging

Near-infrared chemical imaging (NIR-CI) using high-speed cameras based on the push-broom acquisition principle is a strongly evolving technology in food, waste or mineral sorting. Yet, in the pharmaceutical industry, it is mainly used in lab settings.

Here we present a NIR-CI system, which is capable of real-time monitoring of approx. 500.000 tablets per hour and forwarding aggregated data (mean API content, standard deviation (SD) and number of analyzed tablets) to a process control system (PCS). Alternatively, the obtained chemical information can be used to identify and eject single out-of-spec tablets. The system consists of a push-broom spectrometer (EVK Helios G2 CLASS), a software for image analysis and pneumatic valves for ejection of tablets.

The camera (312 pixels spatial-, 248 pixels spectral resolution, 500 Hz measurement rate) generates 49 MB/s of spectral raw data. To reduce the data, chemometric models developed in Simca 13 are transferred to the spectrometer. They are executed in real-time on the onboard FPGA (Field Programmable Gate Array). The resulting video stream (0.29 MB/s) is further analyzed via openCV, an image processing library. For each detected tablet the mean and SD of API content, predicted via PLS, are calculated and averaged for one second. The results (3×4 byte/s) are written to an OPC-DA server, a standard interface to PCS, enabling a reaction to measured deviations. For ejection of out-of-specs tablets the video stream is analyzed via a DSP/TI System. Objects, which have to be ejected are forwarded to an IO subsystem, which controls the pneumatic valves.

Thereby the overall quality of tablet manufacturing can be raised by removal of potential out-of-spec tablets from the product stream at full manufacturing speed.