Adaptive data-driven predictive control using behavioral approach for autonomous powder compaction

Dr. Vikas Kumar Mishra
Gottlieb-Daimler-Str. 42
67663, Kaiserslautern
Phone: +49 (0)631/205-3592
Fax: +49 (0)631/205-4201
vikas.mishra(at)mv.uni-kl.de
 

Prof. Dr. Naim Bajcinca
Dr. Vikas Kumar Mishra
Prof. Dr. Markus Thommes (TUDo)

Powder compaction performed on a rotary tablet press is a dry granulation method to transfer powder materials consisting of several components (drug, lubricant and other excipients) into compacts (tablets). This process is usually integrated in a multi-stage manufacturing line for the product design. The product quality indices refer to the dose, hardness, disintegration and dissolution of the compacts. These attributes are maintained during processing by adaptation of the subsequent unit operations in the rotary tablet press. Feeding, blending and filling are executed continuously and determine the chemical composition. Compression and ejection, on the other hand, are performed semi-continuously and determine the mechanical properties of the compacts (e.g. hardness). The interconnection of the five process steps leads to a complex control task. During start-up, the production rate is sequentially step-wise increased until the desired operating point is reached, which typically leads to a considerable product waste. During continuous production, a predefined operating point is maintained and the product quality is checked ex-situ periodically. Thereby, the product quality may be deteriorated due to the inherent process disturbances in form of fluctuations of the material feed rates or diverse material properties (e.g. particle size). All established strategies for this purpose, including the critical adjustment of production parameters for purposes of disturbance compensation are currently based on human operator interventions. However, human intervention is generally costly and prone to slow and inaccurate (re)actions. The present research programme finds here the motivation for developing an autonomous process control system, which is to replace the human intervention into the process course. On the basis of in-situ process monitoring, its goal consists in enhancing the process efficiency by minimizing start-up material waste and maximizing the production rate, while guaranteeing predefined quality marks of the compact production.
 

• In-situ process monitoring for real-time detection of quality deviations:
  Process monitoring aims for the detection of product deviations from a specifed target function (tablet dose and hardness). A set of sensors (UV-Vis spectroscopy, NIR spectroscopy, machine data) is developed to characterize the corresponding product properties (weight, weight fraction, porosity and lubrication) during processing.
• Data-driven modeling for control:
  Mathematical models (ARX, NARX) for the powder compaction process shall be developed based on the measured data.
• Adaptive predictive data-driven policies:
  We develop offline and online data-driven predictive control (DPC) policies utilizing the behavioral system theory. A theoretical base for the decision policies subject to stochastic and nonlinear data-based process models needs to be established in both scenarios, offline and online adaptation, which go beyond the state-of-the-art of data-driven control.
   

Powder compaction
Data-driven control
Fundamental Lemma
Autonomous Control
 

Deutsche Forschungsgemeinschaft (DFG)
 

Jan 2023 - Dec 2025