(732d) Image-Based Control of a Material Healing Process with Blender

Industry 4.0 heralds the use of big data not only in the amounts of it generated, but also the different forms that can be captured, stored, and processed. With advances in computing capabilities, using image-based sensors and control algorithms on-site become a viable option for plant-wide applications. Image-based control is popular in the field of robotics, such as the control of unmanned aerial vehicles [1], and in chemical engineering applications, as in [2], where images are used to classify flow regimes in aerated tanks. A process where images can be extremely useful to determine states is Directed Self-Assembly (DSA), where individual particles arrange themselves into desired structures under an external control action. Image-based control of this process has been explored in the work of [3], where colloidal particles suspended in a liquid are assembled with an image-based, closed-loop control system. There is a need to be able to simulate such applications in a virtual environment, in order to benchmark and test a proposed system before deployment to the physical plant.

In this work, we present the use of Blender, a free and open source 3D modeling and animation software, as a test bed for image-based control applications. We have explored the use of Blender in similar applications, such as level control in a tank, disturbance handling due to weather phenomena, and the control of a nanorod under Brownian motion [4, 5]. Here we simulate and control the motion of a particle in a fluid medium in order to repair a layer present on the inner wall of a cylindrical pipe. Virtual cameras in Blender are used to capture images of the particle as it travels downstream to the repair site, and control actions are executed that cause the particle to move to the site in order to repair the layer. This simulation functions as a precursor to a self-healing corrosion inhibitor layer. This demonstrates the use of Blender to reconstruct the 3D location of the particle from images captured by separate cameras, and execute the control action using a Stochastic Model Predictive Controller (SMPC). The SMPC takes into the account the motion of the particle due to flow in the pipe, as well as Brownian motion deviations (modeled as random variables) due to impacts from particles in the medium.

References:

[1] Rafik Mebarki and Vincenzo Lippiello. “Image-based control for aerial manipulation”. In: Asian Journal of Control 16.3 (2014), pp. 646–656.
[2] Corinna Kr ̈oger, Valentin Khaydarov, and Leon Urbas. “Data-driven, Image-based Flow Regime Classification for Stirred Aerated Tanks”. In: Computer Aided Chemical Engineering. Vol. 51. Elsevier, 2022, pp. 1363–1368.
[3] Xun Tang and Martha A Grover. “Control of microparticle assembly”. In: Annual Review of Control, Robotics, and Autonomous Systems 5 (2022), pp. 491–514.
[4] Henrique Oyama et al. “Development of directed randomization for discussing a minimal security architecture”. In: Digital Chemical Engineering 6 (2023), p. 100065.
[5] Akkarakaran Francis Leonard et al. “Virtual Test Beds for Image-Based Control Simulations Using Blender”. In: Processes 12.2 (2024), p. 279