Here I introduce some of the projects I have been worked on or currently focus in. A overview of the projects interest me, the pain points, solutions being investigated.
#Physical Simulation #Software Collaboration
The initiative of cooperation is to find a flexible solution to tackle the new non-standard problems. In particular with the partnership between UVR and Hembach Photonik, it could bring optic geometric ray tracing(forward) method together with computer graphic path tracing(backward) method. Therefore to respond to growing customer demands, and offer engineers, designers and stylists access to the combined tool for developing a broad range of innovative products.
The technical feasibility study was scheduled as a two-week workshop. Furthermore, a feasibility report will be shared and a few projects in automotive industryhave been proposed to continue the study. Eventually verify the simulation with laboratory tests.
This is an exciting and challenging consulting project; what really appeals to me is the idea of simulation by script rather than GUI; most established industrial simulation software is powerful and efficient, they are good for solving standard problems; however, when it comes to nonconventional problems, adapting the simulation program using a GUI has itslimination.
The investigation was conducted from several perspectives. It takes into account of the companies: their projects, services, clients in the context of industry and market sector and their technical intelligence specialty, competence and capabilities. Along with the high-level observations, the investigation covers a deeper dive into the concept and format of the ray data, which is the keyto resolve the interoperation. Additionally, two project ideas were proposed to kickoff the colaborative cooperation. The identification of risks and improvement of programs were also underlined.
#Physical Simulation; #Automotive
The key question for physical simulation being asked most oftenly is how does it correlate with measurements. I have used several different optic simulation programs in my career as optical design engineer for automotive and general lighting companies, frankly speaking, each program does great job, the calculation was rather rapid and the result is correlate with laboratory about five to ten percent. Of course it strongly depends on how close your simulation geometry, material properties and settings to the reality. Until here, the conclusion above is more on the value of the simulated irradiance (illuminance), and radiance intensity (luminous intensity). What is missing, the raidance(luminance), visual aspect rendering is always a tricky topic, calculation takes long time, and simulation value may differ hugely to the measurement.
To solve this issue, the well-known approach is the "backward simulation", compare with the previously mentioned conventional "optical ray tracing", where the rays start from source to the sensor. In the "backward simulation"(could be called ray tracing, path tracing, etc depends on the techniques), rays starts from the camera, so much less "useless" ray was excluded to the calculation. It's much faster, and depends on the technique material data, the rendering can be true-to-earth.
In comparison to the speed and visual aspect was greatly improved, how about the radiance(lumiance) simulation data and measurement correlation? This was one of the main topic I was focused on with UVR company. It is a start-up with several developers, all of them has Ph. D. degree in physics, informatic, mathematic. Their product Predict Suite is powerful simulation package which offers real time true-to-earth rendering scene, with multiple layer of physical information: colors, spectra, polarisation states, etc. Apparently, it isn't simply a photo rendering engine, but has a great potential by using the physical components being calculated for engineering and development work.
Now come the question: what to do with the vast amount of multi-dimensional valuable physical data, how to filter the useful data chanel and present them to engineers intuitively. This was another main topic I worked in. The specification proposal on functions and features in the analysis module in Predict Suite.
#Medical
This is the first project I have taken as an independent optical consultant. The project was in collaboration with consultant from INSERM. The request came from a pharmaceutical research company. The most recent viscosity test results of their product appeared to be off a reasonable consistency. There are many types of viscosity test methods, the most common and well accepted one in pharmaceutic is a rotational test, wherein a volumn of fluid is sheared at a controlled rate and temperature.
The pain point is that the procedure differs with laboratory, test machine settings and other factors. Moreover, this specific product is a sorts of gel which will be applyed on the cornea and necessarily stay for several minutes before the next operation. So seemingly a rotational test is a dynamic test which isn't suitable for a stationary gel medicalement. Furthermore, the gel is a non-newtonian fluid, which means the fluid does not follow Newton's law of viscosity, its viscosity varies dependent on stress.
It was a challenge for me in a few perspect, the hardest one was to understand the problem, quickly learn the chemi-physical perperty of fluid and the standard measurement approach, etc. The optical method I have proposed and tested involve some classical setup: with camera, back light, standard convex lens, image processing and analysis. It is a non-standard approach for stationary viscosity test.
The proof of concept was carried during several months. I have been away for this project a while now, but there are still some left work to do. For example to apply the same test method in clinical with rabbits and check whether the result coorelates with the laboratory test.
It was a great opportunity to test my multi-discipline knowledge with optic, sensor, lighting and image processing. Most important, it opens a door for me to medical domaine.
Copyright © Tous droits réservés
