Photoelectric sensor setup and characterization for general interface detection in laboratory scale liquid-liquid extraction processes

Liquid-liquid extraction is a well-known and widely used process in chemistry and laboratory settings.

The manual version of this process relies on visual inspection of the mixture, looking for interface formation and stabilization, and for draining the liquids separately. Different automatic devices have been developed for the purpose of separating immiscible liquids, however these solutions are often designed with a specific chemical process in mind. An automatically adjustable LLE device, that can detect and adapt to different substances can shorten adjustment time and enable automatic experimentation in laboratory settings. These kinds of devices are a key part of AI orchestrated, automatic experimentation and screening systems.

To enable adaptation in a robust way, an interface detection setup that is able to detect a range of interfaces is presented in this work. The use of a light sensor, that encompasses an array of photo detectors for different light wavelengths, and an arrangement that takes advantage of the optical properties of liquids is investigated for detecting different interfaces with varying optical features. Light sensors have the advantage of not coming in contact with the chemicals involved, as they can be located outside glass vessels. An array of simple sensors reduces the complexity of the signals to be analyzed while at the same time providing enough relevant information to feed more elaborated detection techniques, i.e., machine learning approaches.

A dataset is created from sensor measures, characterized, and related to different interface features. Data is automatically generated using a normal glass separatory funnel, connected to a peristaltic pump for controlling the level of liquid inside, but the setup can be adapted for different kinds of vessels and existing automation tools. An analysis of light refraction and reflection inside the funnel is performed. The light source is located on the opposite side to the light sensor and the whole setup is shielded from external light. The dataset obtained will allow the creation of machine learning models (e.g. neural networks) to improve the detection capabilities of the sensor and generalize it to not seen before interfaces.

This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 957189