Validation of Artificial Intelligence Time Resolved Fluorescence method for the real-time monitoring of critical pollutants in industrial and municipal effluents – VAI-TRF

we focused on a few key points: (i) development of the method for the identification and quantification of polluting substances by time-resolved fluorescence, (ii) development of the device for sample preparation and fluorescence measurement , (iii) first tests on real water containing pollutants.

We have optimized the measurement parameters of the time-resolved fluorescence device as well as the analysis of the collected data. The idea was to find the optimal ratio between the pollutant and the fluorescent probe. This ratio varies with the pollutant studied as well as the environmental conditions (water, salinity, concentration, LOD, presence of other substances). To proceed step-by-step, we focused on two families of simple pollutants as indicated in the project table: chelating molecules and pesticides. The limit of quantification is for the moment fixed at one hundred ppb given that once the process has been validated under these conditions, it will then be easy to extend it to lower concentrations.
AI learning models have been developed in order to correctly identify contaminants. Among the proposed models, the neural type was preferred because it allows the simultaneous estimation of two or more contaminants. The first results of automatic identification are encouraging but more raw data for the training of the algorithm and an optimization of the code are necessary in order to obtain a functional identification on site. We also focused on the sample preparation robot and the spectrophotometer. As for the sampler, the robot is driven via software developed on python which considers at each measurement the set of optimal parameters for quantification, based on a previously developed database which includes contaminants in different environmental conditions. The basic idea of the algorithm is to avoid unnecessary measurements, for example when there is no variation in the concentration of contaminants and to propose the optimal parameters for each site. Regarding the spectrophotometer, the current configuration is being modified to include a device for controlling the excitation wavelength by a monochromator.

The group will mainly work on optimizing the software, integrating the sampling robot with the measuring device. Several series of mesocosm experiments will be conducted. At the same time, tests will be carried out on emerging contaminants specific to sites of interest. A first prototype of the final device is under construction and its installation is scheduled for the end of 2023 for long-term site measurements (6 months).

a patent on the developed method will be drafted

Due to growing urbanization, industrialization and agriculture, natural water is increasingly threatened by the release of synthetic chemicals into the environment. Both diffuse pollution (e.g., agricultural activities) and point pollution (e.g., domestic and industrial activities) are of high concern because they alter water ecosystem functions and constitute risks for human health (e.g., contaminated food and drinking water), welfare and economy (e.g., access restriction to recreational waters or prohibition of commercial fishing). This project aims at tackling the environmental problem of domestic and industrial point sources of pollution by monitoring critical chemicals in effluents and by providing decision-makers with timely actionable knowledge in order to reduce the release of these compounds in the environment.
While conventional laboratory-based methods usually have good analytical performances, they are poorly adapted to the provision of timely and representative information. They rely on sophisticated sampling strategy (expensive and time-consuming sample transfer to offsite laboratories and sample preparation), thus needing time to provide results. In situ and online analytical techniques are thus urgently needed.
To overcome these limitations, we will validate a new concept of measure based on time-resolved fluorescence for continuous in-situ monitoring of critical chemicals such as glyphosate, phosphonates, sulphonates, chelating surfactants and PFASs. This project will be based on an earlier patented method on the dosage of limestone and corrosion inhibitors via TRF technique. The technique will be supported by using a standard addition approach and artificial intelligence training (AI-TRF). The quantification procedure will be optimized to monitor a large variety of contaminants commonly present in effluents by both laboratory and mesocosm assays feedbacks. The performances of AI-TRF prototype will be compared to those of conventional laboratory-based techniques (e.g., LC-MSMS, ICP-MS). In a second phase, we will deploy AI-TRF in real effluents to acquire information on the release of contaminants in sewage networks (signaling of exceedance of limit concentrations and production of information on the types of potential sources). In this study, we will focus on (pre)treated wastewaters because they have a less complex matrix than raw wastewaters, but the ultimate goal is to provide a tool that can be deploy at any site of a sewage network to obtain the maximal amount of relevant information.
With this project, we would like to prove that first AI-TRF should be an alternative technology for timeliness quantifications of pollutants comparing to conventional analytical methods currently used by environmental agencies. Secondly, the in-situ TRF installations could drastically increase the spatial coverage of monitoring campaign, thus generating deeper knowledge on critical pollutants, facilitating the development of policies concerning chemical production and consumption as well as on wastewater treatment strategies; thus sustainably preserving the environment. Third, artificial intelligence could support, not only the hard-working tasks normally done continuatively by technical staff, but also the implementation of new generation of state-of-the-art devices.