We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
µFTIR detection of tyre wear particles: A new analytical process using neural network modeling to attempt their identification
Original title: µFTIR detection of tyre wear particles: A new analytical process using neural network modeling to attempt their identification
Summary
Every time tires roll on the road, they shed tiny plastic particles that wash into our waterways and environment—but scientists have struggled to accurately detect and count them. This study developed a smarter tool using AI (a computer model trained to recognize patterns) combined with light-based scanning to spot these tire particles more reliably than older methods, successfully finding them in real pond sediment samples. Better detection tools like this are an important step toward understanding how much of this microplastic pollution is out there, which matters since we're still learning how these particles might affect ecosystems and, ultimately, human health.
• Mean spectral absorbance and first derivative of spectral standard deviation discriminate CMTTs • Dedicated neural network defines a classification zone specific to tyre wear particles • Non-flat stainless steel filters can lead to false positives or omissions • Consistent particles sizes across imaging modes demonstrate qualitative reliability • The deep learning-based methodology shows strong potential for identifying TWP Tyre wear particles (TWPs) are a major source of microplastic contamination. Detection and quantification of TWPs in the environment is a major challenge. However, several methods are currently used, with their own advantages, disadvantages and limitations. In this study, a new method based on infrared micro spectroscopy was developed. In contrast to more conventional approaches based on spectral matching of characteristic peaks, a multi-criteria designation using a heatmap was chosen. This method was developed using absorbance and average spectral variation. These criteria were chosen due to their high potential for distinguishing carbon black filled. A neural network deep learning model was trained using spectral data from various microplastics and sediments. To test the detection performance of the model, the model was tested on some samples containing a known number of tyre particles. The method demonstrated high detection efficiency without false positives. However, the few particles omitted were associated with the use of a filter not perfectly flat. Analyses testing the method on a filter containing sediment from a stormwater pond revealed the presence of numerous particles corresponding to TWPs. This method is an innovative and promising approach for further studies on TWPs. Repeated application and continuous model training should improve its reliability, including more complex environmental matrices.