Client context
For alkyd resins in solvents, key quality parameters – viscosity and acid number – are most often verified using laboratory methods on samples taken after filtration. This approach ensures precision but introduces a significant delay between production and quality decision-making.
The project aimed to verify whether the Spectrally platform based on Raman spectroscopy could take on the role of a fast and repeatable „quality sensor,” delivering results in a short time and forming the basis for quality control automation and ultimately process stabilization.
Challenge.
- Lack of real-time product quality information – laboratory results are available with a delay.
- Risk of detecting deviations only after batch production is completed, which can generate losses and delay product release.
- The complex matrix of resin and solvents causes signal overlap in the spectrum, making simple assignment of a single band to one parameter difficult.
- Flammable solvents are present in the process environment, so future implementation must consider requirements for explosion-hazardous zones.
Project Objectives
- Assess the feasibility of measuring viscosity and acid number based on Raman spectra for alkyd resins in solvents.
- Build a preliminary chemometric model for predicting quality parameters.
- Define the target implementation concept, with a preferred in-line measurement after filtration, directly in the pipeline.
- Establish the direction for further model validation for predicting intermediate values.
Solution
We conducted feasibility measurements in an at-line configuration using the Spectrally platform, recording Raman spectra for samples with varied matrices and performing replicates to confirm repeatability. We optimized acquisition to integration times below 1 s and averaged spectra multi-scan to improve the signal-to-noise ratio.
Due to spectrum complexity, we applied a whole-spectrum analysis approach and tested two types of predictive models: linear regression and a neural network-based model. Ultimately, we recommend in-line measurement with probe installation after the filtration unit, which provides continuous reading, fast process response, and readiness for further process automation.
Results and Benefits
- Results demonstrated the high potential of Raman technology combined with chemometrics for monitoring alkyd resin quality parameters.
- A coefficient of determination R² > 0.99 was achieved for both viscosity and acid number.
- The predicted measurement uncertainty range for viscosity was approximately ±0.25 Pa·s, with potential for further improvement after expanding the sample database.
- Quality decisions can be accelerated from hours to seconds or tens of seconds, shifting the operational model from reactive control to proactive process monitoring.
- The recommended next step is to expand the calibration set and validate under near-process conditions.
