This paper presents a sensor system that uses low-cost spectral sensor technology and artificial intelligence to record and evaluate the composition and condition of cooling lubricants online in the process. For this purpose, MEMS-based spectral detectors are used, which measure the near-infrared spectrum of the cooling lubricant directly in the process. The recorded spectra are preprocessed on the sensor system and then analyzed using AI methods. AI-based methods offer innovative possibilities for evaluating NIR spectra, which offers various advantages when analyzing MEMS-based spectral data. First of all, classification is used to select a suitable analysis model for further processing. The next step involves the use of artificial neural networks to improve the analysis of NIR spectra with regard to various coolant parameters such as type, concentration, pH value and nitrate. This type of concatenated analysis can significantly improve the reliability and robustness of the methods used. The condition of the cooling lubricant can thus be determined directly in the process, moreover forecasts and recommendations for action can be determined in order to increase the service life. By integrating the sensor system into a cloud environment, data can also be stored and processed centrally to meet the requirements of a smart factory.

In addition to the design and functional principle of the sensor system, the paper also provides an insight into the evaluation and analysis of the sensor's spectral data. Typical challenges are discussed and the use of AI to improve the analysis is explained. Various AI-based methods and approaches are presented, evaluated and their results are discussed in conclusion. The article thus provides an insight into an innovative sensor system which, through the targeted use of AI, opens up new possibilities for automated monitoring in metalworking production plants.