The assessment of odour and/or smell in indoor environments has basically relied on sensory testing by human volunteers. In recent years, there has been active development of olfactory sensing devices and electronic noses based on semiconductor sensors. On the other hand, the development of a digital twin, which mathematically describes the olfactory mechanism in the human nasal cavity and reproduces it on a computer, is also being researched. This paper presents the current status and future prospects for the development of an in silico nose based on computational fluid dynamics and physiologically based pharmacokinetic models.

In recent years, with the advancement of AI technology and big data analysis techniques, expectations have grown for the wide range of applicability of e-Nose-type odor sensors.

In this article, a brief history of e-Nose type odor sensor is explained as well as a rogress of transducer technologies.

In addition, Aromabit products and their application examples are introduced, showcasing their capabilities and application examples. The paper further explores the potential of this technology in odor database creation and early fire detection, serving as illustrative examples of its diverse applications.

In this study, we have developed a highly sensitive and compact thermally controlled artificial olfactory system (TC-OSS) consisting of 16 sensitive membranes (PCM) with excellent heat resistance and a microheater. Five reference VOCs with odor intensity of 2 were measured using TC-OSS. A machine learning model was constructed based on the feature values (height, slope, etc.) calculated from the shape of time series data (5 classes×40 data). The accuracy of the learning model was evaluated by cross-validation (number of repetitions＝100), and it was successfully discriminated with a high accuracy rate of 95.8±2.7%.

Expectations for the practical application of smell sensors are growing year by year in various fields. There are several principles of smell sensors already in the market and under research and development, each with own characteristics. In this study, we evaluated the odor exhaled of by smoker using an smell sensor utilizing MEMS technology of I-PEX Inc.. 180 different types of chemically sensitive films are available at I-PEX Inc., and the odor response patterns of each sensitive films are different. To enhance sensing, we report the measurement results of odor exhaled without cigarettes and with four different types of cigarettes, using the smell sensor by utilizing different measurement and data analysis methods.

Taking into account the calculation procedures provided by Japan Environment Agency Notification No.63 of 1995, this study focuses on multiple odor index calculation procedures to elucidate the characteristics of the distribution properties of the odor index obtained according to these calculation procedures. Specifically, we present the distribution of an odor index obtained without considering the distribution of odor thresholds among people or accidental correct answers, along with the distribution of an odor index obtained by considering these factors using a probabilistic approach. Our findings reveal that the odor index calculated using individual thresholds, whether as trimmed mean or median, is more likely to be closer to the value obtained from the average olfactory sensitivity of panels, resulting in a unimodal distribution centered around this value. This probability decreases as the value diverges from this average. In contrast, the calculation procedure based on average correctness rates demonstrates a tendency for the probabilities of the odor index to be more widely dispersed than those procedures using trimmed mean or median.