AIOS-2030 Artificial Intelligence Olfactory System

AIOS-2030 Artificial Intelligence Olfactory System——Constructing a new generation of olfactory perception and identification evaluation system

Introduction to Artificial Intelligence Olfactory System Technology for Juxin Chasing Wind

This system is a laboratory level artificial intelligence olfactory system, consisting of three parts:

Sample pre-treatment injection system

This section is composed of a dynamic headspace processing system. The main task is to place the sample into a headspace bottle, heat and extract odor substances into a low-temperature trap, concentrate them, and then sample them into the olfactory system. Specific technical parameters can be found in the product brochure. This system can collect odor substances at the concentration level of PPT, and more comprehensive collection and injection is an important part of the artificial intelligence olfactory system.

AIOS-2030 Artificial Intelligence Olfactory System：

This section consists of at least 10 sets of sensor arrays to form the olfactory nasal cavity. The olfactory sensor detects the interaction between odor molecules and sensor materials, converting odor information into electrical signals. Sensors include: (1) acetone, (2) organic sulfide nitrogen compounds, (3) toluene, aldehydes, ketones and alcohols, alkyl aromatic compounds, (4) aliphatic hydrocarbons, halogenated hydrocarbons, ethers, esters, pyridines, phenols and alcohols, (5) alcohols, ketones, aldehydes and aromatic compounds, (6) methane and hydrogen sulfide, (7) phenols, ketones, ethyl acetate, cyclohexanone, chlorobenzene, toluene and ethers, (8) alkanes, alkenes and aromatic compounds sensitive; Alkanes, olefins and hydrogen, (9) alkanes, carbon monoxide, aldehydes, alcohols, nitrogen oxides, ketones and aldehydes, (10) sulfides, nitrides, carbides, hydrocarbons and nitrogen oxides. The selection of sensors needs to be optimized according to specific application scenarios.

In addition, a good nasal airway rotation system can better respond to all gas components, reduce residue through inerting treatment, and avoid cross contamination.

Olfactory neural network

Olfactory neural networks are responsible for feature extraction and pattern recognition of odor signals, generating olfactory perception data. Olfactory neural networks typically include the following levels:

Input layer: receives multidimensional data from sensors, such as gas concentration, temperature, humidity, etc

Hidden layer: Extract odor features through multi-layer perceptrons or convolutional neural networks. The design of hidden layers needs to consider the depth and width of the network to balance computational complexity and feature extraction capabilities.

Output layer: Generate odor classification or concentration prediction results. The design of the output layer needs to be optimized based on specific tasks, such as multi classification tasks or regression tasks.

High quality training data is the key to the performance of neural networks. The digital signals generated by the nasal system require data preprocessing, such as denoising and normalization, to improve the quality and consistency of the original data. Then perform data annotation, which involves manually or automatically annotating odor data to generate labels required for supervised learning.

Pattern recognition is the process of classifying or predicting the concentration of odors through neural networks. Our company's artificial intelligence olfactory system currently integrates four pattern recognition algorithms:

1. The KNN (K-Nearest Neighbors) algorithm is an instance based supervised learning classification method that calculates the distance between the sample to be classified and the K nearest neighbors in the training data, and determines its class assignment based on the majority voting principle.

2. Support Vector Machine (SVM) is a supervised learning binary classification algorithm, whose core idea is to find the optimal hyperplane by maximizing the classification interval. It is suitable for linearly separable and nonlinearly separable data and is widely used in fields such as pattern recognition.

3. Random forest is an ensemble learning algorithm based on decision trees, which improves the accuracy and robustness of models by constructing multiple trees and combining their prediction results. It is widely used in tasks such as classification, regression, and feature selection.

4. Gradient Boosting is an ensemble learning algorithm that optimizes model performance by iteratively training weak learners (usually decision trees) and accumulating their predicted results, suitable for regression and classification tasks.

New algorithm models can also be introduced for sample identification and evaluation according to testing needs.

Finally, in order to successfully complete the training of the olfactory system, our company also provides "Companion Run" technical services to ensure the smooth implementation of the project system. The main task of the "Companion Run" service is to assist users in training the olfactory system, seeking better sensor combinations and optimizations according to project requirements, selecting big data model algorithms for testing, algorithm fusion, and improving the ability and stability of discrimination rating.