This project mainly belongs to the field of computer science and technology, involving machine learning, online intelligent sensing, precision measurement and control, and bioengineering theory and technology. It has been funded by the ldquo; 863rdquo; plan, the National Natural Science Foundation of China, and the Shanghai City Science and Technology Research Project for many times. Electronic nose technology is a core application technology in industries such as production, food, light industry, and environmental protection. It has been included in the Ministry of Science and Technology's ldquo;863 rdquo;, scientific and technological support and key research and development plans for many times. Current national standards such as edible vegetable oil and industry standards such as paraffin wax all include air (smell) quality indicators, but the evaluation method completely relies on human sense of smell. On the one hand, the breath (smell) sensory assessment method has been criticized for its poor operability and lack of objectivity and fairness. The current standards and evaluation indicators are in name; on the other hand, long-term smelling of breath (smell) will cause serious harm to the body. Biological fermentation engineering involves medicine, food, environmental protection, energy and other fields, such as antibiotics, amino acid production, etc. Biological fermentation utilizes the growth and metabolic activities of microbial cells. It is a typical nonlinear time-varying process with complex influencing factors. Our country is a big fermentation country, but it is not a strong fermentation country. Today, in addition to a few physical parameters such as temperature, tank pressure, and stirring speed, many physical, chemical and biological parameters in the fermentation process cannot be detected and analyzed online. Electronic nose technology uses an array of multiple gas-sensitive sensors with sensitivity up to ppm to realize the identification and quantitative analysis of odorous substances, such as smell (smell) recognition and quantitative evaluation of quality levels; online detection and optimization control of fermentation processes and odorous pollutants, etc. In view of the current situation of foreign electronic nose instruments that are expensive, harsh using conditions, narrow application range, and lack of related products, this project has carried out research on machine learning theory and electronic nose technology and application, and achieved the following key method innovations and technological innovations: For the first time, modular machine learning cascade models and algorithms are proposed to achieve qualitative and quantitative simultaneous online analysis of multiple odorant substances; creatively use the first layer of the cascade model mdash; multiple single-output shallow neural network modules through online learning to realize on-site identification of multiple gas (smell) flavor types and origin; Use the second layer of the cascade model mdash; multiple single-output deep neural network modules to offline learning and online fine-tuning to realize online simultaneous quantitative prediction of the main gas (smell) flavor components and multiple sensory quality indicators. Invented automatic continuous online detection technology for application objects such as fermentation processes and environmental odors and on-site detection technology for application objects such as food and paraffin, including: selection and equivalent replacement of gas-sensitive sensors, annular arrangement with small sections at intervals such as gas-sensitive sensor arrays and precise constant temperature, large sample volumes and large volumes of volatile gas generation and precise constant temperature, large-flow headspace automatic sampling and precise control, and responds to environmental changes with internal ldquo; unchanged rdquo;, solving the problem of gas sampling being greatly affected by the environment, artificial dilution of components, and poor repeatability. Invented modularization of key components and integration of electronic nose instruments and intelligent big data analysis technology, including: Components such as gas sensor arrays, large-capacity volatile gas generation and automatic sampling systems, and computer control and analysis systems are modularized and integrated into one to form a small electronic nose instrument, establish big data on gas (smell) smell, and machine learning cascade models. Qualitative and quantitative analysis of complex odors solves the problems of large size, poor stability, cumbersome operation, and poor analytical capabilities of existing electronic nose instruments. This project applied for 13 national invention patents and authorized 11; applied for and disclosed 2 PCT international patents; and published more than 150 related papers, including 45 SCI papers, which he cited more than 1,600 times. Relevant achievements have been used in biological fermentation, petrochemical industry, environment, food and other fields, and have been promoted and applied in Fushun Research Institute of Petrochemical Engineering, Institute of Microbiology, Chinese Academy of Sciences, State Key Laboratory of Bioreactor Engineering and other units, producing obvious economic and social benefits.