Over the past ten years, various machine learning techniques have been incorporated into a range of measurement systems for multiphase flow measurement and combustion process monitoring. Such techniques in conjunction with low-cost sensors and sensor arrays provide either unique solutions to some measurement challenges or offer more cost-effective or complementary options to other possible methods. The established or potential applications of machine learning in measurement and instrumentation appear wide-ranging, but the underlining principle, advantages, and limitations are similar. This presentation will review recent advances in the applications of data-driven modeling techniques to the measurement of gas-liquid, gas-solids, and liquid-solids mixture flows and the advanced monitoring of combustion processes. These include the mass flow rate measurement of air-oil two-phase flow, carbon dioxide two-phase flow, slurry flow, and pneumatically conveyed pulverized fuel in a range of industrial sectors. Meanwhile, systems that incorporate machine learning algorithms for the online continuous identification of pulverized fuel, burner condition monitoring, and combustion plant optimization will be introduced. Results from recent experimental programs and trials on industrial-scale test plants will be reported.
