Researchers at the NSU Physics Department’s Youth Laboratory for Photonics Technologies and Machine Learning for Sensor Systems, in collaboration with colleagues at the Institute of Laser Physics SB RAS (ILP), have developed an algorithm that helps stabilize the operation of optoacoustic gas sensors.
The primary element in this type of sensor is a cell containing the gas or gas mixture being studied. It is also a resonator for acoustic waves. When testing for concentration, the gas absorbs the radiation of the laser source and heats up. With pulsed radiation, the gas either heats up or cools down. In this case, sound waves are emitted, the amplitude is captured and measured by a special microphone. It is important that the repetition rate of laser pulses coincide with the resonant frequency of the gas cell. If this condition is met, the amplitude of the sound waves increases and the researchers can detect it and determine the concentration of the gas in the cell with a high accuracy.
The researchers explained that it is important to note that these sensors work is stable and accurate during short-term studies (several tens of seconds), but with longer ones (lasting from 10 minutes to several hours) they can give incorrect results. Long-term stability in the operation of an optoacoustic gas sensor is required in studies aimed at mapping an area when searching for oil and gas fields, in medical diagnostics (for analyzing the air exhaled by a patient), and for assessing air safety at industrial enterprises. It is possible to make their operation stable by applying mathematical algorithms. It was this aspect that the researchers at the Physics Department Youth Laboratory addressed.
Anastasia Bednyakova, Candidate of Physical and Mathematical Sciences and Senior Researcher at the Youth Laboratory described their work,
We proposed a solution to the stability problem. At the same time, we applied an optimization algorithm, namely, an algorithm for controlling the search for an extremum. It allows you to control the repetition rate of laser pulses in real time so that it corresponds to the resonant frequency of the gas cell at each moment in time. That is how we solved the problem.
The algorithm developed by the scientists at the Laboratory was introduced into the prototype of the gas sensor. After this, the NSU researchers worked with their ILP colleagues and conducted a series of experiments and tests.
Bednyakova added,
During the experiment, the temperature of the gas cell changed for a long time and over a wide range and it was demonstrated that the measured gas concentration remains constant, which means that the algorithm works correctly. In the future, we are faced with the task of optimizing the parameters of the algorithm and further improving the characteristics of the sensor using machine learning algorithms.
This research was supported by the Priority 2030 program and its results were published in the Infrared Physics & Technology journal.
