IRLAM Overview
The most distinctive elemental technology in IRLAM is HORIBA's unique Concentration Calculation Algorithm (World's first*).
In gas measurement, it is important to reduce interference and various effects caused by gases other than the measurement target.
HORIBA has established a method that extracts "features" from the measured gas absorption signal and uses them to correct the interference with a small amount of calculation.
The advantages of this feature-based concentration calculation is that it needs dramatically smaller amounts of calculation and will not lose accuracy compared to the conventional spectrum fitting method. Due to this, while a high-performance computer is required to perform real-time measurements using conventional methods, the concentration calculation by IRLAM can be fully supported by a general-purpose microcomputer embedded in the device.
From here on, Concentration Calculation Algorithm is described in a way that is easier for everyone to understand.
*According to our in-house research, 2021. Registered patents are as follows (as of May 2024)
(Japanese patent No. 06886507, US patent No. 11030423, European patent No. 3674690B, Chinese patent No. 201911335317.X, Indian patent No. 515821B)
Materials have the property of absorbing certain types of light. For example, a red apple looks red because the surface of the apple absorbs light with wavelengths other than red, and conversely, reflects light with red wavelengths.
This phenomenon occurs not only with visible light such as red and yellow, but also with infrared light, which is widely used to measure the concentration of gases. Using this principle, infrared gas analyzers measure gases by placing the gas in a gas cell to be measured, passing the light through it, and detecting the amount of absorbed light by a detector at the endpoint.
The wavelength of the infrared light that is easily absorbed differs depending on the gas, so by determining the amount of absorbed light, it is possible to measure "which type of gas" and "how much concentration” is present.
Measurement of gas concentration using infrared absorption
When multiple gases exist, the measured absorption signal is divided into multiple peaks, making it difficult to determine which type of gas is absorbed with the amount of infrared light (interference effect). Therefore, in order to accurately measure the concentration of the gas to be measured, various computational devices are used.
Here, we call the gas to be measured as the "target gas," and the other gases the "interfering gases”.
Difference in absorption signal between single gas component and multiple gas components
In conventional infrared gas analysis methods, spectral fitting and multivariate analysis are used to calculate the concentration. This method is used to calculate the concentration of the target gas by (1) converting the raw signal from the detector obtained from the measurement into an absorption spectrum in the wavelength domain, and (2) by comparing the absorption spectrum with a reference spectrum prepared in advance for each gas component through numerous calculation steps.
While this method can calculate the concentration correctly, it takes a long time to calculate, and a high-performance computer is required to calculate the results in real time.
In IRLAM, the raw signal of the detector obtained from the measurement is (1) turned into an absorption signal in the time domain, and (2) the features of the absorption signal are replaced by ’feature quantities,’ which are abstracted as numerical values. This is a very unique concept of HORIBA in the field of gas analysis.
By comparing these feature quantities with the pre-measured unique feature quantities of the target gas and interfering gas, the system recognizes the degree of mixing of each component based on the relationship and calculates the concentration.
Compared to conventional spectrum fitting method, which requires hundreds of data points to calculate, this feature-based approach can calculate the concentration using only a few numerical values, thus significantly reducing the amount of calculations. However, this does not degrade the accuracy of the measurement because the feature quantity contains enough information necessary for measurement.
Comparison of concentration calculation methods between conventional technology and IRLAM
As described above, HORIBA's Concentration Calculation Algorithm can significantly reduce the amount of calculation, and thus a microcomputer embedded in a device can sufficiently handle the processing. Therefore, products equipped with IRLAM can realize an unprecedentedly compact device configuration while ensuring safety and robustness as an industrial instrument.
An example of microcomputer
In addition to the interfering gas, various factors such as wavelength shift of the laser and change in absorption characteristics of the gas due to disturbance effects such as ambient temperature and pressure also affect the measurement signal in actual measurement.
Then, how do you capture such disturbance effects as features? In fact, this is also where HORIBA's expertise lies.
HORIBA actually measures the absorption signals of target gases and interfering gases under various environmental conditions such as temperature and pressure, and stores the feature quantities that take into account disturbance effects in the IRLAM analyzer in advance. What kind of features are extracted is based on HORIBA's experience and knowledge of gas analysis accumulated over many years. This know-how has enabled us to realize highly reliable measurement with high resistance to environmental impact.
This enables products equipped with IRLAM to provide stable measurement results even in real-time measurements in harsh environments, such as automobile exhaust gas testing and oil production processes that require explosion-proofing.
IRLAM analyzers resistant to environmental influences
A feature quantity is a numerical value that represents the characteristics of the object to be analyzed.
The term feature quantity is often used in the field of AI (machine learning), such as image recognition. In the same way, the detector is able to recognize what is in the image. In the same way, by extracting the feature quantities of the target component and the interference component, which have been learned (calibrated) in advance, from the raw signal of the detector, it is possible to recognize (quantify) how much concentration of the target component and the interference component are mixed respectively from the relationship between the feature quantities.
Masz pytania lub prośby? Skorzystaj z tego formularza, aby skontaktować się z naszymi specjalistami.
