IOT Based Gas Leakage Detector

IOT Based Gas Leakage Detector — Topics for IoT Students

A COMPACT, EASY-TO-USE LEAK DETECTOR FOR VACUUM SYSTEMS

BASED ON THE MQ-8 SENSOR S.M. Kovalov,1, ∗ I.V. Beznosenko,1 A.V.

Vasyliev,1 and G.V. Sotnikov1 National Science Center Kharkiv Institute of Physics and Technology Institute of Plasma Electronics and New Acceleration Methods 1, Akademichna St., Kharkiv, 61108, Ukraine (Dated: November 24, 2025) This paper presents the development of a compact leak detector for vacuum systems that operates using hydrogen as a tracer gas, detected by a semiconductor MQ-8 sensor. The sensor is connected to an Arduino microcontroller, enabling digital signal processing and real-time visualization through arXiv:2511.16686v1 [physics.ins-det] 14 Nov 2025

the Processing software environment. The designed device is capable of detecting even small hydro- gen leaks caused by imperfect sealing of vacuum connections. It is characterized by simplicity, low manufacturing cost, and suitability for laboratory use as an alternative to helium leak detectors.

Experimental tests confirmed the effectiveness and stability of the developed leak detector during long-term measurements. PACS: 07.30.Dz, 07.07.Df, 07.05.Hd

I. INTRODUCTION propagate along the entire length of the chamber without

significant energy losses or trajectory deviations, ensur- In experiments involving electron acceleration within ing the stability and reproducibility of the experiment. vacuum chambers, particularly in dielectric laser accel- During one of the experiments with the electron gun, erator (DLA) systems, a stable high vacuum is essential a decrease in the VIT-2 vacuum gauge reading to 1.3 mV to minimize scattering and ensure stable propagation of was recorded, which, according to its calibration curve, the electron beam.

For conducting physical experiments corresponds to a pressure of about 1.5 × 10−1 mmHg on DLA, the computer simulation of which is described (≈20 Pa). At this pressure, in a chamber with a volume in [1–3], the operating pressure in the vacuum chamber of 200 L, air electric breakdown processes begin to occur, must be maintained at the level of 10−4 –10−5 Pa. At such making it impossible to conduct experiments with the low pressures, the mean free path of gas molecules greatly electron gun.

This prompted a diagnostic check of the exceeds the physical dimensions of the vacuum chamber, vacuum system to identify the location of the leak. allowing the electron beam to pass through with negli- gible energy loss or defocusing. The mean free path ℓ is determined by an expression derived from the kinetic II. ANALYSIS OF EXISTING METHODS AND theory of gases [4, 5]: DEVELOPMENT PREREQUISITES

kB T In modern vacuum technology, there is a wide range ℓ= √ , (1) 2d2 pπ of methods for leak detection. The most common and highly accurate solution is helium mass spectrometer where kB is the Boltzmann constant, T is the gas tem- leak detectors, with sensitivities reaching 10−11 –10−12 perature in the chamber, d is the effective diameter of an m3 ·Pa/s. Their operating principle is based on the detec- air molecule, p is the pressure. tion of helium atoms that penetrate the vacuum system In our experiments, the operating pressure in the through microdefects and are measured via the ion cur- chamber is on average p = 5 × 10−5 Pa at a temperature rent in the mass spectrometer.

Such systems are used of T = 293 K, which corresponds to high-vacuum condi- for the certification and diagnostics of sealed setups in tions. For air, which consists of 99 % nitrogen and oxy- the electronics, cryogenic, and nuclear industries [7, 8]. gen molecules, the effective molecular diameter is taken However, the practical use of helium leak detectors in lab- as d = 3.6 × 10−10 m [6]. Substituting these values into oratory conditions, especially under limited resources, is equation (1), we obtain the mean free path of the gas associated with a number of significant challenges.

First, molecules of approximately 137 m, which is many times these instruments are expensive: the price of typical greater than the characteristic dimensions of the used models on the Ukrainian market is at least around USD vacuum chamber (0.5 m). This indicates an extremely 2,500 [7]. Second, their operation requires constant ac- low probability of collisions between electrons and resid- cess to a source of high-purity helium (99.999%) [8, 9]. ual gas molecules.

Consequently, the electron beam can Additionally, regular calibration and maintenance of vac- uum pumps are necessary, which further increases oper- ational costs. ∗ Corresponding author Moreover, mass spectrometer-based systems are bulky Email address: covalov.sergiy@gmail.com (S.M. Kovalov) and heavy (15–25 kg) [7], making them inconvenient for

rapid use in laboratories with limited space. Their start- of explosion, making the method particularly promising up and stabilization to operational mode take several for laboratory conditions. minutes, and monitoring is primarily stationary. In the For hydrogen detection, a semiconductor sensor MQ- case of a sudden vacuum loss or the need for rapid diag- 8 was selected.

Its operation is based on the change in nostics, such systems prove inefficient, as their use con- resistance of the tin dioxide (SnO2 ) sensitive layer upon sumes significant time and resources. interaction with reducing gases [10]. This sensor is low- In practical vacuum research, classical leak detection cost, easy to interface, and sufficiently sensitive (in the methods, such as blowing with volatile liquids (alcohol, range of 100 – 10,000 ppm [10]) for detecting leaks in freon) or applying soapy solutions, are used to localize laboratory settings.

When combined with an Arduino leaks at moderate pressures. However, their effectiveness Uno microcontroller, it enables continuous monitoring of sharply decreases in the case of significant leaks, since the gas concentration and real-time data visualization using system pressure becomes too high to detect small changes the Processing software environment. in vacuum readings. Under these conditions, the evapo- ration of liquid indicators occurs too quickly, and the sen- sitivity of standard thermocouple vacuum gauges (VIT- 2, VIT-1, VIT-3) is insufficient to register such changes IV.

OPERATING PRINCIPLE OF (in our case, at a pressure of about 20 Pa, the ionization- SEMICONDUCTOR SENSORS (MQ FAMILY) thermocouple gauge VIT-2 did not detect any signal fluc- tuations). Furthermore, the use of soapy solutions is often limited MQ-type sensors implement a chemiresistive principle. by the design features of the equipment. If the setup has The sensitive element consists of a semiconductor film, a complex geometry or double walls (inner and outer, as typically based on SnO2 with catalytic additives [10], in the considered vacuum chamber), the space between which is heated by an integrated heater to several hun- them becomes inaccessible for visual inspection.

This dred ◦ C. In the presence of reducing gases (H2 , etc.), makes it impossible to use foaming methods to determine molecules adsorb onto the surface, altering the charge the leak location. carrier density in the near-surface layers. This leads to Thus, in our case, traditional methods for leak diag- a change in the resistance RS of the sensitive element. nostics are ineffective in the presence of significant leaks Thus, the gas concentration is converted into an electri- and a complex vacuum chamber design.

The absence of a cal signal. helium mass spectrometer leak detector and the imprac- According to the manufacturer’s technical documenta- ticality of acquiring one quickly in a laboratory setting tion (MQ-8 Datasheet) [12], the sensor has the following may necessitate the search for an alternative approach characteristics. The operating temperature of the sen- capable of providing sufficient sensitivity at low cost and sitive element is maintained by the built-in heater and minimal preparation time. should be 300 ◦ C.

A warm-up period of at least 60 sec- Below, we describe the development of a compact leak onds is required after power-on to stabilize the sensor’s detector technology that can be quickly fabricated in lab- characteristics. The sensitivity curve is nonlinear and is Rs oratory conditions without significant material expenses. described by a power-law dependence R 0 on hydrogen concentration in the range of 100 ppm to 10,000 ppm [10].

The sensor demonstrates high selectivity to hydro- III. HYDROGEN AS A TRACER GAS AND gen, but exhibits cross-sensitivity to carbon monoxide MQ-8 SENSOR AS DETECTOR and methane, as indicated by the standard curves in the datasheet. An analysis of existing approaches for detecting leaks An important operational factor is the drift of the sen- in vacuum chambers has shown that using hydrogen sor’s resistance over time due to aging of the sensitive as a tracer gas represents an optimal compromise be- layer and exposure to interfering gases, which necessi- tween sensitivity, cost, and ease of implementation.

This tates periodic recalibration. In addition, the readings are method has been successfully applied in a number of in- influenced by ambient temperature and humidity. There- dustrial and research setups [10], confirming its reliability fore, the manufacturer recommends conducting measure- and reproducibility. ments at relative humidity not exceeding 85% and tem- Hydrogen has a minimal molecular diameter (about peratures between 20 and 40 ◦ C. 2.9 Å) [11] and high diffusivity, allowing it to penetrate Unlike mass-spectrometric helium leak detectors that even the tiniest defects in vacuum joints.

Furthermore, operate on the physical principle of detecting helium ions, the gas can be easily generated using a laboratory elec- the MQ-8 sensor is based on changes in the electrical con- trolyzer, eliminating the need for expensive cylinders and ductivity of a semiconducting SnO2 layer upon interac- auxiliary equipment. Thus, a hydrogen-based method tion with hydrogen molecules. This approach provides a provides a combination of high efficiency, compactness, simple design, low cost, and the possibility of implement- and low cost, while the supply of hydrogen in small flows ing a compact digital device with sufficient sensitivity for (∼10 L/h) ensures safe operation and eliminates the risk laboratory applications.

V. ELECTRICAL MODEL AND CONNECTION VI. METHOD FOR LEAK DETECTION AND

SCHEME OF THE MQ-8 SENSOR LOCALIZATION

To localize leaks in the vacuum system, a gas-blowing

The sensor output is formed by a voltage divider Vout method was implemented using hydrogen as a tracer across RS (the sensor resistance) and an external resistor gas. The hydrogen source was a laboratory electrolyzer RL (”load resistor”), connected to ground: (Fig.1), consisting of a unit with plate electrodes im- mersed in an aqueous electrolyte and a separate gas collection flask. When a direct current was applied to the electrodes, electrolysis of water occurred, releasing RL Vout = VCC · , (2) gaseous hydrogen at the cathode.

The electrolyzer design RL + Rs allowed continuous operation and enabled adjustment of the gas generation rate by varying the current. The gas outlet was connected to a fitting with a flexible hose used which describes a voltage divider, where VCC is the to blow the external surface of the chamber. To prevent supply voltage (typically 5 V), RL is the external load overpressure and enhance safety, a water seal (acting as resistor, and RS is the resistance of the sensor’s sensitive a check valve) was employed.

Hydrogen was supplied to element, which depends on the hydrogen concentration. the system as a slow and steady flow, which allowed con- Under normal conditions, at low hydrogen concentration trolled blowing and minimized background fluctuations. RS is high, so the voltage drop across the load resistor is small, and the output voltage Vout remains low. As the hydrogen concentration increases, RS decreases, leading to a rise in the output voltage (2).

Thus, changes in the chemical composition of the air near the sensor are di- rectly converted into an electrical signal suitable for mea- surement. The MQ-8 sensor output can be connected to an analog input of an Arduino Uno microcontroller (e.g., pin A0). The microcontroller contains a built-in 10-bit analog-to-digital converter (ADC), which allows convert- ing voltages from 0 to 5 V into digital codes ranging from 0 to 1023.

This means that small changes in the sensor output voltage can be detected with a resolution of ap- proximately 4.9 mV [10]. For more accurate detection of weak signals, the inter- nal reference voltage of the Arduino (1.1 V) can be used, FIG. 1. Laboratory hydrogen electrolyzer used in the experi- which increases the ADC sensitivity almost fivefold.

In ment. this case, the value of the load resistor RL , must be ad- justed so that the sensor output voltage does not exceed The gas was directed onto the outer surface of the vac- 1.1 V even at maximum hydrogen concentration. uum chamber in the area of the suspected leak. At the same time, the forevacuum pump created a pressure drop The operation of the entire system can be described inside the system, ensuring air intake through the defect. as follows.

The built-in heater of the MQ-8 sensor main- If a leak was present in the blown area, part of the hydro- tains a high temperature of the sensitive layer, enabling gen penetrated into the chamber and was then carried by adsorption and desorption of gases. When the hydro- the air flow toward the forevacuum pump outlet. gen concentration changes, the layer resistance changes, At the outlet line of the forevacuum pump, an MQ-8 and according to the voltage divider formula, the output sensor (Fig.2) was installed, operating in the continuous voltage varies.

This voltage is converted by the Arduino hydrogen concentration monitoring mode. To increase ADC into a digital code, which can then be processed sensitivity, an additional reservoir (Fig.3) was used, in in software: averaged, filtered, and analyzed to detect which the gas flow rate decreased and partial accumula- threshold exceedances. tion of hydrogen occurred. The sensor was placed in the Thus, the combination of the sensor’s analog model upper part of this reservoir, which ensured more reliable and digital processing allows detection of both large leaks detection of gas concentration.

Its electrical signal was (hydrogen concentrations above 1,000–2,000 ppm) and read by an Arduino Uno microcontroller [13] (Fig.2) at a small leaks, where hydrogen changes are on the order of sampling frequency of about 10 Hz, after which the data 50–100 ppm relative to the background level. Visualiza- in digital form were transmitted to a personal computer tion in the Processing environment [14] further facilitates via a USB interface (Fig.4). operation by displaying real-time concentration changes To improve the informativeness of measurements, dig- as graphs. ital filtering algorithms implemented on the Arduino mi-

crocontroller were used. The main data processing tool time, Xt is the current measurement from the sensor, and was the Exponential Moving Average (EMA), which al- α is the smoothing coefficient (0 < α < 1). lowed smoothing out rapid noise fluctuations of the sen- To eliminate false spikes, a jump limiter was applied. sor output signal. The algorithm can be expressed as a When a sharp change in readings exceeded the preset recurrent equation: threshold ∆max the value was corrected according to the formula:

St = α · Xt + (1 − α) · St−1 , (3)

Xt′ = Xt−1 + sign (Xt − Xt−1 ) · ∆max , (4) where St is the smoothed signal value at the current which made it possible to minimize the effect of ran- dom impulsive noise. The threshold for sensor triggering was set dynami- cally and defined as the sum of the baseline level and a value proportional to the current standard deviation of the signal:

T = B + k · σ, (5)

where T is the threshold, B is the baseline signal value, FIG. 2. MQ-8 sensor and Arduino microcontroller. σ is the standard deviation over a fixed time interval, and k is the coefficient determining system sensitivity. This approach allowed adaptive consideration of background variations and increased the reliability of detecting low hydrogen concentrations.

Data visualization was performed in the Processing environment (Fig.5) [14]. On the computer screen, a real-time graph of the sensor signal changes was dis- played. When hydrogen entered the system, a character- istic spike significantly exceeding the level of background fluctuations was registered.

This allowed the operator to promptly record the moment of leakage and adjust the position of the blowing point. A software-controlled sequence of blowing different sections of the chamber sur- face was implemented. The appearance of a sharp signal FIG. 3.

Additional reservoir installed at the outlet of the spike during hydrogen supply to a specific area made it forevacuum pump. possible to localize the defect with high accuracy and confirm the correctness of the applied method.

VII. EXPERIMENTAL RESULTS

During the tests, characteristic changes in the sensor

output signal were recorded when blowing specific areas of the chamber. The reproducibility of the response dur- ing hydrogen supply confirmed the presence of a leak. A typical example of real-time observation of the sig- nal during leak detection is shown in Fig.5.

Along the t-axis (horizontal), time was plotted, and along the A0- axis (vertical), the analog value of the sensor output sig- nal was plotted. Before the start of measurements, the sensor was software-calibrated to establish the baseline FIG. 4. Diagram for leak detection in the chamber: 1 is signal value corresponding to normal conditions without vacuum chamber; 2 is forevacuum pump; 3 is hydrogen elec- hydrogen (in our case, this value was 15).

When this trolyzer; 4 is MQ-8 sensor; 5 is direction of hydrogen flow; level was exceeded (the curve going above the baseline), 6 is Arduino (digitization and data transfer); 7 is personal the program registered the event ”DETECT = YES” and computer with visualization software. marked the leak occurrence.

Thus, the change in hydrogen concentration was vi- signal level for about 30 seconds. After a temporary in- sualized as a characteristic spike on the graph, allowing terruption of hydrogen supply at the 69th second, the quick identification of a vacuum chamber defect at the analog signal dropped from 48 to 34 counts, and when sensor’s position and evaluation of the sensor’s sensitiv- hydrogen was supplied again at the 105th second to the ity.

At the 45th second, the baseline level of the analog same location, the signal increased to 115 counts within signal (15 counts) was crossed, indicating the presence 60 seconds, clearly demonstrating leak localization and of a leak. The graph showed a gradual increase in the dynamics.

FIG. 5. Dependence of the Arduino analog output signal (ADC, from 0 to 1023) on time.

VIII. CONCLUSIONS venience and safety of transport and operation, despite

the use of hydrogen during leak detection.

The developed device based on the MQ-8 sensor and

Arduino platform demonstrated efficiency as a simple and low-resource tool for detecting leaks in vacuum sys- tems. The tests were conducted on our vacuum setup ”KASPAR” (Fig.6), representing a cylindrical chamber of 200 L volume equipped with a forevacuum pump 2NVR-5DM and a diffusion pump, with pressure mea- surement using thermocouple and ionization gauges VIT- 2. The device proved convenient in operation, provided real-time data visualization, and allowed recording of hy- drogen entering the working area of the setup.

Initial measurements revealed pressure instability and the in- ability to achieve the target vacuum level of 10−4 –10−5 Pa, suggesting the presence of leaks. With the devel- oped device, the leakage areas were localized and subse- quently eliminated, after which the vacuum chamber was restored to a pressure of about 4 × 10−5 Pa. After elim- inating leaks, the pumping rate increased significantly, ensuring stable and rapid achievement of the required vacuum for further experiments on dielectric laser ac- celeration (DLA) of electrons.

The device can also be used in other laboratory vacuum chambers due to its compactness, simplicity, low production cost, and con- FIG. 6. Vacuum setup ”KASPAR”.

The working range of the sensor was 100–10,000 ppm, conditions, where simplicity and minimal cost are criti- which corresponds to the technical documentation data. cal. During the experiments, the sensitivity to other gases (methane, butane) was also tested, confirming cross- selectivity. However, the highest response was observed ACKNOWLEDGMENTS specifically for hydrogen, which allows the developed leak detector to be considered as an affordable alternative to The study is supported by the National Research Foun- helium mass-spectrometric leak detectors in laboratory dation of Ukraine under the program ”Excellent Science in Ukraine” (project # 2023.03/0182).

[1] A. Vasyliev, G. Sotnikov, I.

Beznosenko and O. Svis- of Gases // Microporous and Mesoporous Materials. Vol. tunov.

Comparison of the Performance of Electron Guns 343, 2022, ISSN 1387-1811. https://doi.org/10.1016/ with BaO and Nanotipe Thermionic Cathodes for use j.micromeso.2022.112156 in DLA Experiments // In 2024 IEEE 7th International [7] Helium Leak Detector, Pribory.Online. Conference on Smart Technologies in Power Engineering URL: https://pribory.online/ and Electronics (STEE). Kyiv, Ukraine, 24-26 Septem- p2201536685-gelievyj-techeiskatel-vakuuma.html ber 2024, pp.

TT3.17.1-TT3.17.5. https://doi.org/10. [8] Pfeiffer Vacuum GmbH. Leak Detection 1109/STEE63556.2024.10747818 Compendium. Asslar (Germany), April [2] I.V.

Beznosenko, A.V. Vasyliev, G.V. Sotnikov.

Simu- 2013. https://www.pfeiffer-vacuum.com/ lation of Deflecting System Based on Permanent Mag- media/documents/leak-detection-know-how/ nets with a Non-Uniform Magnetic Field to Register Ac- leak-detection-compendium-pfeiffer-vacuum.pdf celerated and Decelerated Electrons Generated in Di- [9] INFICON. How to Save Helium in (Vacuum) Leak electric Laser Accelerator // Problems of Atomic Sci- Testing. 2023: https://www.inficon.com/en/news/ ence and Technology. 2023, No. 6(148), pp. 110-115. 2023-12-at-how-to-save-helium-in-vacuum-leak-testing https://doi.org/10.46813/2023-148-110 [10] D.A.H.

Fakra, D.A.S. Andriatoavina, N.A.M.N. [3] I.V. Beznosenko, A.V.

Vasyliev, G.V. Sotnikov, and G.O. Razafindralambo, K.A.

Amarillis, J.M.M. Andria- Krivonosov. Comparative analysis of simulation results of mampianina.

A simple and low-cost integrative sensor dielectric laser acceleration of non-relativistic electrons system for methane and hydrogen measurement // Sen- in transparent and reflective periodic structures // Prob- sors International. 2020, Vol. 1, 100032, ISSN 2666-3511. lems of Atomic Science and Technology. 2024. No. 6(154), https://doi.org/10.1016/j.sintl.2020.100032 pp. 42-48. https://doi.org/10.46813/2024-154-042 [11] D.P.

Halter, R.A. Klein, M.A. Boreen, B.A.

Trump, C.M. [4] L.D. Landau, and E.M. Lifshitz.

Physical Kinetics. Ox- Brown, J.R. Long.

Self-adjusting binding pockets en- ford and New York: ”Pergamon Press”, 1981, Vol. 10 of hance H2 and CH4 adsorption in a uranium-based metal- Course of Theoretical Physics, pp. 35–38 (Section: Ki- organic framework // Chemical Science. 2020, Vol. 11, netic Theory of Gases – Mean Free Path). pp. 6709–6716. https://doi.org/10.1039/d0sc02394a [5] Tec-Science. Mean Free Path and Collision Fre- [12] Hanwei Electronics. MQ-8 Hydrogen Gas Sensor. quency – Derivation from Kinetic Theory of Datasheet. https://www.mouser.com/datasheet/2/ Gases.

URL: https://www.tec-science.com/ 830/MQ_8-1951103.pdf thermodynamics/kinetic-theory-of-gases/ [13] Arduino. Arduino Uno. Datasheet / User Manual.

Ar- mean-free-path-collision-frequency/ (accessed duino Official Website: https://store.arduino.cc/ October 2025) products/arduino-uno-rev3 [6] A. Kretzschmar, V. Selmert, H.

Kungl, H. Tempel, R.- [14] Processing. Integrated Development Environment for Vi- A.

Eichel. Application of a Tailorable Carbon Molecular sual Programming. Processing Foundation: https:// Sieve to Evaluate Concepts for the Molecular Dimensions processing.org/