METHODOLOGY FOR BUILDING A NEURAL NETWORK TO PREDICT FALSE ALARMS FROM SMOKE DETECTORS IN A FIRE ALARM SYSTEM

Elina Rahmatullina; Ildar Khafizov; Zukhra Pavlova

doi:10.61260/1998-8990-2026-1-184-192

UDC 614.849

METHODOLOGY FOR BUILDING A NEURAL NETWORK TO PREDICT FALSE ALARMS FROM SMOKE DETECTORS IN A FIRE ALARM SYSTEM

Published in Problems of risk management in the technosphere · Volume 2026, Issue 1 · Pages 184–192 · Rubric: Fire safety

DOI: https://doi.org/10.61260/1998-8990-2026-1-184-192

Received: 13.01.2026 Accepted: 05.02.2026 Published: 16.04.2026

Authors

Elina Rahmatullina

ORCID

Ildar Khafizov

ORCID SPIN

Zukhra Pavlova

Elina Rahmatullina

Ufa State Petroleum Technological University senior lecturer · department of mathematics · Russian Federation · Ufa

Ufa State Petroleum Technological University postgraduate student · department of fire safety · Russian Federation · Ufa

Ufa State Oil Technical University starshiy prepodavatel' · Ufa

ORCID

Ildar Khafizov

Academic degree: doctor of technical sciences

Science rang: professor

Ufa State Petroleum Technical University professor · department of fire and industrial safety · Russian Federation · UFA

ORCID SPIN

Zukhra Pavlova

Academic degree: doctoral candidate of technical sciences

Science rang: docent

Ufa state petroleum technological university professor · electrical engineering and electrical facilities of enterprises · Russian Federation

ABSTRACT

This article develops a Python-based neural network for predicting false alarms in smoke alarms within a fire alarm system. This research is relevant due to the high rate of false alarms, which disrupt facility operations, increase personnel workload, and reduce confidence in the alarm system. The novelty lies in the transfer of reliability and false alarm root cause analysis issues to predictive analytics. A formalized approach to predicting false alarms using telemetry time series and operational logs is proposed, along with a feature structure aligned with typical causes of failures and false alarms. The paper describes the data requirements, episode labeling logic, preprocessing scheme, architecture of a recurrent model, and quality criteria. Particular attention is paid to considering operational factors (installation, ventilation, dust levels, power supply instability, and room conditions) when generating features. The goal of this work is to propose a reproducible method for constructing a false alarm prediction model that can be subsequently integrated into maintenance procedures. The solution utilizes source analysis, comparative analysis of approaches, formalization, and design of a learning algorithm. This article is helpful for developers of fire alarm systems and researchers of applied machine learning in industrial diagnostics.

KEYWORDS

fire alarm, smoke detector, false alarms, prediction, time series, LSTM, Python, machine learning, reliability, diagnostics

Text

Text (PDF)

Read Download

1. Gil'mullin T.M., Gil'mullin M.F. Bystryj poisk anomalij v chislovyh ryadah pri pomoshchi modificirovannogo metoda Hampelya // Modelirovanie, optimizaciya i informacionnye tekhnologii. 2023. T. 11. № 4. DOI:https://doi.org/10.26102/2310-6018/2023.43.4.030

2. Chirkov A.V. Algoritm formirovaniya obuchayushchih i testovyh vyborok dlya analiza haraktera dannyh // Modelirovanie, optimizaciya i informacionnye tekhnologii. 2024. T. 12. № 4. DOI:https://doi.org/10.26102/2310-6018/2024.47.4.014

3. Obespechenie nadezhnosti sistem pozharnoj signalizacii: problemy i perspektivy / A.A. Snezhko [i dr.] // Sibirskij pozharno-spasatel'nyj vestnik. 2024. № 4 (35). S. 143–149. DOI:https://doi.org/10.34987/vestnik.sibpsa.2024.89.77.016

4. Kumratova A.M., Chumarenko K.E. Prognozirovanie slozhnyh processov nejronnymi setyami // Sovremennaya ekonomika: problemy i resheniya. 2023. № 3 (159). S. 27–36. DOI:https://doi.org/10.17308/meps/2078-9017/2023/3/27-36

5. Primenenie iskusstvennyh nejronnyh LSTM-setej dlya operativnogo analiza akusticheskih, magnitnyh i vibracionnyh polej / V.V. Gravirov [i dr.] // Nauka i tekhnologicheskie razrabotki. 2023. T. 102. № 1. S. 40–64. DOI:https://doi.org/10.21455/std2023.1-3

6. Vliyanie koncentracii dyma i lozhnyh faktorov na srabatyvanie avtomaticheskoj pozharnoj signalizacii / E.F. Rahmatullina [i dr.] // Vestnik Sankt-Peterburgskogo universiteta GPS MCHS Rossii. 2025. № 3. S. 1–12. DOI:https://doi.org/10.61260/2218-130X-2025-3-1-12

7. Prognozirovanie lozhnyh srabatyvanij dymovyh pozharnyh izveshchatelej s pomoshch'yu nejronnoj seti / I.F. Hafizov [i dr.] // Sibirskij pozharno-spasatel'nyj vestnik. 2025. № 2 (37). S. 111–120. DOI:https://doi.org/10.34987/vestnik.sibpsa.2025.17.70.008

8. Kim D., Lee S., Jeong Yu. False alarm suppression in fire alarm systems using ensemble learning of heterogeneous sensors // Fire Safety Journal. 2023. Vol. 138. P. 103798. DOI:https://doi.org/10.1016/j.firesaf.2023.103798

9. Lee S., Park J., Kim H. Development of an intelligent fire detector based on convolutional neural network for discriminating false alarms // IEEE Access. 2022. Vol. 10. P. 54567–54578. DOI:https://doi.org/10.1109/ACCESS.2022.3175968

10. Lee S., Park J., Kim H. Development of an intelligent fire detector based on convolutional neural network for discriminating false alarms // IEEE Access. 2022. Vol. 10. P. 54567–54578. DOI:https://doi.org/10.1109/ACCESS.2022.3175968

11. Terehin M.A., Ivaschenko A.V., Kulakov G.A. Konceptual'nyy podhod k integracii iskusstvennogo intellekta v inzhenernuyu deyatel'nost' // Modelirovanie, optimizaciya i informacionnye tehnologii. 2025. T. 13. № 2. DOI:https://doi.org/10.26102/2310-6018/2025.49.2.031

12. Prichiny neispravnostej sredstv protivopozharnoj zashchity obrazcov vooruzheniya i voennoj tekhniki / A.I. Zajcev [i dr.] // Voprosy oboronnoj tekhniki. Ser. 16: Tekhnicheskie sredstva protivodejstviya terrorizmu. 2022. № 7–8 (157–158). S. 20–29. DOI:https://doi.org/10.53816/23061456_2022_5–6_156

13. Terekhin M.A., Ivashchenko A.V., Kulakov G.A. Konceptual'nyj podhod k integracii iskusstvennogo intellekta v inzhenernuyu deyatel'nost' // Modelirovanie, optimizaciya i informacionnye tekhnologii. 2025. T. 13. № 2. DOI:https://doi.org/10.26102/2310-6018/2025.49.2.031

PDF Text JATS XML

Professor of the Department of Physical and Chemical Fundamentals of Combustion and Extinguishing Processes Saint-Petersburg university of State fire service of EMERCOM of Russia

Olga Lozhkina

Head of the Department of Fire, Rescue Equipment and automotive industry of Saint-Petersburg university of State fire service of EMERCOM of Russia

Maksim Sytdykov

277

downloads

GOST

Rahmatullina E. F., Khafizov I. F., Pavlova Z. K. METHODOLOGY FOR BUILDING A NEURAL NETWORK TO PREDICT FALSE ALARMS FROM SMOKE DETECTORS IN A FIRE ALARM SYSTEM // Problems of risk management in the technosphere. 2026. no. 1. pp. 184-192. DOI: https://doi.org/10.61260/1998-8990-2026-1-184-192. (Date of access 28.09.2026).

MLA

Elina Rahmatullina, Khafizov I. F., Pavlova Z. K.. "METHODOLOGY FOR BUILDING A NEURAL NETWORK TO PREDICT FALSE ALARMS FROM SMOKE DETECTORS IN A FIRE ALARM SYSTEM." Problems of risk management in the technosphere 2026.1 (2026): 184-192. DOI: https://doi.org/10.61260/1998-8990-2026-1-184-192.

APA

Rahmatullina E. F., Khafizov I. F., Pavlova Z. K. (2026). METHODOLOGY FOR BUILDING A NEURAL NETWORK TO PREDICT FALSE ALARMS FROM SMOKE DETECTORS IN A FIRE ALARM SYSTEM. Problems of risk management in the technosphere, 2026(1), 184-192. https://doi.org/10.61260/1998-8990-2026-1-184-192

Chicago

Elina Rahmatullina, Khafizov I. F., Pavlova Z. K.. "METHODOLOGY FOR BUILDING A NEURAL NETWORK TO PREDICT FALSE ALARMS FROM SMOKE DETECTORS IN A FIRE ALARM SYSTEM." Problems of risk management in the technosphere 2026, no. 1 (2026): 184-192. https://doi.org/10.61260/1998-8990-2026-1-184-192.

WIKI

{{cite journal | author = Rahmatullina E. F., Khafizov I. F., Pavlova Z. K. | title = METHODOLOGY FOR BUILDING A NEURAL NETWORK TO PREDICT FALSE ALARMS FROM SMOKE DETECTORS IN A FIRE ALARM SYSTEM | journal = Problems of risk management in the technosphere | year = 2026 | volume = 2026 | issue = 1 | pages = 184-192 | doi = 10.61260/1998-8990-2026-1-184-192 }}

Copy

Export: BibTeX EndNote RefMan RusMARC

METHODOLOGY FOR BUILDING A NEURAL NETWORK TO PREDICT FALSE ALARMS FROM SMOKE DETECTORS IN A FIRE ALARM SYSTEM

Confirmation