INTEGRATION OF FUZZY LOGIC IN SCADA SYSTEMS FOR ENHANCED ALARM MANAGEMENT IN PIPELINE MONITORING

ABSTRACT

This article proposes a fuzzy-logic solution for optimizing alarm management in pipeline-monitoring SCADA systems. As industrial processes become more complex, traditional alarm systems fall prey to alarm flooding—large numbers of signals that overwhelm the operator and compromise safety. To counter this, Mamdani-type fuzzy inference is used within SCADA systems to sort alarms by urgency and context.

The method leverages historic alarm records of simulated pipeline operation. Among the attributes used as a basis for fuzzy rules in identifying critical events are alarm frequency, tag repetition, and time intervals. Python with Jupyter Notebook was used as an implementation tool and prototyping as well as visualization environment.

Testing illustrates that this technique drastically reduces alarm noise while preserving high situational awareness. The result is an understandable, scalable solution that enhances decision-making support in industrial control systems.

АННОТАЦИЯ

В этой статье предлагается решение на основе нечеткой логики для оптимизации управления сигнализацией в SCADA-системах мониторинга трубопроводов. По мере усложнения производственных процессов традиционные системы сигнализации становятся жертвой аварийного переполнения — большого количества сигналов, которые перегружают оператора и ставят под угрозу безопасность. Чтобы противостоять этому, в SCADA-системах используется нечеткий логический вывод типа Мамдани для сортировки сигналов тревоги по срочности и контексту.

В методе используются исторические записи аварийных сигналов о работе моделируемого конвейера. В качестве основы для нечетких правил определения критических событий используются частота аварийных сигналов, повторение тегов и временные интервалы. В качестве инструмента реализации и среды прототипирования, а также среды визуализации использовался Python с Jupyter Notebook.

Тестирование показало, что этот метод значительно снижает уровень шума при тревоге, сохраняя при этом высокую ситуационную осведомленность. В результате получается понятное, масштабируемое решение, которое улучшает поддержку принятия решений в промышленных системах управления.

Keywords:  fuzzy logic, SCADA, alarm flood, Mamdani FIS.

Ключевые слова: нечеткая логика, SCADA, аварийное наводнение, Система нечеткого вывода Мамдани.

Introduction

Enlargement within the pipeline networks and regulation requirements make real-time monitoring and anomaly detection with superior advanced techniques necessary. The workhorses of pipeline operation, SCADA systems, are plagued by long-standing alarm management issues—alarm flooding and low-priority allocation - overloading operators with information and numbing situational awareness.

Standards such as ANSI/ISA-18.2 [2] and EEMUA-191 [3] point out that effective alarm systems are meant to allow for timely operator action in response to abnormal conditions. However, in alarm floods when there are more than 10 alarms for a 10-minute period, operator performance is bad and response delay or miss is more probable. This issue is particularly problematic in pipeline networks where transient and recurring alarms will mask real danger, such as with the 2005 BP Texas City explosion and the 2010 Kalamazoo River oil spill.

This study investigates the use of fuzzy logic - a method greatly adaptable in managing uncertainty—through SCADA-based alarm systems. A Mamdani-type fuzzy inference system was created to identify potentially flooded alarms based on frequency, repetition, and timing. Alarm history data from a simulated compressor station built in Siemens WinCC were utilized for testing the model.

The study has theoretical and practical implications: it progresses the application of fuzzy logic to industrial alarm systems and presents an interpretable as well as scalable approach to smart alarm filtering. The solution is anticipated to provide response time, risk improvement, as well as future potential to be combined with IIoT and predictive analytics.

Current Research Trends and Challenges in Industrial Alarm Management

This subchapter describes the most relevant scientific and technical literature of industrial alarm management systems with special focus on applications relevant to pipeline monitoring environments.

Advanced alarm management systems enhance operational productivity, reliability, and safety in increasingly automated industrial plants. No longer warning devices only, alarms are now intelligent decision-support systems in SCADA and DCS applications. Proper alarm management reduces downtime, improves situational awareness, and maximizes asset utilization [2].

Operator Awareness and Safety: Good alarm systems present only suitable, prioritized alarms to prevent cognitive overload and enable timely operator action. Research by Goel [1] shows that systems with dynamic prioritization and root-cause analysis prevent low-level problems to become significant events.

Process Safety and Compliance: Alarm systems are an essential layer of process safety systems. Guidelines like ANSI/ISA-18.2 [2] and EEMUA 191 [3] integrate alarms with safety lifecycles to facilitate timely fault detection and OSHA PSM and COMAH compliance.

Digital Integration and Monitoring: Modern SCADA/DCS platforms facilitate real-time alarm recording, KPI monitoring, and "bad actor" detection. Applications like alarm historians and dashboards facilitate data-driven optimization through PDCA cycles [1].

Financial Benefits: Poor alarm management will cost the industry over $20 billion annually. Sites with rationalization programs—shelving, auditing, deadband tuning—achieve spectacular uptime and performance improvements.

Foundation for Smart Automation: Smart alarm systems are the foundation for predictive maintenance, machine learning, and IIoT. Alarm metadata is used to support failure prediction and adaptive control, making the shift to smart, cyber-physical systems possible.

These articles cover machine learning and fuzzy logic techniques for enhancing alarm filtering, prioritization, and SCADA system integration.

Bote et al. [4] proposed a four-stage machine learning-based method to alleviate alarm flooding in SCADA systems: data collection, preprocessing, training of models, and alarm suppression. SVM, Random Tree, and OneR classifiers were tested using actual plant logs—SVM achieved 99.89% accuracy. Safety-critical alarms were well-prioritized by the system, reducing operator overload and showing adaptive SCADA integration feasibility.

Al-Fadhli and Zaher [5] built a prototype intelligent SCADA for oil refineries with LabVIEW, incorporating wireless sensor data acquisition, mobile interfaces, logging, and redundancy through twin sensors. A deadband method minimized spurious gas leakage alarms. The alarm handling was not emphasized, but the system offered robust human-oriented and hardware-software integration for field operations.

Pellini and Ribeiro [6] applied fuzzy logic to fault diagnosis in substations according to IEC 61850 protocols. They combined a Cause-Effect Network with a Fuzzy Rule-Based Database within a Programmable Logic Controller (PLC) to reason about GOOSE/MMS messages and categorize faults in real-time. Analog signals were translated to comprehensible linguistic words (e.g., "low," "normal"), making it easier for SCADA operators to understand.

Coffman-Wolph [7] introduced fuzzy sorting for alarm prioritization, in which alarms can be in numerous categories (e.g., "mostly critical") with varying degrees. This is a general concept and not SCADA-specific but may be used in dynamic alarm interfaces, where severity can vary based on context and operator interaction.

Zhang et al. [8] presented a hybrid model that employs expert rules and a CNN for alarm identification in electrical grid SCADA systems. It works on raw alarm data and expert outputs to improve classification accuracy, achieving 94.3% accuracy on 8 million records—better than baseline CNN and SVM models. It presents an interpretable and scalable real-time alarm prioritization solution for unfavorable grid conditions.

Design of SCADA System

To test the fuzzy logic algorithm built for alarm flood detection, a simulated Supervisory Control and Data Acquisition (SCADA) system was implemented based on Siemens WinCC platform. The simulation setup mimics primary sectors of an average Compressor Station (CS) equipment used in gas transport pipelines, namely:

• Gas Purification Unit (GPU) to filter and purify gas.

Figure 1. Gas Purification Unit in WinCC

• Pig Transporting Station (PTS), which enables pipeline cleaning and inspection operations by inserting into the main gas pipeline Pipeline Inspection Gauges (PIG).

Figure 2. PTS in WinCC

The composition of the simulation matches conventional industrial norms described in technical literature and GOST normative publications for compressor stations [9;10]

Methodology

This study developed a Mamdani-type fuzzy inference system to assess the likelihood of alarm flooding in SCADA systems used for pipeline monitoring. The algorithm uses three input features: time between alarms (Time_Delta), alarm frequency (Alarm_Frequency), and repetition of the same tag within a short window (Same_Tag_Repeats_1min).

Each input was modeled using linguistic variables with triangular and trapezoidal membership functions. A rule base was constructed based on expert heuristics to evaluate the fuzzy inputs and output a continuous variable: Fuzzy_Flood_Risk. This output represents the degree to which a given alarm may contribute to an alarm flooding event.

The system was implemented in Python using the scikit-fuzzy library. A dataset consisting of historical alarm logs exported from a simulated pipeline compressor station in Siemens WinCC was used for testing. The model processes the log sequentially, assigning flood risk levels to each alarm event.

The fuzzy system provides a flexible and interpretable framework for distinguishing between normal alarms and those likely to occur in overload conditions, with potential integration into SCADA HMIs for improved operator decision-making.

Results

The fuzzy logic-based classifier successfully quantified alarm flood risk in the SCADA system using three input features. A Mamdani-type fuzzy inference system generated a continuous risk score (Fuzzy_Flood_Risk) between 0 and 1 for 118 alarm events (see Figure 3).

The score distribution ranged from 0.10 to 0.87, with a mean of 0.60 and a standard deviation of 0.30. Risk levels were grouped as Low (≤0.33), Medium (0.34–0.66), and High (>0.66). Results showed 51% of alarms were high risk, 25% medium, and 24% low—indicating a tendency toward alarm clustering.

Several high-risk cases involved repeated alarms from the same tag within short intervals (e.g., 3 seconds), suggesting poor deadband settings or filtering logic. These were common in differential pressure transmitters.

The fuzzy risk scores were added to the original alarm dataset and visualized via a dedicated HMI screen. This enriched output provides more actionable insights than traditional thresholding, supporting future real-time monitoring, rationalization, or suppression strategies.

Figure 3. Alarm Flood Risk Report with Results

Conclusion

This study employed a Mamdani-type fuzzy logic model with three characteristics: Time_Delta, Alarm_Frequency, and Same_Tag_Repeats_1min to detect alarm flooding in pipeline SCADA systems. Using scikit-fuzzy and tested on real CSV alarm logs, the model produced a continuous Fuzzy_Flood_Risk score. 50% of 118 events were flagged as high risk, indicating repetitive or closely spaced frequent alarms—very likely a sign of under-tuned deadbands. As opposed to hard threshold methods, the fuzzy approach produces graded risk estimates that are better suited to operator decision-making and can be integrated with rationalization tools or HMIs. Disadvantages include off-line deployment, lack of expert-labeled optimization, and limited feature set. Real-time integration, rule‐base tuning, and incorporation of additional parameters (e.g., priorities, deadbands, acknowledgment times) will be addressed in upcoming work.

References:

1. Goel P., Datta A., Mannan M.S. Industrial alarm systems: Challenges and opportunities // Journal of Loss Prevention in the Process Industries. – 2017.
2. ANSI/ISA-18.2-2016. Management of alarm systems for the process industries. – International Society of Automation, 2016. – 80 p.
3. Engineering Equipment and Materials Users' Association (EEMUA). Alarm Systems – A Guide to Design, Management and Procurement. – 3rd ed. – London: EEMUA, 2013. – (EEMUA Publication No. 191).
4. Bote A.S., Kshirsagar D., Madkaikar A., Shah B. Intelligent Based Alarm Management System for Plant Automation // Proceedings of the 3rd IEEE International Conference on Recent Trends in Electronics, Information & Communication Technology (RTEICT-2018). – 2018. – P. 1569–1572. – DOI: 10.1109/RTEICT42901.2018.9012661.
5. Al-Fadhli M., Zaher A.A. A Smart SCADA System for Oil Refineries // 2018 IEEE Conference on Instrumentation and Measurement Technology. – 2018.
6. Pellini E.L., Ribeiro L.G. Fuzzy Logic Applied to Registration of Alarms and Events in Substations with IEC 61850 // Proceedings of the IEEE Instrumentation and Measurement Society. – 2016.
7. Coffman-Wolph S. Proof-of-Concept: Creating “Fuzzy” Sorting Algorithms // Proceedings of the Midwest Artificial Intelligence and Cognitive Science Conference. – 2017.
8. Zhang M., Shen P., Zhao Y., Liu Y. Grid Monitoring Alarm Event Recognition Method Integrating Expert Rules and Deep Learning // Proceedings of the 2019 International Conference on Mechanical, Control and Computer Engineering (ICMCCE). – IEEE, 2019. – P. 389–393. – DOI: 10.1109/ICMCCE48743.2019.00093.
9. Электронный ресурс: https://online.zakon.kz/Document/?doc_id=32856903 (дата обращения: 14.04.2025)
10. Электронный ресурс: https://online.zakon.kz/Document/?doc_id=30056588 (дата обращения: 14.04.2025)