IM
I. Makropoulos
info
Please Note
<p>This page displays the records of the person named above and is not linked to a unique person identifier. This record may need to be merged to a profile.</p>
1 records found
1
Heating and cooling represent about half of Europe’s final energy consumption, with industry using a large share and fossil fuels still supplying most heat demand. This makes reliable industrial heat pump operation important for improving efficiency and reducing emissions. However, faults, degradation, sensor issues, and changing operating conditions can reduce system performance and increase the need for manual diagnosis.
This thesis develops a digital-twin-based hybrid fault detection and diagnosis methodology for industrial thermal systems. The Digital Twin part focuses on a component-level empirical model of an air-cooled condenser that estimates expected behaviour using manufacturer data. The fault detection and diagnosis part focuses on an alarm-assisted hybrid framework that interprets existing industrial alarms and warnings using operational measurements, thermodynamic principles, rule-based logic, and Digital Twin models.
To support performance monitoring, a general methodology for developing component-level empirical Digital Twin models was created using manufacturer and operational data. The methodology includes automatic extraction of healthy performance data, identification of dominant operating variables, normalization with respect to design conditions, and correction-based empirical modelling to estimate expected component behaviour. As an application example, the methodology was implemented for an air-cooled condenser and applied to operational data from three plants, where it identified one condenser with clear signs of underperformance.
The fault detection and diagnosis part was developed to reduce the dependence on manual alarm interpretation in industrial operation. For this purpose, an alarm-assisted hybrid framework was implemented using a Branch A / Branch B structure, where Branch A diagnoses existing alarms and warnings and Branch B is proposed as a future extension for hidden degradation detection. The first branch combines operational measurements, thermodynamic reasoning, rule-based diagnostic logic, and Digital Twin models to automatically translate alarms and warnings into structured diagnostic indications. The framework was implemented in Python in a modular form, so that new alarms, plant-specific settings, and additional diagnostic checks can be added without changing the overall structure. The results showed that the framework can provide transparent and physically traceable diagnostic support.
The most important conclusion of this thesis is that the methodology developed for component-level Digital Twin modelling, together with the proposed Branch A / Branch B FDD structure, showed strong potential for supporting monitoring and diagnosis in industrial heat pump systems. Additionally, Branch A provided a practical method for translating alarms and warnings into clear and traceable diagnostic indications. Overall, this methodology could help industrial operators diagnose faults more effectively while reducing the time and resources required for manual alarm interpretation.
...
This thesis develops a digital-twin-based hybrid fault detection and diagnosis methodology for industrial thermal systems. The Digital Twin part focuses on a component-level empirical model of an air-cooled condenser that estimates expected behaviour using manufacturer data. The fault detection and diagnosis part focuses on an alarm-assisted hybrid framework that interprets existing industrial alarms and warnings using operational measurements, thermodynamic principles, rule-based logic, and Digital Twin models.
To support performance monitoring, a general methodology for developing component-level empirical Digital Twin models was created using manufacturer and operational data. The methodology includes automatic extraction of healthy performance data, identification of dominant operating variables, normalization with respect to design conditions, and correction-based empirical modelling to estimate expected component behaviour. As an application example, the methodology was implemented for an air-cooled condenser and applied to operational data from three plants, where it identified one condenser with clear signs of underperformance.
The fault detection and diagnosis part was developed to reduce the dependence on manual alarm interpretation in industrial operation. For this purpose, an alarm-assisted hybrid framework was implemented using a Branch A / Branch B structure, where Branch A diagnoses existing alarms and warnings and Branch B is proposed as a future extension for hidden degradation detection. The first branch combines operational measurements, thermodynamic reasoning, rule-based diagnostic logic, and Digital Twin models to automatically translate alarms and warnings into structured diagnostic indications. The framework was implemented in Python in a modular form, so that new alarms, plant-specific settings, and additional diagnostic checks can be added without changing the overall structure. The results showed that the framework can provide transparent and physically traceable diagnostic support.
The most important conclusion of this thesis is that the methodology developed for component-level Digital Twin modelling, together with the proposed Branch A / Branch B FDD structure, showed strong potential for supporting monitoring and diagnosis in industrial heat pump systems. Additionally, Branch A provided a practical method for translating alarms and warnings into clear and traceable diagnostic indications. Overall, this methodology could help industrial operators diagnose faults more effectively while reducing the time and resources required for manual alarm interpretation.
...
Heating and cooling represent about half of Europe’s final energy consumption, with industry using a large share and fossil fuels still supplying most heat demand. This makes reliable industrial heat pump operation important for improving efficiency and reducing emissions. However, faults, degradation, sensor issues, and changing operating conditions can reduce system performance and increase the need for manual diagnosis.
This thesis develops a digital-twin-based hybrid fault detection and diagnosis methodology for industrial thermal systems. The Digital Twin part focuses on a component-level empirical model of an air-cooled condenser that estimates expected behaviour using manufacturer data. The fault detection and diagnosis part focuses on an alarm-assisted hybrid framework that interprets existing industrial alarms and warnings using operational measurements, thermodynamic principles, rule-based logic, and Digital Twin models.
To support performance monitoring, a general methodology for developing component-level empirical Digital Twin models was created using manufacturer and operational data. The methodology includes automatic extraction of healthy performance data, identification of dominant operating variables, normalization with respect to design conditions, and correction-based empirical modelling to estimate expected component behaviour. As an application example, the methodology was implemented for an air-cooled condenser and applied to operational data from three plants, where it identified one condenser with clear signs of underperformance.
The fault detection and diagnosis part was developed to reduce the dependence on manual alarm interpretation in industrial operation. For this purpose, an alarm-assisted hybrid framework was implemented using a Branch A / Branch B structure, where Branch A diagnoses existing alarms and warnings and Branch B is proposed as a future extension for hidden degradation detection. The first branch combines operational measurements, thermodynamic reasoning, rule-based diagnostic logic, and Digital Twin models to automatically translate alarms and warnings into structured diagnostic indications. The framework was implemented in Python in a modular form, so that new alarms, plant-specific settings, and additional diagnostic checks can be added without changing the overall structure. The results showed that the framework can provide transparent and physically traceable diagnostic support.
The most important conclusion of this thesis is that the methodology developed for component-level Digital Twin modelling, together with the proposed Branch A / Branch B FDD structure, showed strong potential for supporting monitoring and diagnosis in industrial heat pump systems. Additionally, Branch A provided a practical method for translating alarms and warnings into clear and traceable diagnostic indications. Overall, this methodology could help industrial operators diagnose faults more effectively while reducing the time and resources required for manual alarm interpretation.
This thesis develops a digital-twin-based hybrid fault detection and diagnosis methodology for industrial thermal systems. The Digital Twin part focuses on a component-level empirical model of an air-cooled condenser that estimates expected behaviour using manufacturer data. The fault detection and diagnosis part focuses on an alarm-assisted hybrid framework that interprets existing industrial alarms and warnings using operational measurements, thermodynamic principles, rule-based logic, and Digital Twin models.
To support performance monitoring, a general methodology for developing component-level empirical Digital Twin models was created using manufacturer and operational data. The methodology includes automatic extraction of healthy performance data, identification of dominant operating variables, normalization with respect to design conditions, and correction-based empirical modelling to estimate expected component behaviour. As an application example, the methodology was implemented for an air-cooled condenser and applied to operational data from three plants, where it identified one condenser with clear signs of underperformance.
The fault detection and diagnosis part was developed to reduce the dependence on manual alarm interpretation in industrial operation. For this purpose, an alarm-assisted hybrid framework was implemented using a Branch A / Branch B structure, where Branch A diagnoses existing alarms and warnings and Branch B is proposed as a future extension for hidden degradation detection. The first branch combines operational measurements, thermodynamic reasoning, rule-based diagnostic logic, and Digital Twin models to automatically translate alarms and warnings into structured diagnostic indications. The framework was implemented in Python in a modular form, so that new alarms, plant-specific settings, and additional diagnostic checks can be added without changing the overall structure. The results showed that the framework can provide transparent and physically traceable diagnostic support.
The most important conclusion of this thesis is that the methodology developed for component-level Digital Twin modelling, together with the proposed Branch A / Branch B FDD structure, showed strong potential for supporting monitoring and diagnosis in industrial heat pump systems. Additionally, Branch A provided a practical method for translating alarms and warnings into clear and traceable diagnostic indications. Overall, this methodology could help industrial operators diagnose faults more effectively while reducing the time and resources required for manual alarm interpretation.