EP4639511A1

EP4639511A1 - Automatic parameterset selection

Info

Publication number: EP4639511A1
Application number: EP23820766.6A
Authority: EP
Inventors: Stephanie Brandt; Ruud Geelen; Ulrich Kuhn-Matysiak; Martin Zimmermann
Original assignee: Siemens Schweiz AG
Current assignee: Siemens Schweiz AG
Priority date: 2022-12-21
Filing date: 2023-11-24
Publication date: 2025-10-29
Also published as: WO2024132375A1; CN120418843A; EP4390892A1

Abstract

A method and an arrangement for automatically providing suitable parameter sets for a hazard detection system (e.g. fire alarm system) to improve the detection performance of the detection system to increase the level of safety within the building.

Description

202212321 1 AUTOMATIC PARAMETERSET SELECTION FIELD OF THE INVENTION The present invention relates to a method and an arrangement for providing parameter sets for a hazard detection system to improve the detection performance of the detection system. BACKGROUND Almost all households or buildings are nowadays equipped with hazard detectors, especially smoke detectors. It happens again and again that hazard detectors, especially smoke detectors set off alarms without there actually being a hazard, like a fire. In the case of a false alarm, e.g. a smoke detector triggers a fire alarm even there is no fire. If a hazard de- tector is too sensitive, false alarms could be caused. On the other hand, if a hazard detector is too robust, false alarms are avoided, but a lower level or fire detection safety or protection is provided. This is annoying and can be expensive in case that the fire brigade is notified. Reasons for false alarms can be faulty measurement technology of the detector, or incorrect or not optimal settings of the detector. SUMMARY OF THE INVENTION The object of the invention is to provide optimal parameter settings for the detector to reduce the number of false alarms of a hazard detector at maximum possible sensitivity. 202212321 2 A first aspect of the invention a method for providing pa- rameter sets for a hazard detection system, comprising a haz- ard control panel connected to hazard detectors, especially fire detectors, to improve the detection performance of the detection system, the method comprising: providing measured values of at least one of the detec- tors representing the operation status in case no hazard is detected (“normal status”) by the respective detector; providing measured values of at least one of the detec- tors representing events (“different from normal status”) in case a hazard (“real hazard”) or a potential hazard (“false alarm”, “No hazard”) is detected by the respective detector; annotating (e.g. tagging) the measured values represent- ing an event in case of occurrence of a real hazard and in case of a potential hazard; analyzing the measured values for the case no hazard is detected and the annotated measured values representing events in case a hazard or a potential hazard is detected to identify a parameter set for the hazard detection system to reduce the number of false alarms and/or to minimize the detection time without increasing the number of false alarms. A second aspect of the invention is an arrangement for providing parameter sets for a hazard detection system to im- prove the detection performance of the detection system, the arrangement comprising: a hazard control panel connected to hazard detectors; an analysis engine for analyzing measured values of at least one of the detectors; wherein at least one of the hazard detectors is config- ured to provide measured values representing the operation status (“normal status”) in case no hazard is detected to the analysis engine; 202212321 3 wherein at least one of hazard detectors is config- ured to provide measured values in case a hazard (“real haz- ard”) or a potential hazard (“false alarm”, “No hazard”) is detected to the analysis engine; means for annotating (e.g. tagging) the measured values representing an event in case of occurrence of a real hazard and in case of a potential hazard; wherein the analysis engine is configured to analyze the measured values for the case no hazard is detected and the measured values representing events in case a real hazard or a potential hazard is detected and the annotated measured values representing events in case a hazard or a potential hazard is detected to identify a parameter set for the hazard detection system to reduce the number of false alarms and/or to minimize the detection time without increasing the number of false alarms. A further aspect of the invention is a data processing system comprising instructions or means to carry out the steps of the inventive method. BRIEF DESCRIPTION OF THE DRAWINGS The above mentioned and other concepts of the present inven- tion will now be addressed with reference to the drawings of the preferred embodiments of the present invention. The shown embodiments are intended to illustrate, but not to limit the invention. The drawings contain the following figures, in which like numbers refer to like parts throughout the descrip- tion and drawings and wherein: 202212321 4 FIG 1 illustrates a first arrangement for providing parameter sets for a hazard detection sys- tem, FIG 2 illustrates a second exemplary arrangement for providing parameter sets for a hazard detection sys- tem, FIG 3 illustrates a third exemplary arrangement for providing parameter sets for a hazard detection sys- tem, FIG 4 illustrates an exemplary flowchart of a method for providing parameter sets for a hazard detection sys- tem, FIG 5 illustrates exemplary time series of exemplary meas- ured values, and FIG 6 illustrates three exemplary types of danger signal patterns. DETAILED DESCRIPTION False alarms (e.g. false fire alarms) in a building are annoy- ing for the occupants and can be expensive in case that the fire brigade is notified. Reasons for false alarms can be faulty measurement technology of the detector (e.g. improperly working sensor or detector), or incorrect or not optimal sen- sitivity settings of the detector. In most of the cases when the customer is faced with a couple of false alarms the most robust parameter sets are applied but not an appropriate parameter set which could be as sensitive as possible in this environment. 202212321 5 Today the hazard detection are often delivered with a medium sensitive parameter set. Very often this is not the most sensitive parameter set which would be possible in this specific environment to provide optimal protection. One aspect of the invention is to provide a process of auto- mated selecting an optimal parameter set from a couple of al- ready approved and/or certified parameter sets for hazard de- tectors and/or for hazard control panels. Figure 1 illustrates a first exemplary arrangement for auto- matically providing parameter sets for a hazard detection sys- tem HDS1 to improve the detection performance of the detection system. The arrangement according to figure 1 comprises: a hazard control panel HCP1 connected to hazard detectors HD1 – HD3 via a detector line DL; an analysis engine AE for analyzing measured values MV_NS, MV_RH, MV_PH of at least one of the detectors HD1 – HD3; wherein at least one of the hazard detectors HD1 – HD3 is configured to provide measured values MV_NS representing the operation status (“normal status”) in case no hazard is de- tected to the analysis engine AE; wherein at least one of the hazard detectors HD1 – HD3 is configured to provide measured values MV_RH, MV_PH in case a hazard (“real hazard”) or a potential hazard (“false alarm”, “No hazard”) is detected to the analysis engine AE; means TM1 for annotating or tagging the measured values MV_RH, MV_PH representing an event in case of occurrence of a real hazard and in case of a potential hazard; wherein the analysis engine AE is configured to analyze the measured values MV_NS for the case no hazard is detected and the annotated measured values MV_RHT, MV_PHT representing events in case a hazard or a potential hazard is detected to 202212321 6 identify a parameter set IPS the hazard detection system to reduce the number of false alarms and/or to minimize the detection time without increasing the number of false alarms. Advantageously the analysis engine AE is determining the opti- mal parameter setting for the detector to achieve the highest level of fire detection safety/protection (=highest sensitiv- ity) without causing risks for false alarms or for too many false alarms. A parameter set or a parameter setting is for instance a named collection of numbers or attributes describing the configura- tion of the hazard detection algorithm in the detection sys- tem. It may for instance comprise filter parameters, weights, and/or thresholds for conditional behaviors. Advantageously also historic measured values MV_RH, MV_PH rep- resenting an event in case of occurrence of a real hazard and in case of a potential hazard will be annotated and tagged to enlarge the set of data for further analysis. Advantageously the analysis determines settings of the hazard detector to reduce the number of false alarms of a hazard de- tector at maximum possible sensitivity of the detector. In the illustration according to figure 1 the means TM1 for annotating or tagging the measured values MV_RH, MV_PH repre- senting an event in case of occurrence of a real hazard and in case of a potential hazard are implemented by a tagging mecha- nism TM1 which is integrated in the hazard control panel HCP1. The hazard control panel HCP1 receives via the detector line DL the following measured values from at least one of the haz- ard detectors HD1 - HD3: MV_NS Measured Values Normal Status 202212321 7 MV_RH Measured Real Hazard MV_PH Measured Values Potential Hazard. Advantageously the measured values MV_NS, MV_RH, MV_PH com- prise in each case information (e.g. metadata) which enables to assign these measured values to the respective hazard de- tector HD1 – HD3. By using the tagging mechanism TM1 a person B (e.g. facility manager, service person) is tagging or annotating the measured values MV_RH (Measured Values Real Hazard) and MV_PH (Measured Values Potential Hazard) in case if a real hazard (e.g. fire, smoke) is present. The tagging or annotating the measured val- ues MV_RH (Measured Values Real Hazard) and MV_PH (Measured Values Potential Hazard) is confirming that a real hazard is present. Advantageously the measured values MV_NS (Measured Values Nor- mal Status) are tagged or annotated accordingly that the nor- mal status or situation is present, and no hazard is detected. The tagging or annotating of the measured values MV_NS (Meas- ured Values Normal Status) can be performed manually by the person B or automatically by the tagging mechanism TM1. The tagging (or annotating) mechanism TM1 can comprise input means (e.g. input panel, touch screen) for manual tagging. Ad- vantageously the tagging mechanism TM1 is performing the tag- ging or annotating automatically. By comparing and evaluating received measured values MV_NS, MV_RH, MV_PHfrom more than one of the hazard detectors HD1 – HD3. Advantageously evaluat- ing and tagging the received measured values MV_NS, MV_RH, MV_PH also comprises inputs from other sources like manual call points (MCP). 202212321 8 Via an appropriate connection CC1 (e.g. radio connection, Internet) the hazard control panel HCP1 is in data connection with the analytics engine AE. The hazard control panel HCP1 is sending the following data to the analytics engine AE: MV_NS Measured Values Normal Status MV_NST Measured Values Normal Status Tagged MV_RHT Measured Values Real Hazard Tagged MV_PHT Measured Values Potential Hazard Tagged. Advantageously the measured values MV_NS, respectively the tagged measured values MV_NST, MV_RHT, MV_PHT comprise in each case information (e.g. metadata) which enables to assign the respective measured values, respectively the tagged measured values to the respective hazard detector HD1 – HD3. The analytics engine AE is hosted by a server S. The server S comprises appropriate communication means, processing means, I/O means, and memory means. The server S has access to a suitable database DB. The database DB comprises BIM-data (BIM: Building Information Model) of the hazard detection system and the building where the hazard detection system is installed. The database DB further comprises a set of parameter sets (e.g. configurations or configuration data) SPS for the hazard detection system. The parameter sets comprising configuration data for the hazard control panel HCP1 and for the respective detectors HD1 – HD3. Advantageously all parameter sets are certified and comply with the respective standard. The analytics engine AE comprises appropriate simulation means and/or analytics means (e.g. machine learning mechanism, rule based reasoning mechanisms, and/or simulation mechanisms) to analyze the received data MV_NS, MV_NST, MV_RHT, MV_PHT to 202212321 9 identify an improved parameter IPS for the hazard detec- tion system which reduces the number of false alarms or mini- mizes the detection time of a hazard situation (e.g. fire, smoke, gas) without increasing the number of false alarms. Based on the received data the analytics engine AE is select- ing the most suitable parameter set from the set of parameter sets (e.g. configurations or configuration data) SPS which re- duces the number of false alarms or minimizes the detection time of a hazard situation (e.g. fire, smoke, gas) without in- creasing the number of false alarms. Advantageously the analytics engine AE and the Building Infor- mation Model BIM are representing a digital twin of the hazard detection system. Based on the received data the analytics en- gine AE can run simulations on the digital twin to identify an improved parameter set IPS. The improved parameter set IPS is sent via the communication connection CC1 to the hazard control panel HCP1. The hazard control panel HCP1 transmits the respective improved parameter set IPS to the respective hazard detector HD1 – HD3. The im- proved parameter set IPS can comprise configuration data for the hazard control panel HCP1 and/or the respective hazard de- tector HD1 – HD3. Advantageously the configuration for HCP1 and HD1-HD3 may also comprise multi-detector dependency, eg. HCP1 considers MV_PH received simultaneously from both HD1 and HD2 located in the same room as a real hazard. Advantageously the analytics means of the analytics engine AE are implemented by appropriate software programs running on the server S. Advantageously the server S and the database DB are implemented or hosted in Cloud infrastructure C. 202212321 10 In principle the analytics engine AE and the database DB can be hosted or implemented on the hazard control panel HCP1 if the hazard control panel HCP1 comprises appropriate processing and memory power. Exemplary advantages of the invention: - Finding optimal parameter set to decrease the number of false alarms and to increase the level of safety in the building. - Improvement of the detection performance of the hazard detection system: making the detectors more sensitive or less sensitive. Advantageously identifying an improved parameter set IPS for the hazard detection system is not a one-time job. Advanta- geously this will be performed frequently or periodically (e.g. by continuously performing a suitable simulation). Optionally identifying an improved parameter set IPS for the hazard detection system is performed on demand. For instance, if there are changes or modifications in the building (e.g. new rooms arrangements, open office spaces in the building, new air flows in the building). Identifying an improved parameter set IPS for a hazard detec- tion system can depend on the criticality of the building where the hazard detection system is implemented. If the building is part of critical infrastructure (e.g. airport, train station) identifying an improved parameter set IPS can be performed for instance daily, weekly or monthly. For highly critical infrastructure it makes sense to identify an improved parameter set IPS for a hazard detection system every hour. 202212321 11 For a new building or for a installed hazard detection system identifying an improved parameter set IPS for a hazard detection system should be performed in the beginning more frequently. Later in the lifetime of the building weekly, monthly or if the usage of the building has changed. E.g. af- ter rebuilding the building, changes on the building. Advantageously the analysis engine AE is configured to analyze the measured values to identify the best fitting parameter set IPS for the hazard detection system to reduce the number of false alarms by simulating a digital twin of the hazard detec- tion system and/or by simulating a digital twin of the hazard detectors. Advantageously the analysis engine AE is a rule-based analytic engine configured to run scenarios of existing parameter sets to identify the parameter set able to reduce the number of false alarms of a hazard detector at maximum possible sensi- tivity. Advantageously the analysis engine AE is implemented in a cloud infrastructure C. Advantageously the hazard detectors HD1 – HD3 are fire detec- tors or smoke detectors or gas detectors or heat detectors or presence detectors. One or more hazard detectors HD1 – HD3 can also be embodiments of a multi-sensor or a multi-criteria de- tector comprising a combination of means for smoke and/or gas and/or heat and/or flame detection. Figure 2 illustrates a second exemplary arrangement for auto- matically providing parameter sets for a hazard detection sys- tem HDS2 to improve the detection performance of the detection 202212321 12 system. In the hazard detection HDS2 according to fig- ure 2 the hazard control panel HCP2 does not communicate di- rectly with the server S and the analytics engine AE. In the hazard detection system HDS2 according to figure 2 the hazard control panel HCP2 communicates to the server S and the analytics engine AE via an edge gateway EG. The edge gateway EG is part of a building network. The communication of the network-nodes (e.g. devices connected by the building network) of the building network to the cloud C is routed via the edge gateway EG. The hazard control panel HCP2 is a device (node) of the building network. In the arrangement of the hazard detection system HDS2 accord- ing to figure 2 the hazard control panel HCP2 receives via the detector line DL the following measured values from at least one of the hazard detectors HD1 - HD3: MV_NS Measured Values Normal Status MV_RH Measured Values Real Hazard MV_PH Measured Values Potential Hazard. Advantageously the measured values MV_NS, MV_RH, MV_PH com- prise in each case information (e.g. metadata) which enables to assign these measured values to the respective hazard de- tector HD1 – HD3. By using the tagging mechanism TM2 a person B (e.g. facility manager, service person) is tagging or annotating the measured values MV_RH (Measured Values Real Hazard) and MV_PH (Measured Values Potential Hazard) in case if a real hazard (e.g. fire, smoke) is present. The tagging or annotating the measured val- ues MV_RH (Measured Values Real Hazard) and MV_PH (Measured Values Potential Hazard) is confirming that a real hazard is present. 202212321 13 Advantageously the measured values MV_NS (Measured Values Nor- mal Status) are tagged or annotated accordingly that the nor- mal status or situation is present, and no hazard is detected. The tagging or annotating of the measured values MV_NS (Meas- ured Values Normal Status) can be performed manually by the person B or automatically by the tagging mechanism TM2. For instance, the tagging mechanism TM2 comprises input means (e.g. input panel, touch screen) for manual tagging and veri- fication of the respective hazard event. Advantageously the tagging mechanism TM2 is performing the tagging or annotating automatically by comparing and evaluating received measured values MV_NS, MV_RH, MV_PH from more than one of the hazard detectors HD1 – HD3. Advantageously evaluating and tagging the received measured values MV_NS, MV_RH, MV_PH also comprises inputs from other sources like manual call points (MCP). Optionally the tagging can be performed remotely by using suitable communication mechanisms. Via an appropriate communication connection CC2 (e.g. radio connection, Internet) the hazard control panel HCP2 is sending the following data to the edge gateway EG: MV_NS Measured Values Normal Status MV_NST Measured Values Normal Status Tagged MV_RHT Measured Values Real Hazard Tagged MV_PHT Measured Values Potential Hazard Tagged. Advantageously the measured values MV_NS, respectively the tagged measured values MV_NST, MV_RHT, MV_PHT comprise in each case information (e.g. metadata) which enables to assign the respective measured values, respectively the tagged measured values to the respective hazard detector HD1 – HD3. 202212321 14 Since the hazard control panel HCP2 and the edge gateway EG are nodes or subscribers of a building network the communica- tion connection CC2 depends on the communication protocol of the building network (e.g. BACnet, BACnet IP, IP protocol). The edge gateway EG is sending these data via an appropriate communication connection CC3 (e.g. radio connection, Internet) to the analytics engine AE which is hosted by a server S. The server S has access to a suitable database DB. The data- base DB comprises BIM-data (BIM: Building Information Model) of the hazard detection system and the building where the haz- ard detection system is installed. The database DB further comprises a set of parameter sets (e.g. configurations or con- figuration data) SPS for the hazard detection system. The pa- rameter sets comprising configuration data for the hazard con- trol panel HCP and for the respective detectors HD1 – HD3. Ad- vantageously all parameter sets are certified and comply with the respective standard. The analytics engine AE comprises appropriate simulation and/or analytics means (e.g. machine learning mechanism, rule based reasoning mechanisms, and/or simulation mechanisms) to analyze the received data MV_NS, MV_NST, MV_RHT, MV_PHT to identify an improved parameter set IPS for the hazard detec- tion system which reduces the number of false alarms or mini- mizes the detection time of a hazard situation (e.g. fire, smoke, gas) without increasing the number of false alarms. Advantageously the analytics engine AE determines settings of the hazard detector to reduce the number of false alarms of a hazard detector at maximum possible sensitivity of the detec- tor. 202212321 15 Advantageously the analysis engine AE is determining the opti- mal parameter setting for the detector to achieve the highest level of fire detection safety/protection (=highest sensitiv- ity) without causing risks for false alarms or for too many false alarms. Based on the received data the analytics engine AE is select- ing the most suitable parameter set from the set of parameter sets (e.g. configurations or configuration data) SPS which re- duces the number of false alarms or minimizes the detection time of a hazard situation (e.g. fire, smoke, gas) without in- creasing the number of false alarms. Advantageously the analytics engine AE and the Building Infor- mation Model BIM are representing a digital twin of the hazard detection system. Based on the received data the analytics en- gine AE can run simulations on the digital twin to identify an improved parameter set IPS. The improved parameter set IPS is sent via the communication connection CC3 to the edge gateway EG. The edge gateway EG transmits the improved parameter set IPS to the hazard control panel HCP2. The hazard control panel HCP2 transmits the re- spective improved parameter set IPS to the respective hazard detector HD1 – HD3. The improved parameter set IPS can com- prise configuration data for the hazard control panel HCP and/or the respective hazard detector HD1 – HD3. Advantageously the analytics means of the analytics engine AE are implemented by appropriate software programs running on the server S. Advantageously the server S and the database DB are implemented or hosted in Cloud infrastructure C. 202212321 16 Advantageously the analysis AE is configured to analyze the measured values to identify the best fitting parameter set IPS for the hazard detection system to reduce the number of false alarms by simulating a digital twin of the hazard detec- tion system and/or by simulating a digital twin of the hazard detectors. A criterion for the best fitting parameter set can be the num- ber of changes which have to be performed in the currently ex- isting setting. Advantageously the analysis engine AE is a rule-based analytic engine configured to run scenarios of existing parameter sets to identify a parameter set able to reduce the number of false alarms and/or to minimize the detection time without increas- ing the number of false alarms. Optionally the analytics engine AE and the database DB can be hosted or implemented on the hazard control panel HCP2 if the hazard control panel HCP comprises appropriate processing and memory power. Optionally the analytics engine AE and the database DB can be hosted or implemented on the edge gateway EG if the edge gate- way EG has appropriate processing and memory power. Figure 3 illustrates a third exemplary arrangement for auto- matically providing parameter sets for a hazard detection sys- tem HDS3 to improve the detection performance of the detection system. In the hazard detection system HDS3 according to fig- ure 3 the hazard control panel HCP3 does not communicate with the server S and the analytics engine AE for automatically providing parameter sets for a hazard detection system HDS3 to improve the detection performance of the detection system. 202212321 17 In the exemplary hazard detection system HDS3 according to figure 3 the hazard detectors HD1 - HD3 are configured to com- municate to the server S and the analytics engine AE via an edge gateway EG. The edge gateway EG is part of a building network. The communication of the network-nodes (e.g. devices connected by the building network) of the building network to the cloud C is routed via the edge gateway EG. hazard detec- tors HD1 - HD3 are devices (nodes) of the building network. In the arrangement of the hazard detection system HDS3 accord- ing to figure 3 the edge gateway EG receives via the respec- tive communication connection CC4, CC5 the following measured values from at least one of the respective hazard detectors HD1 - HD3: MV_NS Measured Values Normal Status MV_RH Measured Values Real Hazard MV_PH Measured Values Potential Hazard. Advantageously the measured values MV_NS, MV_RH, MV_PH com- prise in each case information (e.g. metadata) which enables to assign these measured values to the respective hazard de- tector HD1 – HD3. In the arrangement of the hazard detection system HDS3 accord- ing to figure 3 the edge gateway EG comprises a tagging mecha- nism TM3 for tagging or annotating the measured values MV_RH (Measured Values Real Hazard) and MV_PH (Measured Values Po- tential Hazard) in case if a real hazard (e.g. fire, smoke) is present. The tagging or annotating the measured values MV_RH (Measured Values Real Hazard) and MV_PH (Measured Values Po- tential Hazard) is confirming that a real hazard is present. 202212321 18 Advantageously the measured MV_NS (Measured Values Nor- mal Status) are tagged or annotated accordingly that the nor- mal status or situation is present, and no hazard is detected. Advantageously the tagging mechanism TM3 is performing the tagging or annotating automatically. By comparing and evaluat- ing received measured values MV_NS, MV_RH, MV_PH from more than one of the hazard detectors HD1 – HD3. Advantageously evaluating and tagging the received measured values MV_NS, MV_RH, MV_PH also comprises inputs from other sources (e.g. from a building management station). Automatically evaluating and tagging the received measured values MV_NS, MV_RH, MV_PH can also be performed based on hazard event messages HEM from manual call points MCP. In the arrangement of the hazard de- tection system HDS3 according to figure 3 the edge gateway EG is configured to receive hazard event messages HEM from manual call points MCP via an appropriate communication connection CC7. Via an appropriate communication connection CC6 (e.g. radio connection, Internet) the edge gateway EG is sending the fol- lowing data to the analytics engine AE which is hosted by a server S: MV_NS Measured Values Normal Status MV_NST Measured Values Normal Status Tagged MV_RHT Measured Values Real Hazard Tagged MV_PHT Measured Values Potential Hazard Tagged. Advantageously the measured values MV_NS, respectively the tagged measured values MV_NST, MV_RHT, MV_PHT comprise in each case information (e.g. metadata) which enables to assign the respective measured values, respectively the tagged measured values to the respective hazard detector HD1 – HD3. 202212321 19 Since the hazard detector HD1 – and the edge gateway EG are nodes or subscribers of a building network the communica- tion connections CC4, CC5 depend on the communication protocol of the building network (e.g. BACnet, BACnet IP, IP protocol). The server S has access to a suitable database DB. The data- base DB comprises BIM-data (BIM: Building Information Model) of the hazard detection system and the building where the haz- ard detection system is installed. The database DB further comprises a set of parameter sets (e.g. configurations or con- figuration data) SPS for the hazard detection system. The pa- rameter sets comprising configuration data for the hazard con- trol panel HCP and for the respective detectors HD1 – HD3. Ad- vantageously all parameter sets are certified and comply with the respective standard. The analytics engine AE comprises appropriate analytics means (e.g. machine learning mechanism, rule based reasoning mecha- nisms, and/or simulation mechanisms) to analyze the received data MV_NS, MV_NST, MV_RHT, MV_PHT to identify an improved parameter set IPS for the hazard detection system which re- duces the number of false alarms or minimizes the detection time of a hazard situation (e.g. fire, smoke, gas) without in- creasing the number of false alarms. Based on the received data the analytics engine AE is selecting the most suitable parameter set from the set of parameter sets (e.g. configura- tions or configuration data) SPS which reduces the number of false alarms or minimizes the detection time of a hazard situ- ation (e.g. fire, smoke, gas) without increasing the number of false alarms. Advantageously the analytics engine AE and the Building Infor- mation Model BIM are representing a digital twin of the hazard 202212321 20 detection system. Based on the data the analytics en- gine AE can run simulations on the digital twin to identify an improved parameter set IPS. The improved parameter set IPS is sent via the communication connection CC6 to the edge gateway EG. The edge gateway EG transmits the respective improved parameter set IPS to the re- spective hazard detectors HD1 – HD3. In case an improved pa- rameter set IPS for the hazard control panel HCP3 is provided by the analytics engine AE, said improved parameter set IPS will be transmitted via the detector line DL to the hazard control panel HCP3. Advantageously the analytics means of the analytics engine AE are implemented by appropriate software programs running on the server S. Advantageously the server S and the database DB are implemented or hosted in Cloud infrastructure C. Advantageously the analysis engine AE is configured to analyze the measured values to identify the best fitting parameter set IPS for the hazard detection system to reduce the number of false alarms by simulating a digital twin of the hazard detec- tion system and/or by simulating a digital twin of the hazard detectors. Advantageously the analysis engine AE is a rule-based analytic engine configured to run scenarios of existing parameter sets to identify the parameter set able to reduce the number of false alarms and/or to minimize the detection time without in- creasing the number of false alarms. Optionally the analytics engine AE and the database DB can be hosted or implemented on the edge gateway EG if the edge gate- way EG has appropriate processing and memory power. 202212321 21 Figure 4 illustrates an exemplary flowchart of a method for providing parameter sets for a hazard detection system (HDS1 – HDS3), comprising a hazard control panel connected to hazard detectors, especially fire detectors, to improve the detection performance of the detection system. The method comprising: (S1) providing measured values (MV_NS) of at least one of the detectors representing the operation status in case no hazard is detected (“normal status”) by the respective detec- tor; (S2) providing measured values (MV_RH, MV_PH) of at least one of the detectors representing events (“different from nor- mal status”) in case a hazard (“real hazard”) or a potential hazard (“false alarm”, “No hazard”) is detected by the respec- tive detector; (S3) annotating (e.g. tagging) the measured values (MV_RH, MV_PH) representing an event in case of occurrence of a real hazard and in case of a potential hazard; (S4) analyzing the measured values (MV_NS) for the case no hazard is detected and the annotated measured values (MV_RHT, MV_PHT) representing events in case a hazard or a po- tential hazard is detected to identify a parameter set for the hazard detection system to reduce the number of false alarms and/or to minimize the detection time without increasing the number of false alarms. Annotating or tagging increases the robustness of the hazard detectors and of the hazard detection system in the long term. The ability to build up a database of tagged or annotated measured values (MV_NST, MV_RHT, MV_PHT) for continuous devel- opment and application of improved parameter sets based on 202212321 22 high quality field data new opportunities for identi- fying and tackling important field challenges as well as for monitor success of implementation. Advantageously the database of tagged or annotated measured values (MV_NST, MV_RHT, MV_PHT) is filled automatically by the server S. An aspect of the invention is to identify and to select a pa- rameter set from a set of existing parameters to determine a new parameter set which is better than others. The use and in- tegration of better parameter sets means faster and more reli- able detection of hazard situations. Advantageous embodiments are that also the measured values (MV_NS) representing the operation status in case no hazard is detected (“normal status”) are annotated to indicate the case no hazard is detected or the case a real hazard is detected; and that said annotated measured values (MV_NST) are used in the analyzing step to identify a parameter set for the hazard detection system to reduce the number of false alarms and/or to minimize the detection time without increasing the number of false alarms. Advantageously analyzing the measured values to identify a pa- rameter set for the hazard detection system to reduce the num- ber of false alarms and/or to minimize the detection time without increasing the number of false alarms is performed by simulating a digital twin of hazard detection system and/or by simulating a digital twin of the hazard detectors. Advantageously identifying the parameter set to reduce the number of false alarms and/or to minimize the detection time without increasing the number of false alarms is performed by the respective digital twin by evaluating existing parameter sets to select the best suitable parameter set. 202212321 23 Advantageously identifying the parameter set to reduce the number of false alarms and/or to minimize the detection time without increasing the number of false alarms is performed by a rule-based analytic engine running scenarios of existing pa- rameter sets. The hazard detectors can be fire detectors or smoke detectors or gas detectors or presence detectors. Advantageously the measured values (MV_RH, MV_PH) representing events (“different from normal status”) in case of a real haz- ard (“real hazard”) or a potential hazard (“false alarm”, “No hazard”) are provided based on threshold triggering. Advantageously annotating or tagging the measured values (MV_RH, MV_PH) representing an event in case of occurrence of a real hazard and in case of a potential hazard is performed by a user by entering a respective acknowledgment on the haz- ard control panel. Optionally annotating (e.g. tagging) the measured values (MV_RH) representing an event in case of occurrence of a real hazard is performed by automatic approval of a second hazard detection system and/or by analyzing hazard events (HE) from manual call points (MCP). Optionally annotating (e.g. tagging) the measured values (MV_RH) representing an event in case of occurrence of a real hazard is performed when a defined number of further hazard detectors are also providing measured values (MV_RH) repre- senting an event in case of occurrence of a real hazard. 202212321 24 The method can be realized by configured compo- nents (hazard detectors, detector line, panel, server, commu- nication means), equipped with appropriate computing power, memory space, and software. Advantageously the server is im- plemented in a cloud infrastructure. Roughly explained, an event is a defined status of the detec- tor, which is different from “normal status”. For example, in an average office environment, the fluctuations of raw data are in the range of plus / minus 1 digit and so more or less boring and without value. Events are basically periods of sig- nals related to an upcoming false alarm or real alarm or a status, which is slightly different from “normal status”. Im- plementing such a procedure makes it obsolete to transmit huge amounts of data without any value and to focus only on the in- teresting periods of detector status. Figure 5 illustrates exemplary time series of exemplary meas- ured values or signals. The exemplary measured values or sig- nals (e.g. life time long signals) are illustrated in a time series chart TSC. The X-axis of the time series chart TSC rep- resents the time t. The Y-axis of the time series chart TSC represents the danger counts DC. In figure 5 the embedded small chart TSNS represents a period of time for a time series of signals representing “normal status”. The embedded small chart TSDL3E represents a period of time collecting “events” (e.g. danger level 3 events), in this case exemplary danger level 3 events DL3E1 to DL3E4. The right side of the time se- ries chart TSC represents a legend for the chart TSC for fur- ther information illustrated in the chart TSC: - Danger Min to Max (DMintoMax), - Danger 5% to 95% percentile (D5to95), - Danger Median (DM) - Danger Level 3 (DL3). 202212321 25 If the relevant data records from one or more detectors are available to be analyzed (e.g. in a cloud server), these data records can be used to improve the detector performance in two directions (increase the sensitivity or decrease the sensitiv- ity) to adapt the detectors sensitivity by changing the param- eter set. Case 1: Increase the sensitivity Increase the sensitivity in the sense of "minimizing the time until a real or potential hazard is identified". Such an im- provement needs several data records from “normal status”. Such data records should be recorded at certain times (e.g. specific times during the day or during the night, when the room is populated or without any person inside or to be done with the customer based on specific situations). Also, the length of these normal status records needs to be defined. Some statistical insights or experience could be helpful to define the point of time and the length for the record. Pro- posed here is one minute to one hour, preferably five minutes. Limitation to five minutes or less mainly because of the re- quired space in the cloud. Here we propose to record such normal status events during an observation period over a timeframe of 1 to 12 months prefera- bly 12 months to cover all seasons to have such records avail- able with a certain reliability for further improvement of the parameter sets. These times need to be implemented in the pro- cedures for improvement for collecting these records. Advantageously for the method “recording by threshold” it should be defined how many events we need for an improvement, e.g. 10 recorded data sets, better more, maybe 100 might be 202212321 26 sufficient. In addition, the of such events could be defined also by the timeframe as described for normal status. In fact, more important than the number of events collected, is collection of “normal status signals” which are representa- tive for the site, and which cover all operation modes of the site. Therefore, a combination of both methods might be bene- ficial which could be trying to get records by threshold over a time as described between 1 to 12 months. If not sufficient events are detected then use the method to get records on pre- defined times as described above. Sufficient is again a number of records 10 and 100 but could be defined also specifically. Normally the detectors are installed with a parameter set at medium or even minimal sensitivity. By using the data records recorded with the described methods above, we could increase the sensitivity by applying more sensitive parameter sets un- til a certain level of pre-alarm is achieved. In this improvement procedure a simulator for detection behav- ior is used. This is basically a digital twin of the detector. All records need to run in this digital twin for all parameter sets which are more sensitive than the currently applied pa- rameter set (but we should not limit this only to the more sensitive parameter sets because sensitivity might not be the only property of a parameter set.). The sensitivity could be increased by selecting a certain pa- rameter set as far as the simulation results will not exceed a certain limit. The simplest way would be to use the limit of danger level 1 (DL 1). A certain limit of signal counts, for any sensor raw signal or danger signal, would also be possible. If we keep 202212321 27 the limit significantly below level 3 (DL 3) the simu- lated parameter set might be good enough to provide sufficient false alarm sensitivity. Sensitivity could be increased as far as all simulations of recorded data run below danger level 1 (or danger level 2 or 3) depending on the risk of false alarms the customer is willing to accept. The more precise and more general procedure could be achieved by keeping the results of simulations below a certain level of counts. This level could be agreed with the customer. The result of a possible improvement with higher sensitivity, respectively shorter time to alarm, could be documented in a report and agreed with the customer. Case 2: Decrease the sensitivity One prerequisite to decrease the sensitivity is the method of recording events by threshold triggering. Recording of data for a longer timeframe with hope to catch critical events (an- alogue to the procedure described in Case 1 seems not benefi- cial as it is less reliable to catch the critical time periods with events close to DL 3. But the general method is similar: We need to run simulations at least for a certain number of data records with events or for all events related to one de- tector in the digital twin for all parameter sets. All events need to be tagged. This means that we need to know if the event is related to a real fire or to a false alarm. Only tagged events should be used for simulations to provide a pa- rameter set selection. The target of such simulations is to get the lowest number of false alarms within a certain timeframe or the lowest level for the pre-alarm danger signal. 202212321 28 Simulations must run of all events tagged with “false alarm” for all parameter sets preferably for all param- eter sets with lower sensitivity than the currently applied parameter set (but we should not limit this only to the less sensitive parameter sets because sensitivity might not be the only property of a parameter set.). Then, you can select that parameter set, which results in the lowest number of events exceeding DL3 (or lower threshold e.g. DL2 or DL1 or a defined number of danger signal counts) to decrease the sensitivity by applying a more robust or appropriate parameter set. The result of a possible improvement with lower or more appro- priate sensitivity could be documented in a report and agreed with the customer. The operator could define or propose the maximum acceptable number of false alarms and the algorithm could propose an ap- propriate parameter set (if this is possible at all). Normally such an optimization will reduce the sensitivity in general within the allowed limits as all parameter sets are approved. Of course, the optimization algorithm should propose not the most robust parameter set but the parameter set, which will generate only a few accepted false alarms or the parameter set where all events of this detector won’t create any false alarm. It is basically not necessary to run simulations with all events tagged with alarm to prove that with the new setting the detector is still sensitive enough to detect all fires. All parameter sets are approved and should work properly. Nevertheless, we can do such simulation runs to show the cus- tomer, if a(n) (allowed) delay in fire detection will occur 202212321 29 when the new setting is This can be accomplished in case you have some events from real fires for an optimized de- tector, especially when test fires were performed after set- ting the fire detection system was set into operation. If these events are correctly tagged, then we can run a simula- tion with the optimized parameter. With a report showing the simulation results before and after the optimization we can prove, that the alarm will still take place within allowed limits or to show the customer a possible allowed additional delay in alarming time compared to the originally applied pa- rameter set. Figure 6 illustrates three exemplary types of recorded danger signal patterns or danger signals for alternative detector configurations. On the left side of figure 6 the signals for three different situations (or types) DPST1 to DPST3 as rec- orded are shown. On the right side of figure 6 the possible optimization of the respective situation is shown: ODPST1 to ODPST3. - Type 1 (situation shown at the top of figure 6): several parameter sets are applied but none of them could sup- press the first peak which lead to a danger level 3 (e.g. alarm situation). Nevertheless, the time until DL3 will be achieved is prolonged. - Type 2 (situation shown in the middle of figure 6): sev- eral optimization options are shown on the right side to suppress a danger level. Time is also delayed. - Type 3 (situation shown at the bottom of figure 6): The peak which achieves DL3 cannot be suppressed by optimiza- tion. Nevertheless, the picture on the right side shows, that the amplitude of the peak after optimization is lower. 202212321 30 Optimizing a hazard detector also be accomplished by im- proved handling of alarms and by optimal parameter set selec- tion at first day of operation. Case 3: Improve handling of alarms In some cases, a parameter set for suppressing a false alarm may not exist. Then, alternative measures can be defined for parameter set selection to improve the situation. Increase time between danger level DL1 (or danger level DL2) and danger level DL3. In some regions the fire panel triggers a pre-alarm allowing for local inspections before an automatic alarm is transmitted to the fire brigade. Typically, such in- spection procedure is triggered at danger level DL3. Similarly, it may be desirable to issue very early local warn- ings eg. at DL1 or DL2, to achieve a longer time for local in- spections, eg. if distance to inspection area is very long. Therefore, selecting a parameter set triggering danger level DL1 or danger level DL2 faster may be beneficial. Propose parameter set considering multi-detector dependencies If more than one detector is available in the area of interest (room, space, corridor, …), a fire panel can be configured to evaluate those detectors as a group. Such groupings for multi- detector dependencies, meaning that more than one detector would need to exceed a defined danger signal threshold, may also be proposed as a result of multiple detector simulations. Therefore, collecting the field data including its semantic tags is particularly valuable. The semantic information de- scribing relationships between detectors (e.g. located in room x, is-neighbor) can be derived from the P2-topology, existing 202212321 31 engineering tools but may also to be enriched from e.g. BIM data as walls within a fire zone need to be considered. Advantageously the selection or configuration of parameter sets for hazard control panels and for hazard detectors may also comprise multi-detector dependency. For instance, a panel considers measured values received simultaneously from other detectors located in the same room as a real hazard. Case 4: Select optimal parameter set at first day of operation Over time, the cloud-based approach for configuration of fire detection systems and collection of detector raw data provides a knowledge base from previous installations and their perfor- mances. Such experiences to some extend exist today, but are distributed across locations, or specific to individual ex- perts and therefore transferring such knowledge is more ef- fort. Less transparent is the performance / false alarm rate of sites. Each site may have its own characteristics, but after data from plenty of sites has been collected, each site can be cat- egorized into groups of sites which are sharing similarities, eg. kindergarden, shopping mall, office, hotel room, kitchen, restaurant, smoking area, various types of industrial sites, … or also geographic regions where people may have specific hab- its or different types of constructions. Preferably, the sites are tagged appropriately, or BIM models (Building Information Model) provided information on the type of building and rooms. State of the art signal clustering methods may assist. The groups are validated with respect to its similarities using also state of the art data analysis methods. The system can propose for a new site a typical parameter set being most used / having least false alarm rates, in case the 202212321 32 site matches existing groups on tags. This could then be applied by a service technician or automatically by the sys- tem. The initial settings will then be validated during / at the end of the observation period. A further aspect of the invention is annotating or tagging events (e.g. alarms) but also signals representing the “normal status”. The basic idea is as follows: The record only receives the full information content when it has been evaluated. This is particularly important if the alarm threshold has been ex- ceeded. Then it would be especially important to know whether it was a real fire or a false alarm. This can best be done by evaluating by a human expert. In order for the evaluation of a human to take place, the following implementations on the higher-level system are conceivable: 1. At the end of the recording of a record, the superor- dinate system compulsorily requires an evaluation of the event. For this purpose, a call is made in a spe- cific menu to evaluate the event. 2. An alarm must be acknowledged today at the panel. This acknowledgement takes place at a standardized panel. It would be conceivable to extend this acknowledge- ment, e.g. by a second button. However, it would have to be clarified whether this is permissible. 3. An automatic evaluation could also take place by the fact that with additional activation of one or more manual call points the one alarm is classified as a correct alarm. 4. An automatic evaluation could also be made (but with even greater uncertainty than variant 3) if several automatic fire detectors are additionally activated. 202212321 33 Then it can be assumed higher probability that it is a real alarm and not a false alarm. 5. All records are transferred to a maintenance center (e.g. a server configured to accomplish this task). This investigates in detail and evaluates the records. The inventive optimization procedures could run periodically, e.g. daily, weekly, monthly, or even every hour. If the opti- mization procedure finds detectors with parameter sets clearly below the optimal sensitivity (which means the parameter sets are more robust than necessary) the operator or the system could propose the customer to improve the protection of the system. This could be sold as a service. If the customer has an appropriate service contract the find- ings on parameter sets could be applied automatically or the customer will get an offering for improvement. In case of false alarms, we could run such an analysis as de- scribed above. The customer will get an offer to improve the false alarm robustness, if a more robust parameter set is available, which would provide at least a suppression of a part of false alarms. Exemplary advantages of the invention - Hazard protection will be improved. - False alarm suppression will be improved without going onsite. - Finding optimal parameter set to decrease the number of false alarms and to increase the level of safety in the building. 202212321 34 - Improvement of the performance of the hazard detection system: making the detectors more sensitive or less sensitive. A method and an arrangement for automatically providing suita- ble parameter sets for a hazard detection system (e.g. fire alarm system) to improve the detection performance of the de- tection system to increase the level of safety within the building.

202212321 35 Reference Signs HDS1 – HDS3 Hazard Detection System HCP1 – HCP3 Hazard Control Panel S Server DB Database AE Analytic Engine BIM Building Information Model SPS Set of Parameter Sets CC1 – CC7 Communication Connection HD1 - HD3 Hazard Detector DL Detector Line TM1 – TM4 Tagging Mechanism MV_NS Measured Values Normal Status MV_RH Measured Values Real Hazard MV_PH Measured Values Potential Hazard MV_NST Measured Values Normal Status Tagged MV_RHT Measured Values Real Hazard Tagged MV_PHT Measured Values Potential Hazard Tagged IPS Improved Parameter Set B User EG Edge Gateway MCP Manual Call Point HEM Hazard Event Message TSC Time Series Chart DC Danger Counts DM Danger Median Leg Legend DLLT1 Danger Level less than 1 DL1 Danger Level 1 DL2 Danger Level 2 DL3 Danger Level 3 D5to95 Danger 5% to 95% Percentile DMintoMax Danger Min to Max TSNS Timeserie Normal Status TSDL3E Timeserie DL3 Events DL3E1 – DL3E4 DL3 Events 202212321 36

Claims

202212321 37 1. A method for providing parameter sets for a hazard detec- tion system (HDS1 – HDS3), comprising a hazard control panel connected to hazard detectors, especially fire detectors, to improve the detection performance of the detection system, the method comprising: (S1) providing measured values (MV_NS) of at least one of the detectors representing the operation status in case no hazard is detected (“normal status”) by the respective detec- tor; (S2) providing measured values (MV_RH, MV_PH) of at least one of the detectors representing events (“different from nor- mal status”) in case a hazard (“real hazard”) or a potential hazard (“false alarm”, “No hazard”) is detected by the respec- tive detector; (S3) annotating the measured values (MV_RH, MV_PH) repre- senting an event in case of occurrence of a real hazard and in case of a potential hazard; (S4) analyzing the measured values (MV_NS) for the case no hazard is detected and the annotated measured values (MV_RHT, MV_PHT) representing events in case a hazard or a po- tential hazard is detected to identify a parameter set for the hazard detection system to reduce the number of false alarms and/or to minimize the detection time without increasing the number of false alarms. 2. The method according to claim 1, wherein also the measured values (MV_NS) representing the operation status in case no hazard is detected (“normal sta- tus”) are annotated to indicate the case no hazard is detected or the case a real hazard is detected; and 202212321 38 wherein said annotated values (MV_NST) are used in the analyzing step to identify a parameter set for the haz- ard detection system to reduce the number of false alarms and/or to minimize the detection time without increasing the number of false alarms. 3. The method according to one of the previous claims, wherein analyzing the measured values to identify a parameter set for the hazard detection system to reduce the number of false alarms and/or to minimize the detection time without increas- ing the number of false alarms is performed by simulating a digital twin of hazard detection system and/or by simulating a digital twin of the hazard detectors. 4. The method according to claim 3, wherein identifying the parameter set to reduce the number of false alarms and/or to minimize the detection time without increasing the number of false alarms is performed by the respective digital twin by evaluating existing parameter sets to select the best suitable parameter set. 5. The method according to one of the previous claims, wherein identifying the parameter set to reduce the number of false alarms and/or to minimize the detection time without increas- ing the number of false alarms is performed by a rule-based analytic engine running scenarios of existing parameter sets. 6. The method according to one of the previous claims, wherein the hazard detectors are fire detectors or smoke detectors or gas detectors or presence detectors. 7. The method according to one of the previous claims, wherein the measured values (MV_RH, MV_PH) representing events (“dif- ferent from normal status”) in case of a real hazard (“real 202212321 39 hazard”) or a potential hazard alarm”, “No hazard”) are provided based on threshold triggering. 8. The method according to one of the previous claims, wherein annotating (tagging) the measured values (MV_RH, MV_PH) repre- senting an event in case of occurrence of a real hazard and in case of a potential hazard is performed by a user by entering a respective acknowledgment on the hazard control panel. 9. The method according to one of the previous claims, wherein annotating (tagging) the measured values (MV_RH) representing an event in case of occurrence of a real hazard is performed by automatic approval of a second hazard detection system and/or by analyzing hazard events (HE) from manual call points (MCP). 10. The method according to one of the previous claims, wherein annotating (tagging) the measured values (MV_RH) rep- resenting an event in case of occurrence of a real hazard is performed when a defined number of further hazard detectors are also providing measured values (MV_RH) representing an event in case of occurrence of a real hazard. 11. An arrangement for providing parameter sets for a hazard detection system to improve the detection performance of the detection system, the arrangement comprising: a hazard control panel connected to hazard detectors; an analysis engine for analyzing measured values of at least one of the detectors; wherein at least one of the hazard detectors is config- ured to provide measured values (MV_NS) representing the oper- ation status (“normal status”) in case no hazard is detected to the analysis engine; 202212321 40 wherein at least one of hazard detectors is config- ured to provide measured values (MV_RH, MV_PH) in case a haz- ard (“real hazard”) or a potential hazard (“false alarm”, “No hazard”) is detected to the analysis engine; means for annotating (tagging) the measured values (MV_RH, MV_PH) representing an event in case of occurrence of a real hazard and in case of a potential hazard; wherein the analysis engine is configured to analyze the measured values (MV_NS) for the case no hazard is detected and the measured values (MV_RH, MV_PH) representing events in case a real hazard or a potential hazard is detected and the anno- tated measured values (MV_RHT, MV_PHT) representing events in case a hazard or a potential hazard is detected to identify a parameter set for the hazard detection system to reduce the number of false alarms and/or to minimize the detection time without increasing the number of false alarms. 12. The arrangement according to claim 11, wherein the analysis engine is configured to analyze the measured values to identify the parameter set for the hazard detection system to reduce the number of false alarms by simu- lating a digital twin of the hazard detection system and/or by simulating a digital twin of the hazard detectors. 13. The arrangement according to claim 11 or claim 12, wherein the analysis engine is a rule-based analytic engine configured to run scenarios of existing parameter sets to identify the parameter set able to reduce the number of false alarms. 14. The arrangement according to one of the claims 11 to 13, wherein the analysis engine is implemented in a cloud in- frastructure. 202212321 41 15. The arrangement according one of the claims 11 to 14, wherein the hazard detectors are fire detectors or smoke de- tectors or gas detectors or presence detectors. 16. A data processing system comprising instructions or means ^{to carry out a method according to one of the claims 1 to 10.}