In one example, generating the event device optimization analysis can include determining an optimal detection state for a detector of the event device 106-1 via the machine-learning model. In some examples, event device 106-1 can include multiple operational states, each having sensitivity levels that can vary within each state. For instance, the event device 106-1 can be a detector that can detect multiple fire phenomena, such as heat and CO (e.g., the event device 106-1 can be a dual optical heat CO detector). Such a device type may have a higher sensitivity than, for instance, a device that detects a single fire phenomenon (e.g., a smoke detector). The machine-learning model can determine that the event device 106-1 detects too many false alarm events because the event device 106-1 is located too proximate to a steam pipe in a space in the facility with engine exhaust being emitted, allowing the event device 106-1 to detect the heat from the pipe and the exhaust to interact with the event device 106-1. In such an instance, the machine-learning model can determine that the detection state of the event device 106-1 may be modified and/or the sensitivity level within the detection state can be modified (e.g., turned down) to prevent the event device 106-1 from activating in an instance when the pipe is hot and exhaust fumes are emitted.

Additionally, the event device optimization analysis can include a propensity rating surrounding the proposed detection state and/or optimal sensitivity level. For example, the machine-learning model can determine that turning down the sensitivity level for the dual optical heat CO detector device would have a 60% propensity rating for the likelihood of preventing the event device 106-1 from activating in an instance when the pipe is hot and exhaust fumes are emitted.

In another example, the event device 106-1 can be a device that detects a single event phenomenon. the machine-learning model can determine that the event device 106-1 detects many false alarm events because the event device 106-1 is an air quality sensor located too proximate to a door that is constantly being utilized (e.g., being opened), allowing unclean air to consistently interact with the event device 106-1. In such an instance, the machine-learning model can determine that the detection state of the event device 106-1 may be turned down to prevent the event device 106-1 from activating from so many false alarms.

Additionally, the event device optimization analysis can include a propensity rating surrounding the proposed optimal sensitivity level. For example, the machine-learning model can determine that turning down the sensitivity level for the air quality sensor would have an 80% propensity rating for the likelihood of preventing the event device 106-1 from activating in an instance when the door is opened and unclean air interacts with the event device 106-1.

In another example, generating the event device optimization analysis can include determining an alternative event device type for event device 106-1 via the machine-learning model. For example, the event device 106-1 can be a heat detector which may be located too proximate to a steam pipe (e.g., the temperature of which is provided to the computing device 102 via the building system information), and the machine-learning model can determine that the event device 106-1 detects too many false alarm events because the event device 106-1 unintentionally detects the heat from the steam pipe. In such an instance, the machine-learning model can determine that the event device type for event device 106-1 (e.g., a heat device) should be replaced with an alternative event device type, such as a smoke detector event device. Accordingly, the smoke detector event device would provide fire detection capabilities without being triggered by the heat of the proximate steam pipe.

Additionally, the event device optimization analysis can include a propensity rating surrounding the proposed alternative event device type. For example, the machine-learning model can determine that changing the heat detector to a smoke detector would have a 98% propensity rating for the likelihood of preventing the event device 106-1 from activating in an instance when the pipe is hot.

In a further example, generating the event device optimization analysis can include determining an alternative engineering configuration for event device 106-1 via the machine-learning model. Similar to the previous example, the event device 106-1 can be a heat detector which may be located too proximate to a steam pipe (e.g., the temperature of which, determined by a sensor, is provided to the computing device 102 via the building system information from the EMS 104), and the machine-learning model can determine that the event device 106-1 detects too many false alarm events because the event device 106-1 unintentionally detects the heat from the steam pipe. In such an instance, the machine-learning model can determine that event device 106-1 (e.g., a heat device) should include a co-incidence detection mechanism such that the event device 106-1 does not activate, even when an event is detected (e.g., a temperature is detected that exceeds a threshold temperature), unless event device 106-2 also detects the event. For example, if event device 106-1 detects the event, and event device 106-2 is also a heat detector that detects the heat and is not located proximate to the steam pipe, the event devices 106-1 and 106-2 can activate. As another example, if event device 106-1 detects the event, and event device 106-2 is a smoke detector that detects smoke, the event devices 106-1 and 106-2 can activate. As a further example, if event device 106-1 detects the event, and event device 106-2 is also a heat detector that does not detect the heat and is not located proximate to the steam pipe, the event devices 106-1 and 106-2 will not activate.

Additionally, the event device optimization analysis can include a propensity rating surrounding the proposed alternative engineering configuration. For example, the machine-learning model can determine that including co-incidence detection would have a 95% propensity rating for the likelihood of preventing the event device 106-1 from activating in an instance when the pipe is hot.

While determining an alternative engineering configuration is described above as including co-incidence detection, embodiments of the present disclosure are not so limited. For example, alternative engineering configurations can also include day night detection modules, alarm delay processes, etc.

In a further example, generating the event device optimization analysis can include determining whether to unlock a detector feature of the event device 106-1 via the machine-learning model. For instance, the event device 106-1 can be a detector that can detect multiple fire phenomena, such as heat and CO (e.g., the event device 106-1 can be a dual optical heat CO detector). However, the firmware of the event device 106-1 may allow only the heat detector to be utilized. The machine-learning model can determine that the event device 106-1 detects too many false alarm events utilizing only the heat detector because the event device 106-1 is located too proximate to a steam pipe, allowing the event device 106-1 to detect the heat from the pipe. In such an instance, the machine-learning model can determine that the CO detector feature of the event device 106-1 may be unlocked for use so that in order for the event device 106-1 to activate, the event device 106-1 has to detect both heat and CO. Such a feature unlock can prevent the event device 106-1 from activating in an instance when the pipe is hot.

Additionally, the event device optimization analysis can include a propensity rating surrounding the proposed feature unlock. For example, the machine-learning model can determine that unlocking the locked feature would have a 85% propensity rating for the likelihood of preventing the event device 106-1 from activating in an instance when the pipe is hot.

Additionally, while generating the event device optimization analysis is described above as including determining an optimal detection state, determining an alternative event device type, and determining an alternative engineering configuration, embodiments of the present disclosure are not so limited. For example, generating the event device optimization analysis can further include determining that another event device should be included, the location of the event device in the facility should be changed, etc. Further, the event device optimization analysis may include a combination of an optimal detection state, alternative event device type, and alternative engineering configuration. For instance, the event device optimization analysis may include changing the sensitivity of the detector of the event device 106-1 with a propensity rating of 60%, and an alternative event device type for the event device 106-1 with a propensity rating of 98%. Accordingly, a user reviewing the event device optimization analysis may utilize the propensity rating as a factor in determining which approach to take to reduce false alarm events for the event device 106-1 surrounding the steam pipe.

Upon generating the event device optimization analysis, the computing device 102 can transmit the event device optimization analysis to another computing device. For example, the computing device 102 can transmit the event device optimization analysis to the mobile device 112 and/or any other computing device to cause the event device optimization analysis to be displayed via a user interface.

For example, the mobile device 112 can include a user interface that can display the event device optimization analysis received from the computing device 102. The event device optimization analysis can be utilized to do various things, as is further described herein.

The event device optimization analysis can include a report. The report can summarize the event device optimization analysis, including event devices 106 of interest, possible recommendations for the event devices 106, etc.

In some examples, the report can include a recommendation to change a detection state of the detector of an event device 106 to an optimal detection state. For example, the event device 106-1 may be an air quality sensor having a sensitivity level 8. The event device optimization analysis can determine an optimum sensitivity level to be 7 based on the event device 106-1 having too many false alarms (e.g., from being exposed to unclean air from a door consistently being left open), and can include a recommendation to change the sensitivity level from sensitivity level 8 to the optimum sensitivity level (e.g., sensitivity level 7). Additionally, the event device optimization analysis can include a propensity rating associated with the recommendation to change the detection state.

In some examples, the report can include a recommendation to change the event device 106 to an alternative event device type. For example, the event device 106-1 may be a heat detector located too proximate to a steam pipe. The event device optimization analysis can determine the heat detector detects too many false alarm events because the event device 106-1 unintentionally detects the heat from the steam pipe, and can determine that the event device type for event device 106-1 should be replaced with a smoke detector event device type. The report can accordingly include the recommendation to change the event device 106-1 from a heat detector to a smoke detector. Additionally, the event device optimization analysis can include a propensity rating associated with the recommendation to change the event device 106 to an alternative event device type.

In some examples, the report can include a recommendation to change the engineering configuration for the event device 106 to an alternative engineering configuration. For example, the event device 106-1 may be a heat detector located too proximate to a steam pipe. The event device optimization analysis can determine the heat detector detects too many false alarm events because the event device 106-1 unintentionally detects the heat from the steam pipe, and can determine that the event device 106-1 (e.g., a heat device) should include a co-incidence detection mechanism such that the event device 106-1 does not activate, even when an event is detected (e.g., a temperature is detected that exceeds a threshold temperature), unless event device 106-2 also detects the event. The report can accordingly include the recommendation to include an alternative engineering configuration for the event device 106-1. Additionally, the event device optimization analysis can include a propensity rating associated with the recommendation to change the engineering configuration of the event device 106 to an alternative engineering configuration.

In some examples, the report can include a recommendation to unlock the detector feature in the event device 106. For example, the event device 106-1 may be a combination heat and CO detector with the CO detector being locked in the firmware of the event device 106-1, where the event device 106-1 is located too proximate to a steam pipe. The event device optimization analysis can determine the heat detector detects too many false alarm events because the event device 106-1 unintentionally detects the heat from the steam pipe, and can determine that the event device 106-1 (e.g., a heat device) should have the CO detection feature unlocked such that the event device 106-1 does not activate, even when an event is detected by the heat detector (e.g., a temperature is detected that exceeds a threshold temperature), unless the CO detector also detects the event (e.g., CO associated with an event). The report can accordingly include the recommendation to unlock the detector feature for the event device 106-1. Additionally, the event device optimization analysis can include a propensity rating associated with unlocking the detector feature of the event device 106.

In some examples, the computing device 102 can cause the event device 106-1 to change the detection state. For example, the report including the recommendation to change the detection state of the detector of an event device 106 to an optimal detection state can include a query displayed to a user (e.g., via the mobile device 112) as to whether to accept the recommendation or not. In response to an input to the mobile device 112 accepting the recommendation to change the detection state of the detector, the computing device 102 can cause the event device 106-1 to change the detection state of the detector to the optimal detection state. For instance, the computing device 102 can transmit a signal to the event device 106-1 that can modify the firmware of the event device 106-1 causing the detection state of the detector of the event device 106-1 to change.

While the computing device 102 is described above as changing the detection state of the detector of an event device 106 in response to a user input, embodiments of the present disclosure are not so limited. For example, the computing device 102 can, in some instances, cause the detector to change the detection state to the optimal detection state automatically based on the event device optimization analysis. For example, based on the event device optimization analysis indicating a change in the detection state of the detector, the computing device 102 can cause the event device 106-1 to change the detection state of the detector to the optimal detection state. For instance, the computing device 102 can transmit a signal automatically to the event device 106-1 that can modify the firmware of the event device 106-1 causing the detection state of the detector of the event device 106-1 to change to the optimal detection state determined by the event device optimization analysis.

Additionally, in some examples the computing device 102 can cause the event device 106-1 to change the engineering configuration to an alternative engineering configuration. For example, the report including the recommendation to change the engineering configuration of the detector of an event device 106 to an alternative engineering configuration can include a query displayed to a user (e.g., via the mobile device 112) as to whether to accept the recommendation or not. In response to an input to the mobile device 112 accepting the recommendation to change the engineering configuration of the event device 106, the computing device 102 can cause the event device 106-1 to change the engineering configuration of the detector to an alternative engineering configuration. For instance, the computing device 102 can transmit a signal to the event device 106-1 that can modify the firmware of the event device 106-1 causing the engineering configuration of the detector of the event device 106-1 to change.

Further, while the computing device 102 is described above as changing the engineering configuration of the detector of an event device 106 in response to a user input, embodiments of the present disclosure are not so limited. For example, the computing device 102 can, in some instances, cause the detector to change the engineering configuration to the alternative engineering configuration automatically based on the event device optimization analysis. For example, based on the event device optimization analysis indicating a change in the engineering configuration of the detector, the computing device 102 can cause the event device 106-1 to change the engineering configuration of the detector to the alternative engineering configuration. For instance, the computing device 102 can transmit a signal automatically to the event device 106-1 that can modify the firmware of the event device 106-1 causing the engineering configuration of the detector of the event device 106-1 to change to the alternative engineering configuration determined by the event device optimization analysis.

Moreover, in some examples the computing device 102 can cause the event device 106-1 to unlock the detector feature. For example, the report including the recommendation to unlock the detector feature in an event device 106 can include a query displayed to a user (e.g., via the mobile device 112) as to whether to accept the recommendation or not. In response to an input to the mobile device 112 accepting the recommendation to unlock the detector feature of the event device 106, the computing device 102 can cause the event device 106-1 to unlock the detector feature. For instance, the computing device 102 can transmit a signal to the event device 106-1 that can modify the firmware of the event device 106-1 causing the detector feature to be unlocked.

Further, while the computing device 102 is described above as unlocking the detector feature of the detector of an event device 106 in response to a user input, embodiments of the present disclosure are not so limited. For example, the computing device 102 can, in some instances, cause the detector to unlock the detector feature automatically based on the event device optimization analysis. For example, based on the event device optimization analysis indicating the detector feature should be unlocked, the computing device 102 can cause the event device 106-1 to unlock the detector feature. For instance, the computing device 102 can transmit a signal automatically to the event device 106-1 that can modify the firmware of the event device 106-1 causing the detector feature of the event device 106-1 to be unlocked.

As mentioned above, the computing device 102 can generate the event device optimization analysis via a machine-learning model using building system information. Such information can be received from various sources, as is additionally described in connection with FIG. 2.

FIG. 2 is an example of a computing device 202 having a machine-learning model 214 for generating an event device optimization analysis in an event system, in accordance with one or more embodiments of the present disclosure. The computing device 202 can receive building system information from a BMS 204, event device 206, and a mobile device 212.

As previously mentioned in connection with FIG. 1, the computing device 202 can utilize the machine-learning model 214 to generate the event device optimization analysis. The computing device 202 can further generate the machine-learning model 214 in order to generate such an optimization analysis. For example, the computing device 202 can utilize training data comprised of site data from the facility, and/or test data to generate decision rules for generating the event device optimization analysis. Such decision rules can be generated based on the training and test data, and can be utilized when generating the event device optimization analysis. For example, the training and test data can inform a decision rule such that under certain circumstances (e.g., if an event device encounters an unwanted alarm event over a threshold number of times under certain environmental conditions, which may be determined based on building system information from the BMS 204), a recommendation with respect to detection state, alternative event device type, alternative engineering configurations, and/or unlocking of detector features can be generated.

As illustrated in FIG. 2, the computing device 202 can receive building system information including pre-alarm data and alarm data from an event device 206. The pre-alarm data can be data from the event device