UWB-based intelligent wearable hazard detection system and hazard prediction method using the same

Abstract

A UWB-based intelligent wearable risk detection system and a risk prediction method using the same are disclosed. A UWB-based intelligent wearable risk detection device attached to a safety hook or protective gear worn by a worker may include a communication module that receives a UWB signal periodically transmitted from a tag installed in a risk zone and calculates the distance to the tag using the signal; a first sensor that detects whether the safety hook is fastened and whether the protective gear is worn; and a controller that determines whether safety measures are not implemented based on the detection result and generates a warning if the distance information is within a warning level threshold for each risk type in a non-implementation state.

Description

Technology Field

[0001] The present invention relates to a UWB-based intelligent wearable risk detection system and a risk prediction method using the same. Background Technology

[0003] Serious accidents, such as falls, electric shocks, falling objects, and suffocation, frequently occur in hazardous work environments like industrial sites, construction sites, electrical equipment areas, and confined spaces. These accidents are often caused by human error—such as worker negligence or failure to adhere to safety regulations—rather than by defects in machinery or equipment. Examples include failing to fasten safety harnesses while working at heights, not wearing insulating protective equipment during electrical work, or not wearing oxygen concentration meters in confined spaces.

[0004] Existing industrial safety management methods rely on passive and visual warnings, such as installing warning signs in hazardous areas, placing traffic cones (safety fences), and stationing monitors. However, these methods are ineffective in effectively preventing risks due to factors like reduced vigilance, changes in the work environment, and limited visibility. In particular, the lack of capabilities to detect real-time access to hazardous areas or verify whether workers have taken safety measures results in safety management focused on reactive response rather than accident prevention.

[0005] In the past, some Bluetooth Low Energy (BLE)-based proximity warning systems were introduced, but since BLE uses a distance measurement method based on signal strength indicators (RSSI), signal attenuation by obstacles such as metal, non-metal, and the human body is severe, and measurement errors are large. In addition, because it uses the 2.4 GHz band, radio interference from other wireless devices occurs, and there are limitations in accurately determining the actual distance between the danger zone and the worker.

[0006] In contrast, Ultra Wide Band (UWB)-based distance measurement technology calculates distance based on Time of Flight (TOF) rather than signal strength, allowing it to maintain high precision of ±0.5m even in environments with obstacles. Additionally, the variable frequency band of 3–10 GHz minimizes ambient interference and provides faster response speeds compared to BLE. These characteristics are highly advantageous for industrial safety management, such as detecting approach to hazardous areas, issuing warnings, and verifying worker safety measures.

[0007] However, existing UWB-based systems are often limited to simple distance measurement and warning functions, lacking the capability to analyze habitual risk behaviors occurring in repetitive work environments or individual risk tendencies of workers, and to predict the likelihood of accidents in advance. In particular, there is a shortage of technology that provides customized safety guidance by comprehensively analyzing individual worker behavioral patterns, responses to warnings and alerts, environmental changes, and past safety history. The problem to be solved

[0009] The present invention is intended to provide a UWB-based intelligent wearable risk detection system and a risk prediction method using the same.

[0010] In addition, the present invention aims to provide a UWB-based intelligent wearable risk detection system capable of measuring the distance between a worker and a risk zone with a precision of ±0.5m even in environments with metal, non-metal obstacles and human shielding by applying UWB technology, and a risk prediction method using the same.

[0011] In addition, the present invention aims to provide a UWB-based intelligent wearable risk detection system and a risk prediction method using the same, which can automatically or manually determine whether a worker has performed essential safety measures such as fastening a safety hook, wearing insulating protective gear, or attaching an oxygen concentration meter, and continuously provide a warning if the safety measures are not completed.

[0012] In addition, the present invention aims to provide a UWB-based intelligent wearable risk detection system and a risk prediction method using the same, which enable workers and surrounding personnel to intuitively recognize dangerous situations and respond quickly by providing LED colors, voice guidance, vibration warnings, etc., differentially according to the danger zone approach stage.

[0013] In addition, the present invention aims to provide a UWB-based intelligent wearable risk detection system and a risk prediction method using the same, which can analyze a worker's risk behavior patterns and tendency to approach risk zones and predict potential accident possibilities in advance by having an AI model learn UWB-based location data, warning / notification history, worker profile, environmental sensor data, and past accident / near-accident history.

[0014] In addition, the present invention aims to provide a UWB-based intelligent wearable risk detection system and a risk prediction method using the same, which can provide customized warning messages and safety guides based on the analysis results of each worker's risk tendency, and can convey the worker's risk assessment and the need for intervention to a manager in real time. means of solving the problem

[0016] According to one aspect of the present invention, a system for UWB-based intelligent wearable risk detection is provided.

[0017] According to one embodiment of the present invention, a UWB-based intelligent wearable risk detection device attached to a safety hook or protective gear worn by a worker may be provided, comprising: a communication module that receives a UWB signal from at least one tag installed in a risk zone and calculates distance information with the tag using the UWB signal; wherein the tag periodically transmits a UWB signal including risk information corresponding to the installation location, a distance threshold for a warning stage, and identification information; a first sensor that detects at least one of whether the safety hook is fastened and whether the protective gear is worn; and a controller that determines whether safety measures are not implemented by detecting at least one of whether the safety hook is fastened and whether the protective gear is worn based on the detection result of the first sensor, and controls to generate a warning in the form of audiovisual and tactile signals when the distance information is within a distance threshold for a different warning stage according to the type of risk, wherein the warning is generated with different intensities according to the warning stage.

[0018] The above controller can generate audiovisual and tactile warnings by controlling the LED color, voice guidance message, and vibration pattern differently according to the type of danger received from the tag.

[0019] The above controller can receive UWB signals from multiple tags and, in the event of simultaneous access to multiple danger zones, control the system to prioritize the generation of the warning with the highest warning level.

[0020] The above controller can further consider the risk prediction information and warning conditions for each worker provided by the server to set the distance threshold for each warning stage differently for each worker, and control the generation of warnings according to the distance threshold set differently for each worker.

[0021] The above tag updates the risk type, warning level distance threshold, and path pattern information stored in the tag according to changes in work process stages or field conditions, and the update may be performed by a control signal from a server or an administrator terminal.

[0023] According to another aspect of the present invention, a plurality of tags are installed in a hazardous area of ​​a work site, store risk information corresponding to the installation location, a distance threshold for a warning stage, a warning pattern, and identification information as tag information, and periodically transmit a UWB signal including said tag information; and an intelligent wearable risk detection device is attached to a safety ring or protective gear worn by a worker, receives a UWB signal from at least one of the plurality of tags, calculates distance information with said tag using said UWB signal, determines whether safety measures are not implemented based on at least one detection result among whether the safety ring is fastened and whether protective gear is worn, and generates an audiovisual and tactile warning if the distance information is within a different distance threshold for a warning stage according to a risk type while the safety measures are not implemented. A system may be provided comprising a server that preprocesses data collected from the intelligent wearable risk detection device and the administrator terminal and applies it to a deep learning-based risk prediction model to generate customized risk prediction information and warning conditions for each worker, wherein the intelligent wearable risk detection device additionally considers the customized risk prediction information and warning conditions for each worker provided by the server to set a distance threshold for each warning step differently for each worker, and generates a warning according to the distance threshold set differently for each worker.

[0024] The collected data above may include worker profiles, hazardous area access history, warning and notification history, and environmental sensor data.

[0025] The above server can update the risk type, warning stage distance threshold, and warning pattern information stored in the tag according to changes in the work process stage or site conditions.

[0027] According to another aspect of the present invention, a risk prediction method is provided.

[0028] According to one embodiment of the present invention, a risk prediction method may be provided, comprising: periodically receiving a UWB signal from at least one tag installed in a risk zone, the signal including at least some of risk information corresponding to the installation location, a warning stage distance threshold, a warning pattern, and identification information; calculating distance information with respect to the tag using the received UWB signal; detecting at least one of whether a safety ring is fastened and whether protective gear is worn to determine whether safety measures are not implemented; and, in the state of non-implementation of safety measures, if the distance information is within a different warning stage distance threshold according to the risk type, generating a warning in the form of audiovisual and tactile warnings, wherein the warning is generated with different intensities according to the warning stage.

[0029] The step of generating the above warning may further consider the risk prediction information and warning conditions for each worker provided by the server, set the distance threshold for the warning stage differently for each worker, and generate a warning according to the distance threshold set differently for each worker. Effects of the invention

[0031] By providing a UWB-based intelligent wearable risk detection system and a risk prediction method using the same according to one embodiment of the present invention, high distance measurement precision of ±0.5m can be secured by a UWB-based distance measurement method, thereby reliably determining whether to approach a risk zone.

[0032] In addition, the present invention enables differential warnings based on risk intensity by distinguishing detailed step-by-step access situations within a risk zone using warning step distance thresholds stored in a tag.

[0033] Furthermore, since the present invention can directly detect whether safety measures, such as whether a safety hook is fastened or protective equipment is worn, are being performed, it is possible to provide more accurate and situation-specific warnings than simple location-based warnings. Additionally, because it generates a warning only when safety measures are not implemented, it can reduce unnecessary warnings and avoid hindering work efficiency. Moreover, the present invention has the advantage of enabling rapid response even in multi-risk situations by changing LED colors, voice guidance messages, and vibration patterns based on risk type information stored in tags, allowing workers to intuitively recognize the type of risk, and by prioritizing warnings for the risk with the highest warning level when multiple tags are accessed.

[0034] In addition, the present invention has the advantage of automatically adjusting warning thresholds and intensity according to proficiency, violation history, and risk trend scores by reflecting AI-based risk prediction information and warning conditions for each worker.

[0035] In addition, the present invention allows for the remote updating of risk information, warning stage distance thresholds, and warning patterns stored in a tag via a server or administrator terminal, thereby enabling rapid response to changes in work process stages or on-site environments. It also has the advantage of merging environmental sensor data, such as CO₂, temperature, humidity, and oxygen concentration, to predict risks and adjust warnings in response to environmental changes.

[0036] In addition, the present invention can analyze risk trends by worker and predict the likelihood of accidents by learning past hazardous area access patterns, warning history, and environmental change data.

[0037] In addition, the present invention includes a multilingual voice guidance function in an intelligent wearable risk detection device, thereby providing customized warning messages tailored to the worker's nationality and language. This improves the delivery and comprehension of safety warnings in workplaces with a high proportion of foreign workers and prevents accidents caused by ignoring warnings. Brief explanation of the drawing

[0039] FIG. 1 is a diagram illustrating the configuration of a UWB-based intelligent wearable risk detection system according to an embodiment of the present invention. FIG. 2 is a drawing illustrating an example of attachment of a UWB-based intelligent wearable risk detection device according to an embodiment of the present invention. FIG. 3 is a block diagram schematically illustrating the internal configuration of an intelligent wearable risk detection device according to one embodiment of the present invention. FIG. 4 is a flowchart illustrating a risk prediction method using a UWB-based intelligent wearable risk detection device according to an embodiment of the present invention. FIG. 5 is a drawing illustrating an operation scenario for each risk level according to an embodiment of the present invention. FIG. 6 is a flowchart illustrating a deep learning-based customized risk prediction method according to an embodiment of the present invention. FIG. 7 is a drawing illustrating an example of a tag according to an embodiment of the present invention. Specific details for implementing the invention

[0040] As used in this specification, singular expressions include plural expressions unless the context clearly indicates otherwise. In this specification, terms such as "composed" or "comprising" should not be interpreted as necessarily including all of the various components or steps described in the specification, and should be interpreted as meaning that some of the components or steps may be excluded, or that additional components or steps may be included. Furthermore, terms such as "...part," "module," etc., as used in the specification refer to a unit that processes at least one function or operation, which may be implemented in hardware or software, or a combination of hardware and software.

[0041] Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings.

[0043] FIG. 1 is a diagram illustrating the configuration of a UWB-based intelligent wearable risk detection system according to one embodiment of the present invention, and FIG. 2 is a diagram illustrating an example of attachment of a UWB-based intelligent wearable risk detection device according to one embodiment of the present invention.

[0044] Referring to FIG. 1, a system (100) according to one embodiment of the present invention comprises a plurality of tags (110), an intelligent wearable risk detection device (120), an administrator terminal (140), and a server (130).

[0045] A plurality of tags (110) are installed in each danger zone (see FIG. 7) and are a means for periodically transmitting an identification signal containing danger information of the zone. For example, the plurality of tags (110) store information such as a danger type, warning level, and distance threshold corresponding to the installation location, and can enable real-time distance measurement through UWB communication with an intelligent wearable danger detection device (120).

[0046] For example, multiple tags (110) may be installed in multiple hazardous areas, such as high-altitude work zones, lifting work zones, near power distribution boards, and enclosed workspaces. Each tag (110) stores different warning levels and distance thresholds depending on the type of risk in the installed area. For example, a tag installed in a high-altitude work zone may be set to a warning threshold of 3m and a risk threshold of 1.5m, while a tag in a lifting work zone may be set to 5m and 3m. In this way, the distance threshold of each tag may be variably set according to the risk characteristics and work environment of the corresponding area.

[0047] Additionally, the tag (110) can store risk type, warning level distance threshold, warning pattern, and identification information depending on the installation location during initial installation. However, the data stored in the tag (110) may be changed remotely via a server (130) or an administrator terminal (140) when there is a change in the field situation or work plan. This allows the risk information and warning conditions to be kept up to date without the need for physical reinstallation of the tag (110). The intelligent wearable risk detection device (120) is attached to a safety hook or other protective gear worn by a worker and receives UWB signals periodically transmitted from multiple tags (110).

[0048] According to one embodiment of the present invention, a tag installed in a danger zone and an intelligent wearable danger detection device worn by a worker perform short-range wireless communication using the UWB method rather than the Bluetooth method.

[0049] The existing Bluetooth-based RSSI (Received Signal Strength Indicator) method is vulnerable to environmental noise, human shielding, and multipath interference, which can result in distance measurement errors of ±2 to 5 meters or more. In contrast, UWB enables Time of Flight (TOF)-based distance calculation, providing high measurement precision of ±10 to 50 cm. This allows for more precise setting of danger zone approach thresholds and enhances the reliability of warning timing.

[0050] In addition, since Bluetooth uses a narrow frequency band (several MHz), it suffers from severe interference from reflected and diffracted signals caused by metal structures, concrete walls, and the human body, whereas UWB uses ultra-wideband signals ranging from hundreds of MHz to several GHz, which can distinguish and eliminate individual path delay times, enabling stable measurements even in multipath environments such as construction sites.

[0051] Furthermore, since UWB provides short pulse widths and high symbol transmission speeds, it can detect location changes or approach events within milliseconds (ms). This is advantageous in situations requiring immediate warnings, such as the approach of heavy equipment or falling objects. Additionally, because UWB tags transmit short pulses periodically like beacons, they consume less power compared to Bluetooth based on the same distance measurement cycle, offering the advantage of extending the battery life of field-installed tags.

[0052] The intelligent wearable risk detection device (120) can determine whether a worker is approaching a risk zone by calculating the distance to the tag (110) through time-of-flight (TOF) analysis of the UWB signal received from the tag (110) and comparing the calculated distance with the risk type and level distance threshold stored in the tag. Additionally, the intelligent wearable risk detection device (120) includes a sensor that detects at least one of whether a safety hook is fastened or whether protective gear is worn, and if approach to a risk zone is detected while the safety measures are not implemented, it generates an audiovisual and tactile warning by controlling the intensity of the LED color, voice guidance, and vibration pattern differently according to the warning level.

[0053] Although the present specification describes the intelligent wearable risk detection device (120) as receiving UWB signals from the tag (110), it can measure the round-trip time of propagation by exchanging query-isobaric signals with the tag according to a bidirectional distance measurement protocol and calculate TOF-based distance information by correcting the processing delay on the tag side.

[0054] As another example, the intelligent wearable risk detection device (120) may calculate the distance using a unidirectional TOF method, and in such a case, the time information of the tag (110) and the intelligent wearable risk detection device (120) may be synchronized. That is, the intelligent wearable risk detection device (120) may calculate the TOF using the difference between the transmission timestamp included in the UWB signal transmitted from the tag (110) and the timestamp received by the intelligent wearable risk detection device (120).

[0055] As another example, when multiple tags are time-synchronized with each other, the intelligent wearable risk detection device (120) may calculate the location by measuring the difference in arrival times of UWB signals received from multiple tags, and then calculate the distance to the target tag or the proximity distance to a specific risk zone.

[0056] At this time, when multiple tag (110) signals are received simultaneously, the tag information of the highest risk level is applied first, and the risk information, threshold, and pattern of the tag (110) can be updated remotely through a server or administrator terminal. FIG. 2 illustrates an example in which an intelligent wearable risk detection device (120) is attached to a safety ring.

[0057] The administrator terminal (140) receives and displays location information, warning occurrence history, safety measure status, etc. transmitted from the intelligent wearable risk detection device (120), and can deliver warning notifications to workers or field managers when necessary.

[0058] The server (130) can analyze location data, worker profiles, warning history, and environmental sensor data collected from a plurality of intelligent wearable risk detection devices (120) and a manager terminal (140) using a deep learning-based risk prediction model to predict each worker's risk tendencies and potential accident possibilities, and provide the analysis results to the manager terminal (140) and the plurality of intelligent wearable risk detection devices (120).

[0060] FIG. 3 is a block diagram schematically illustrating the internal configuration of an intelligent wearable risk detection device according to one embodiment of the present invention.

[0061] Referring to FIG. 3, an intelligent wearable risk detection device (120) according to one embodiment of the present invention comprises a communication module (310), a sensor unit (320), a display unit (330), an audio output unit (340), a vibration unit (350), a memory (360), a power supply unit (370), and a controller (380).

[0062] The communication module (310) is a UWB-based communication module and is a means for transmitting and receiving UWB signals. For example, the communication module (310) can receive UWB signals transmitted from a plurality of tags (110), calculate the distance to the tags (110) based on the time of arrival (TOF), and transmit the calculated distance information to the controller (380).

[0063] In one embodiment of the present invention, the communication module (310) is illustrated as being equipped only with a UWB-based communication module, but additionally, a separate communication module capable of long-distance communication may be provided. Through this, data may be transmitted and received with a server or an administrator terminal, etc.

[0064] The sensor unit (320) is a sensor module that detects whether essential safety measures, such as whether a worker fastens a safety hook or wears protective gear, are performed, and may include detection means such as a magnetic sensor, a switch, an RFID reader, or a proximity communication module. The sensor unit (320) is connected to a controller (380) to transmit the detection result as a digital signal, and the controller (380) can determine whether to release the warning based on the detection result.

[0065] Different sensors may be attached depending on the target to which the intelligent wearable risk detection device is attached, and multiple sensors may be provided depending on the implementation method.

[0066] In addition, the sensor unit (320) may further include an environment sensor for detecting the surrounding environment conditions, in addition to a first sensor that detects whether the safety ring is fastened or whether protective gear is worn.

[0067] For example, the environmental sensor can measure physical and chemical environmental variables such as CO₂ concentration, temperature, humidity, oxygen concentration, current, and illuminance in real time. The environmental data collected in this way can be used to determine the warning of the device itself through a controller, and can be transmitted to a server (130) to be used as input data for a deep learning-based risk prediction model. Through this, not only simple distance-based warnings but also advance warnings and predictive warnings considering environmental change trends can be provided, and customized safety management according to environmental factors such as confined space work, electrical equipment inspection, and chemical handling is possible.

[0068] The display unit (330) is a means for providing a visual path by illuminating LEDs of different colors according to the danger zone approach stage under the control of the controller (380). In addition, the display unit (330) may also display the battery status and the device operation status.

[0069] The audio output unit (340) is a means for outputting audio, such as a danger zone access warning, safety measure guidance, and system status guidance, under the control of the controller (380). The audio output unit (340) may be, for example, a speaker.

[0070] Additionally, the audio output unit (340) can store a multilingual voice file in memory (360) and play a warning message in the language set in the worker profile or the language specified in the field manager terminal.

[0071] For example, multiple languages ​​such as Korean, English, Chinese, Vietnamese, and Thai can be stored, and voice guidance can be transmitted in the corresponding language by linking with an ID tag for each worker or a worker profile provided by the server (130).

[0072] The vibration unit (350) generates vibration when approaching a danger zone under the control of the controller (380), so that the worker can recognize the danger situation even in a noisy environment.

[0073] For example, the vibration unit (350) can generate different vibration intensity depending on the danger zone access stage.

[0074] The memory (360) can store various instructions, device identification information, distance threshold, user profile, voice data, system setting values, etc. for performing a risk prediction method using a UWB-based intelligent wearable risk detection device according to one embodiment of the present invention.

[0075] The power supply unit (370) supplies stable power to each component, including a battery, a power management IC (PMIC), and a charging circuit, and can support charging via a USB-C port and downloading a controller program.

[0076] The controller (380) is a means for controlling internal components (e.g., a communication module (310), a sensor unit (320), a display unit (330), an audio output unit (340), a vibration unit (350), a memory (360), a power supply unit (370), etc.) of an intelligent wearable risk detection device (120) according to one embodiment of the present invention.

[0077] Additionally, the controller (380) may determine whether to approach a danger zone based on distance data received from the communication module (310), control the operation of the display unit (330), audio output unit (340), vibration unit (350), etc. according to the result, and, if necessary, determine the warning release condition to control the operation of the device.

[0078] For example, a communication module (310) can receive UWB signals transmitted from a plurality of tags (110) and calculate the distance to the tags (110) based on the transmission arrival time. A controller (380) receives distance information calculated from the communication module (310) and can determine whether the operator has entered the yellow zone or red zone by comparing the distance information with a preset warning threshold or danger threshold. In one embodiment of the present invention, as shown in FIG. 5, the danger zone is divided into two stages and described based on this, but the danger zone may be subdivided into three or more stages. For example, it can be divided into a caution zone, a warning zone, and an emergency zone, and by applying different LED colors, flashing speeds, voice message content, vibration intensity, and patterns to each zone, the operator can intuitively recognize the severity of the danger.

[0079] Additionally, the controller (380) can detect whether the worker is performing safety measures through detection information input from the sensor unit (320). For example, the sensor unit (320) can detect whether the safety hook is engaged through a safety hook fastening sensor. As another example, the sensor unit (320) can detect whether an insulating protective device or an oxygen concentration meter is being worn through a protective device wearing sensor.

[0080] The controller (380) can determine whether to issue a warning by combining the distance determination result and the safety measure detection result. The controller (380) can generate various warning notifications in the form of audiovisual and tactile signals if safety measures are not implemented despite the worker approaching within the danger threshold.

[0081] Additionally, when the risk threshold is approached while safety measures are in place, the controller (380) may release the warning and light the display (330) in green to indicate that it is in a safe state.

[0082] Additionally, the controller (380) may apply an on-delay timer so that the voice guidance is not interrupted when a warning occurs, and may apply an off-delay timer so that the warning is not immediately terminated even when the operator moves out of the danger zone. Through this, warning malfunctions caused by intermittent signal interruptions or temporary position errors can be minimized.

[0083] Additionally, the intelligent wearable risk detection device (120) can minimize battery consumption by switching from an inactive state to a low-power mode. Additionally, the intelligent wearable risk detection device (120) may include a wake-up function that enters sleep mode when no signal from the tag (110) is detected for a certain period of time and is immediately activated upon receiving a signal.

[0085] FIG. 4 is a flowchart illustrating a risk prediction method using a UWB-based intelligent wearable risk detection device according to an embodiment of the present invention.

[0086] In step 410, the UWB-based intelligent wearable risk detection device (120) receives a UWB signal from at least one tag and calculates the distance to the tag using the received UWB signal. As described above, the UWB-based intelligent wearable risk detection device (120) can calculate the distance to the tag based on the time of arrival (TOF) using the received UWB signal.

[0087] The intelligent wearable danger detection device (120) can achieve a measurement accuracy of within ±0.5 m by applying a multi-sample average, Kalman filter, and multipath reflection removal algorithm to the measurement value when performing time-of-flight (TOF) based distance measurement on a UWB signal received from a tag (110). This enables precise determination of approach to a danger zone even in a narrow working environment.

[0088] Additionally, when the intelligent wearable risk detection device (120) receives UWB signals simultaneously from multiple tags (110), it may generate a warning by determining the nearest tag first based on signal strength (RSSI), channel impulse response (CIR), or distance calculation value. Additionally, the intelligent wearable risk detection device (120) may apply a multi-signal crosstalk removal algorithm to prevent unnecessary duplicate warnings.

[0089] According to another embodiment of the present invention, an intelligent wearable risk detection device (110) can receive a UWB signal from a device worn by another worker and calculate the proximity distance to other workers. At this time, if the calculated distance is below a set threshold, the intelligent wearable risk detection device (110) may warn of the risk of collision or entrapment in a confined space or notify a manager terminal (140) of the crowded state of multiple workers within a specific equipment operating area.

[0090] In step 415, the UWB-based intelligent wearable risk detection device (120) compares the calculated distance information with a distance threshold to determine whether the worker has approached a risk zone.

[0091] If you have not approached the danger zone, proceed to Step 410.

[0092] However, if it is determined that a danger zone has been approached, in step 420, the UWB-based intelligent wearable danger detection device (120) detects whether safety measures are performed.

[0093] For example, the UWB-based intelligent wearable risk detection device (120) can detect whether a safety ring is fastened or whether protective gear is worn by using a sensor equipped in the UWB-based intelligent wearable risk detection device, and can detect whether safety measures are performed based on the detected result.

[0094] If safety measures are not implemented, in step 425, the UWB-based intelligent wearable risk detection device (120) generates a warning alarm in the form of audiovisual and tactile when approaching a risk zone in a state where safety measures are not implemented.

[0095] At this time, the UWB-based intelligent wearable danger detection device (120) can generate a different audiovisual warning and a tactile warning alarm with a stronger vibration intensity as the distance to the danger zone decreases. For example, in the first danger zone, a yellow LED light and a warning voice are output, and when entering the second danger zone, a red LED light, a danger warning voice output, and a strong vibration warning are output in combination to induce the worker to immediately recognize and take safety measures. The second danger zone can be set to a distance closer to the tag than the first danger zone. In one embodiment of the present invention, for the convenience of understanding and explanation, it is assumed that the danger zone is divided into two zones and the explanation is centered on this, but it is obvious that the number of danger zones can be set and divided into three or more.

[0096] Referring to Fig. 5, we will explain the operation scenarios for each risk situation.

[0097] (1) Fall (upper) detection and warning

[0098] The intelligent wearable risk detection device (120) receives a UWB signal from a tag (110) installed in an area with a risk of falling, such as a high-altitude work section, end, or opening.

[0099] The intelligent wearable danger detection device (120) can output a yellow LED light and transmit "There is a fall hazard zone nearby. Please be careful." once and provide a vibration once based on the UWB signal received from the tag when the distance is within a first threshold (3.0m).

[0100] Next, the intelligent wearable danger detection device (120) can light a red LED, provide a voice guidance saying “This is a fall hazard zone. Please fasten the safety hook” when the measured distance approaches within a second threshold (1.5m), and repeatedly output vibrations for 3 seconds at 1-minute intervals.

[0101] In one embodiment of the present invention, the warning release condition may be applied by combining one or more of manual release, sensor-linked automatic release, and central control release. For example, in step 1, a field manager manually releases the warning, in step 2, automatically releases it when safety measures are completed in conjunction with a safety hook sensor, etc., and in step 3, releases it after checking whether work is permitted in a central control manner through an app (TBM+MNT)-based server (130).

[0102] The conditions for deactivating the warning vary depending on the device generation; for the 1st generation, it is manually deactivated by an administrator, for the 2nd generation, it is automatically deactivated via the safety hook sensor, and for the 3rd generation, it can be deactivated by verifying work authorization status through central control based on an app (TBM+MNT). The TBM records the work details, risk factors, and safety measures to be performed on the corresponding work day in advance, and stores the work authorization status for each worker in a database.

[0103] The MNT function collects real-time location data, whether a danger zone has been accessed, whether safety measures have been implemented, and a history of warning occurrences from an intelligent wearable danger detection device (120) deployed at the site, and displays them on a central server (130) and an administrator terminal (140).

[0104] The server (130) can combine TBM data and MNT data to comprehensively determine whether a specific worker has been granted permission to work in the relevant hazardous area, is currently implementing safety measures, or if an emergency situation has occurred. In this way, the server (130) can execute a central control-based warning release logic based on TBM+MNT, thereby increasing the efficiency of safety management throughout the site and effectively restricting access to hazardous areas by unauthorized personnel.

[0105] (2) Fall (lower hazard) detection and warning

[0106] The intelligent wearable risk detection device (120) receives a UWB signal containing information on the risk of falling objects caused by upper work from a tag (110) installed on the ground.

[0107] The intelligent wearable danger detection device (120) lights up a yellow LED and provides a voice guidance and one vibration when distance information measured based on a UWB signal approaches within a first threshold (3.0m), saying, “Upper work is in progress nearby. Please be careful.”

[0108] Additionally, the intelligent wearable danger detection device (120) can repeat a red LED light, a voice guidance “Upper work in progress. Please move out of the danger zone,” and a 3-second vibration at 1-minute intervals when the calculated distance information approaches within a second threshold (1.5m).

[0109] In such cases, the warning release conditions can be performed using a first-generation manual release and a third-generation central control method via a server, but the second-generation automatic release conditions may not be applied.

[0110] (3) Detection and warning of lifting (crane) work zone

[0111] The intelligent wearable risk detection device (120) receives a UWB signal containing lifting risk information from a tag (110) installed in an external crane lifting area. If the distance calculated based on the UWB signal received from the tag is within a first threshold (3.0m), the intelligent wearable risk detection device (120) may light up a yellow LED, provide a voice guidance “Crane lifting is in progress nearby. Please be careful.” and provide one vibration.

[0112] In addition, the intelligent wearable danger detection device (120) can repeat a red LED light, a voice guidance “Crane lifting is in progress. Please move out of the danger zone,” and a 3-second vibration at 1-minute intervals when the distance approaches within a second threshold (1.0m).

[0113] In such cases, the warning release condition can be performed by manual release by a site manager (1st generation) or by a central control method through a server (130) (3rd generation), and the 2nd generation automatic release condition may not be applied because it is difficult to determine whether the risk is resolved through linkage with the safety measure detection unit due to the nature of the lifting operation.

[0114] (4) Detection and warning of boarding of high-altitude work equipment

[0115] A tag (110) installed on a high-altitude work facility or a scaffolding ladder may have a first threshold (1.0m) set for detecting boarding. Accordingly, when distance information calculated based on a UWB signal received from the tag approaches within the first threshold, the intelligent wearable risk detection device (120) may light a red LED, provide a voice guidance saying “Please attach the safety hook for high-altitude work,” and repeat a 3-second vibration at 1-minute intervals.

[0116] The warning release conditions may be applied by combining one or more of manual release, sensor-linked automatic release, and central control release. This is as previously stated.

[0117] (5) Switchboard access detection and warning

[0118] The intelligent wearable risk detection device (120) can receive a UWB signal including a risk type, a warning level distance threshold, and an ID from a tag (110) installed at the entrance of the power distribution panel. Then, the intelligent wearable risk detection device (120) calculates distance information based on the UWB signal received from the tag, and if the distance information is within the warning level distance threshold (e.g., 1.0 m), it can light a red LED, provide a voice guidance saying “Please check safety equipment when entering the power distribution panel,” and repeat a 3-second vibration at 1-minute intervals.

[0119] In this embodiment, the warning release conditions may be applied by combining one or more of manual release by a field manager, automatic release via sensor linkage through detection of a live line alarm and wearing insulation, or remote release by a central control system.

[0120] A live line alarm is a device that detects whether voltage is applied to electrical equipment or wires in a non-contact manner and warns a worker. In one embodiment of the present invention, an intelligent wearable risk detection device (120) is linked with a live line alarm to determine whether voltage is applied when approaching a switchboard or electrical equipment, and can release the warning if an inactive state is confirmed.

[0121] (6) Detection and warning of entry into enclosed sections

[0122] The intelligent wearable risk detection device (120) can receive a UWB signal including a risk type, a warning level distance threshold, and an ID from a tag (110) installed at the entrance of a confined space. Then, the intelligent wearable risk detection device (120) calculates distance information based on the UWB signal received from the tag, and if the distance information is within the warning level distance threshold (e.g., 1.0 m), it can light a red LED, provide a voice guidance saying “Please check the oxygen concentration meter when entering or exiting the confined section,” and repeat a 3-second vibration at 1-minute intervals.

[0123] In this case, the warning release conditions may be applied by combining one or more of manual release, sensor-linked automatic release (automatic release when oxygen concentration meter power is turned on), and central control release.

[0124] (7) Detection and warning of heavy equipment access

[0125] The intelligent wearable danger detection device (120) can repeat a red LED light, a voice guidance “Heavy equipment and vehicles are moving. Be careful of pinching and collisions,” and a 3-second vibration at 1-minute intervals based on a UWB signal received from a tag (110) installed in a heavy equipment work zone when the distance information is within a threshold (1.5m).

[0126] In this embodiment, the warning release condition may vary depending on the device generation, and in the first generation, it may be released manually by a field manager, and in the third generation, it may be released via a central control method through a server (130). However, since it is difficult to determine whether the risk of heavy equipment access is resolved through linkage with the safety measure detection unit, the second generation automatic release condition may not be applied.

[0127] (8) Chemical facility open detection and warning

[0128] The intelligent wearable risk detection device (120) can receive a UWB signal containing chemical risk information from a tag (110) installed in the area when the chemical facility (CIB, VMB) is opened. Subsequently, the intelligent wearable risk detection device (120) can provide an LED warning, guidance voice, and vibration according to the type of risk when distance information calculated based on the UWB signal approaches within a threshold value.

[0129] In this embodiment, the warning release conditions vary depending on the device generation, and in the first generation, manual release through confirmation by a field manager, in the second generation, automatic release linked to a sensor that detects whether hazardous material handling equipment is being worn, and in the third generation, release after confirming whether work is permitted and the risk exposure status by linking with a central control system.

[0130] The server (130) stores and manages language profiles for each worker and can transmit voice data packages corresponding to the language together when transmitting warning conditions. For example, when a worker of Vietnamese nationality approaches a high-altitude work zone, “ It can output a warning message in Vietnamese saying "" and provide the same warning in Korean to Korean workers.

[0132] As described above, the intelligent wearable risk detection device (120) can receive a UWB signal from a tag that includes a risk type, a step threshold, an ID, etc. The intelligent wearable risk detection device (120) can calculate distance information using the UWB signal received from the tag and selectively generate a visual, auditory, or tactile warning corresponding to the risk level by comparing the calculated distance information with the step threshold. Through this, even for the same risk type, different LED colors, guidance voices, and vibration patterns are provided depending on the approach distance level, allowing the worker to intuitively recognize the dangerous situation and respond appropriately.

[0134] FIG. 6 is a flowchart illustrating a deep learning-based customized risk prediction method according to an embodiment of the present invention.

[0135] In step 610, the server (130) collects data from a plurality of intelligent wearable risk detection devices (120) and a manager terminal (140).

[0136] For example, the server (130) may collect UWB-based location / movement pattern data including at least one of the frequency of access to the danger zone, duration of stay, speed and direction of movement, history of habitual entry, and proximity patterns between workers, and may also collect warning / notification history data including the time of occurrence, number of occurrences, worker reaction, release method, and time taken. In addition, the server (130) may collect worker profiles including training history, proficiency, and history of safety rule violations, and may also collect environmental sensor data regarding CO₂, temperature, humidity, current, oxygen concentration, etc. In addition, the server (130) may collect accident / near-accident history data including the type of occurrence, time of occurrence, and cause analysis results.

[0137] In step 620, the server (130) preprocesses the collected data.

[0138] For example, the server (130) can normalize data with different measurement units to align them to the same standard and filter out abnormal or error values ​​through outlier removal. Additionally, the server (130) can perform time-series synchronization to align the time axis of each data and can perform missing value processing by correcting missing data or inputting replacement values. Through these preprocessing steps, the server (130) ensures data consistency and reliability and improves the accuracy of AI analysis.

[0139] In step 630, the server (130) inputs the preprocessed data into a deep learning-based risk prediction model to output risk prediction information and warning conditions. The input data includes UWB-based location and movement pattern data, warning and notification history, environmental sensor data, accident / near-accident history, and a worker profile including the worker's training history, proficiency, and history of safety rule violations.

[0140] Deep learning-based risk prediction models may include models based on recurrent neural networks (RNN) or long-short-term memory (LSTM), through which they can learn changes in workers' behavioral patterns from time-series data and predict future risky behaviors based on past history of entering and staying in risk zones.

[0141] In addition, deep learning-based risk prediction models may include regression models, random forests, and gradient boosting-based classification models, which can combine worker profiles and field data to predict the probability of specific risk behaviors occurring, such as not wearing a safety helmet or leaving a designated area.

[0142] In addition, deep learning-based risk prediction models may include anomaly detection models based on Isolation Forest or One-Class SVM, and after learning normal behavior patterns, they can detect abnormal behaviors different from usual (e.g., ignoring warnings, wandering in danger zones, etc.) in real time.

[0143] In addition, deep learning-based risk prediction models may include reinforcement learning-based models and, as a long-term goal, learn the effectiveness of warnings and whether to prevent accidents to independently determine the optimal warning timing, warning content, and the need for manager intervention.

[0144] The server (130) analyzes the characteristics of dangerous behavior patterns by considering the worker profile and can apply different weights depending on the worker's safety habits and proficiency, even for the same dangerous zone access. For example, for a worker with low proficiency and a history of violating safety rules, the warning threshold can be set further and the LED, voice, and vibration patterns can be strengthened, while conversely, for a worker with sufficient safety training history and a quick response to warnings, the threshold can be reduced or the frequency of warnings can be mitigated to reduce unnecessary warnings.

[0145] In addition, the server (130) evaluates the performance of the prediction model by continuously monitoring changes in the worker's behavior (whether safety rules are followed, whether the worker leaves the danger zone, etc.) and whether an actual accident occurs after the prediction warning. The results of this evaluation and newly collected data from the field are periodically used for model retraining, thereby continuously improving prediction accuracy and reliability by adapting to the changing work environment and worker behavior patterns. Furthermore, by reflecting feedback from safety management experts to analyze the causes of incorrect predictions and improving the model structure or learning parameters, the system can be optimized.

[0146] When a predicted warning occurs or a dangerous situation is detected, the server (130) sends a notification to the administrator through the administrator terminal (140), and the administrator can provide feedback through a web dashboard or mobile app.

[0147] Feedback may include evaluations of accuracy (Positive Confirmation), which verifies that prediction or detection results match actual hazardous situations or contributed to accident prevention; false positives, which report a warning even though no risk exists; and false negatives, which report an actual hazardous situation that the deep learning-based risk prediction model failed to detect. Additionally, feedback may include detailed information on risk factors, such as sudden forklift acceleration; worker specifics, such as deteriorating worker condition; and site-specific information, such as insufficient lighting or weather conditions. It may also include an efficiency evaluation regarding the appropriateness of the warning's timing, intensity, content, and delivery method. Furthermore, managers can designate a specific judgment as the top priority (Override) to enforce learning, ensuring that the system prioritizes that judgment in identical or similar situations in the future.

[0148] Additionally, the server (130) can update the AI ​​model based on collected administrator feedback. In this case, through weight adjustment, the weight of the judgment logic where accurate feedback is provided can be strengthened, the weight of the judgment logic where false positive feedback is provided can be decreased, and situations where no positive feedback is provided can be added as new training data to retrain the model. Furthermore, through rule-based learning, explicit warning rules can be generated for situations where false positives occur repeatedly or overrides are applied. Moreover, by applying a combination of online learning and batch learning, false positive feedback can be reflected in real time, and at the same time, all feedback data and field data can be periodically integrated and retrained to improve model performance. In this case, override data can be reflected with the highest priority.

[0149] The risk prediction information derived from this analysis may include risk scores per worker, tendencies to approach hazardous areas, and the likelihood of potential accidents, while warning conditions may include warning thresholds, LED, voice, and vibration patterns and intensities, warning repetition cycles, and customized safety guide messages.

[0150] In step 640, the server (130) finally sets customized warning conditions for each worker based on risk prediction information and warning conditions. For example, the server (130) can adjust the conditions to expand the warning threshold and strengthen LED, voice, and vibration patterns for workers with high risk prediction results, and to minimize unnecessary warnings for workers with low risk prediction results.

[0151] Depending on the implementation method, the server (130) may generate a prediction warning if necessary. For example, the conditions for generating a prediction warning may consist of (1) when the risk trend score exceeds a set threshold (score-based prediction warning), (2) when real-time behavior similar to past risk behavior patterns is detected (pattern-based prediction warning), and (3) when environmental sensor data shows a risk change trend but has not yet reached a warning threshold (environmental change-based prediction warning).

[0152] At this time, the predictive warning may be transmitted in a transmission method that includes one or more of the following: a worker individual notification method that provides customized predictive warnings to the worker, such as voice, vibration, or LED flashing, through a UWB wearable device; a manager notification method that provides risk scores, expected dangerous behaviors, and information on potential dangerous zones through a web dashboard or mobile app; and a preemptive intervention method that induces a manager to issue safety instructions in real time or take additional safety measures for dangerous zones.

[0153] In step 650, the server (130) transmits warning conditions and safety guide messages to the intelligent wearable risk detection device (120) and the administrator terminal (140). Here, the warning conditions and safety guide messages may include detailed setting values ​​such as distance threshold, warning level, LED color pattern, voice guidance phrase, vibration intensity, and period.

[0154] The intelligent wearable risk detection device (120) stores warning conditions received from the server (130) in its internal memory and can control the warning by comparing the distance measurement result with the tag (110) with the corresponding condition. Through this, the intelligent wearable risk detection device (120) can apply different LED, voice, and vibration patterns and warning thresholds to each worker even in the same risk zone, thereby providing a risk warning optimized for the worker's characteristics and the field situation.

[0157] An apparatus and method according to an embodiment of the present invention may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either alone or in combination. The program instructions recorded on the computer-readable medium may be those specifically designed and configured for the present invention, or they may be those known and available to a person skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc.

[0158] The hardware device described above may be configured to operate as one or more software modules to perform the operation of the present invention, and vice versa.

[0159] The present invention has been described above with reference to its embodiments. Those skilled in the art will understand that the present invention may be implemented in modified forms without departing from the essential characteristics of the invention. Therefore, the disclosed embodiments should be considered in an illustrative rather than a restrictive sense. The scope of the invention is defined by the claims, not by the foregoing description, and all variations within the scope of the claims should be interpreted as being included in the invention.

Claims

Claim 1 A UWB-based intelligent wearable risk detection device attached to a safety hook or protective gear worn by a worker, comprising: a communication module that receives a UWB signal from at least one tag installed in a risk zone and calculates distance information with respect to the tag using the UWB signal; wherein the tag periodically transmits a UWB signal including risk information corresponding to the installation location, a distance threshold for a warning stage, warning pattern information, and identification information; and a first sensor that detects at least one of whether the safety hook is fastened and whether the protective gear is worn. The system includes a controller that controls the generation of audiovisual and tactile warnings when the distance information is within a distance threshold of a different warning stage according to the risk type in a state where safety measures are not implemented, and controls the generation of warnings with different intensities according to the warning stage; the tag stores tag information including risk type, warning stage distance threshold, and warning pattern information according to the installation location, and updates the tag information according to the work process stage or changes in the site situation; the controller extracts the tag information from the received UWB signal, verifies whether the safety measures are not implemented detected through the first sensor, and independently determines whether the warning is generated by directly comparing the calculated distance information with the step distance threshold; when the UWB signal is received from multiple tags, controls the generation of the warning with the highest warning stage according to the risk type included in the tag information as a priority; and applies an on-delay timer to prevent voice guidance from being interrupted when the warning is generated, and an off-delay timer to prevent the warning from immediately ending even if the worker temporarily leaves the danger zone or the signal is disconnected. UWB-based intelligent wearable risk detection device that controls the maintenance and termination of notifications. Claim 2 A UWB-based intelligent wearable risk detection device according to claim 1, wherein the controller further considers risk prediction information and warning conditions provided by a server for each worker to set the warning step distance threshold differently for each worker, and controls the generation of a warning according to the distance threshold set differently for each worker. Claim 3 A UWB-based intelligent wearable risk detection device according to claim 1, characterized in that the update is performed by a control signal from a server or an administrator terminal. Claim 4 A plurality of tags installed in a hazardous area of ​​a work site, storing risk information corresponding to the installation location, a distance threshold for a warning stage, warning pattern information, and identification information as tag information, and periodically transmitting a UWB signal including said tag information; an intelligent wearable risk detection device attached to a safety hook or protective gear worn by a worker, receiving a UWB signal from at least one of said plurality of tags, calculating distance information to said tag using said UWB signal, determining whether safety measures are not implemented based on at least one detection result among whether the safety hook is fastened and whether protective gear is worn, and generating an audiovisual and tactile warning if the distance information is within a different distance threshold for a warning stage according to a risk type while safety measures are not implemented. The system comprises a server that preprocesses data collected from the intelligent wearable risk detection device and the administrator terminal and applies it to a deep learning-based risk prediction model to generate customized risk prediction information and warning conditions for each worker, wherein the intelligent wearable risk detection device additionally considers the customized risk prediction information and warning conditions for each worker provided by the server to set distance thresholds for each warning stage differently for each worker, generates a warning according to the distance thresholds set differently for each worker, extracts tag information from the UWB signal to independently determine whether a warning is generated by verifying whether the safety measures are not implemented and comparing distance thresholds, prioritizes generating the warning with the highest warning stage when signals are received from multiple tags, and controls the maintenance and termination of warning notifications by applying an on-delay timer to prevent voice guidance from being interrupted in the middle when the warning is generated and an off-delay timer to prevent the warning from being terminated immediately even if the worker temporarily leaves the danger zone or the signal is disconnected. Claim 5 A step of periodically receiving a UWB signal from at least one tag installed in a danger zone, the signal including at least some of danger information corresponding to the installation location, a distance threshold for a warning stage, warning pattern information, and identification information, and extracting tag information from the UWB signal; a step of calculating distance information to the tag using the received UWB signal; a step of determining whether safety measures are not implemented by detecting at least one of whether a safety hook is fastened and whether protective gear is worn; A risk prediction method comprising the step of, in a state where the above safety measures are not implemented, if the distance information is within a distance threshold of a different warning stage according to the risk type, generating a warning in the form of audiovisual and tactile warnings, generating a warning with a different intensity according to the warning stage, and the step of generating the warning further comprising the step of independently determining whether to generate a warning by verifying whether the above safety measures are not implemented and comparing distance thresholds based on the extracted tag information, and when signals are received from multiple tags, applying the risk information with the highest warning stage first, and controlling the maintenance and termination of the warning notification by applying an on-delay timer to prevent the voice guidance from being interrupted in the middle and an off-delay timer to prevent the warning from being terminated immediately even if the worker temporarily leaves the risk area or the signal is disconnected.

Images