The learning variable may be, for example, a weight value representing connection strength between layers or nodes configuring an artificial neural network. Each of the layers may be a set of nodes. Here, a node may be an element which imitates a biological neural of brains of a person. Each database of the storage medium 52 may further store the learning variable received from a corresponding intelligent edge interrogator.

The long-term safety analysis AI model may include a plurality of AI models which are classified based on types of facilities, and for example, may include a bridge AI model 53A, a building AI model 53B, a tunnel AI model 53C, and a dam AI model 53D.

The bridge AI model 53A may analyze long-term sensing data of the bridges 13, 17, . . . stored in the bridge database 52A to predict and diagnose long-term safety of the bridges 13, 17, . . . on the basis of an analysis result. The building AI model 53B may analyze long-term sensing data of the buildings 10, 14, . . . stored in the building database 52B to predict and diagnose long-term safety of the buildings 10, 14, . . . on the basis of an analysis result. The tunnel AI model 53C may analyze long-term sensing data of the tunnels 12, 16, . . . stored in the tunnel database 52C to predict and diagnose long-term safety of the tunnels 12, 16, . . . on the basis of an analysis result. The dam AI model 53D may analyze long-term sensing data of the dams 11, 15, . . . stored in the dam database 52D to predict and diagnose long-term safety of the dams 11, 15, . . . on the basis of an analysis result.

The memory 53 may provide an execution space for executing long-term safety analysis AI models 53A, 53B, 53C, 53D, . . . stored in the storage medium 52, based on control by the processor 54. In this case, the memory 53 may be a volatile memory and may further include a non-volatile memory.

Long-term safety analysis AI models 55A, 55B, 55C, 55D, . . . executed by the processor 54 may analyze long-term sensing data of a corresponding facility received through a corresponding intelligent edge interrogator to predict and diagnose long-term safety of the corresponding facility.

Long-term diagnosis data obtained by predicting and diagnosing long-term safety may be expressed as scale data where risk situations of corresponding facilities are classified by classes, identically or similarly to the above-described short-term diagnosis data.

The processor 54 may continuously monitor risk situations of facilities distributed based on long-term diagnosis data, and when a risk situation (collapse risk) of a corresponding facility is recognized, the processor 54 may transmit a notification message to a facility management server (not shown). The facility management organization may provide announcement for evacuating inhabitants near the corresponding facility on the basis of the notification message from the server 50, or may transmit a notification message indicating the maintenance and repair of a corresponding facility installed in a specialized organization which performs the maintenance and repair of the corresponding facility.

The processor 54 may be a hardware element which controls and manages overall operations of the communication interface 51, the storage medium 52, and the memory 53 and may be implemented with at least one central processing unit (CPU), at least one graphics processing unit (GPU), or a combination thereof.

The processor 54 may load the long-term safety analysis AI models 53A 53B, 53C, 53D, . . . , stored in the storage medium 52, into the memory 53 to execute the AI models. The processor 54 may generate a notification message which evacuates inhabitants near the corresponding facility or indicates the maintenance and repair of the corresponding facility, based on long-term data generated from the executed long-term safety analysis AI models 53A 53B, 53C, 53D, . . . and may transmit the generated notification message to a relevant organization through the communication interface 51.

FIG. 2 is a block diagram schematically illustrating an internal configuration of an intelligent edge interrogator according to an embodiment of the present invention, and FIG. 3 is a diagram for describing short-term sensing data and long-term sensing data according to an embodiment of the present invention.

For conciseness of the drawings, FIG. 2 illustrates one facility of the facilities illustrated in FIG. 1 and one optical fiber sensor attached on the one facility. The optical fiber sensor illustrated in FIG. 2 is one of the optical fiber sensors 20 to 27 illustrated in FIG. 1. Reference numerals of the facility and the optical fiber sensor illustrated in FIG. 2 are omitted. Also, in FIG. 2, one intelligent edge interrogator 100 connected to the one optical fiber sensor is illustrated. The intelligent edge interrogator 100 may be one of the plurality of intelligent edge interrogators 30 to 37 illustrated in FIG. 1.

The intelligent edge interrogator 100 may include a light generator 101, a light receiver 102, an optical signal measurer 103, a memory 104, a communication interface 105, a processor 106, and a storage medium 107.

The light generator 101 may generate an optical signal of a broad wavelength band on the basis of control by the processor 106 and may transmit the optical signal to an optical fiber sensor through the first optical fiber line L1, and the light receiver 102 may receive an optical signal reflected from the optical fiber sensor, based on control by the processor 106.

The optical signal measurer 103 may measure an optical wavelength of the reflected optical signal received by the light receiver 102 on the basis of control by the processor 106 and may generate and output sensing data corresponding to the measured optical wavelength.

The processor 106 may load a data preprocessor 104A and a short-term safety analysis AI model 104B, stored in the storage medium 107, into the memory 104 and may execute the data preprocessor 104A and the short-term safety analysis AI model 104B. In a case where the data preprocessor 104A and the short-term safety analysis AI model 104B are loaded into the memory 104, the data preprocessor 104A and the short-term safety analysis AI model 104B may each be a software module, but are not limited thereto and may each be a hardware module.

The data preprocessor 104A may extract short-term sensing data and/or long-term sensing data from sensing data output from the optical signal measurer 103, based on control by the processor 106.

As illustrated in FIG. 3, short-term sensing data STSD may be obtained by sampling sensing data at a sampling period T consisting of a predetermined short time unit (for example, second (sec)). Long-term sensing data LTSD may be obtained by calculating an average value and/or a feature value of each of pieces of short-term sensing data STSD1, STSD2, . . . , and STSDn.