Ai-based smart maintenance system for BIM

Abstract

The present invention relates to a BIM-based smart maintenance system utilizing AI, comprising: a collection unit that collects sensor data and location data from a plurality of facilities; a detection unit that inputs the sensor data, the location data, and metadata of the facilities into an artificial neural network model to detect abnormal signs and identify abnormal facilities; a visualization unit that visualizes the operating status and type of abnormality of the abnormal facilities; and a guide unit that provides a response guide according to the type of abnormality of the abnormal facilities to an operator terminal.The system includes a control unit that receives a signal from the operator terminal and corrects the control parameters of the abnormal equipment, and the detection unit includes a learning module that inputs the sensor data, the location data, and metadata of the equipment into the artificial neural network model to learn the range of normal sensor values ​​according to the locations of the equipment; a judgment module that identifies the abnormal equipment based on the range of normal sensor values ​​and the sensor data to determine the type and degree of the abnormality; and a ranking module that generates ranking information regarding the abnormal equipment that requires priority action based on the type of abnormality, the degree of the abnormality, and environmental factors specific to the equipment, wherein the judgment module determines whether equipment adjacent to the abnormal equipment deviates from the range of normal sensor values ​​to track the path of abnormality diffusion, and the environmental factors specific to the equipment include the introduction price of the equipment, the number of sensor data generated by the equipment, the presence or absence of backup equipment capable of replacing the equipment, and the frequency of worker access by zone according to a preset radius centered on the location of the abnormal equipment, wherein the frequency of worker access by zone is defined as follows: a zone less than 50m formed in a concentric shape according to the radius distance centered on the abnormal equipment is designated as Zone 1, and a zone 50m or more The area less than 100m is divided into a second zone and the area greater than 100m is divided into a third zone, and is calculated based on frequency and distance weights of workers who have passed through the first zone, the second zone, and the third zone during a preset time; the visualization unit is characterized by visually representing the abnormal diffusion path in a three-dimensional space according to the ranking information; and the control unit includes a verification module that, upon receiving a signal from the operator terminal, determines whether the signal complies with preset rules and calculates verification information, and a correction module that receives the verification information and corrects the control parameters of the abnormal equipment.

Description

Technology Field

[0001] The present invention relates to a BIM-based smart maintenance system utilizing AI. Background Technology

[0002] Recently, the adoption of smart factories, which collect and analyze data in real time by attaching sensors to equipment, is accelerating in manufacturing and process industries. In particular, since sudden equipment shutdowns directly lead to reduced productivity and massive economic losses, maintenance technology that monitors equipment status in real time and detects and responds to abnormal signs in advance has established itself as an essential element of the smart manufacturing environment.

[0003] Conventional facility management systems rely on a simple monitoring method that sets specific thresholds and triggers alarms when sensor data deviates from them. However, in large-scale facility complexes, the organic interrelationships between equipment and environmental variables related to installation locations interact in a complex manner. Consequently, relying solely on simple threshold settings makes it difficult to accurately identify abnormal signs that reflect location-specific characteristics, leading to the problem of frequent false alarms.

[0004] In particular, when an anomaly originating in a specific piece of equipment is transferred or spreads to adjacent equipment, existing systems provide only fragmentary alarms and fail to provide visualization functions that allow for the intuitive identification of the path of the anomaly's spread. Furthermore, when an operator modifies the control parameters of the equipment in response to an abnormal situation, there is a persistent risk that the operator's operational errors or inappropriate signals may be directly reflected in the system, potentially causing secondary failures or safety accidents. Prior art literature

[0005] Republic of Korea Registered Patent Publication No. 10-2042616 The problem to be solved

[0006] The objective of the present invention is to provide a BIM-based smart maintenance system utilizing AI that solves the aforementioned conventional problems by analyzing sensor data and location data collected from multiple facilities using an artificial neural network model to detect abnormal signs, identifies the abnormal facilities and visualizes their diffusion paths to display them to the operator, and verifies the operator's control signals to correct the parameters of the facilities. means of solving the problem

[0007] The above objective is achieved, according to the present invention, by a collection unit that collects sensor data and location data from a plurality of facilities; a detection unit that inputs the sensor data, the location data, and metadata of the facilities into an artificial neural network model to detect abnormal signs and identify abnormal facilities; a visualization unit that visualizes the operating status and type of abnormality of the abnormal facilities; and a guide unit that provides a response guide according to the type of abnormality of the abnormal facilities to an operator terminal.The system includes a control unit that receives a signal from the operator terminal and corrects the control parameters of the abnormal equipment, and the detection unit includes a learning module that inputs the sensor data, the location data, and metadata of the equipment into the artificial neural network model to learn the range of normal sensor values ​​according to the locations of the equipment; a judgment module that identifies the abnormal equipment based on the range of normal sensor values ​​and the sensor data to determine the type and degree of the abnormality; and a ranking module that generates ranking information regarding the abnormal equipment that requires priority action based on the type of abnormality, the degree of the abnormality, and environmental factors specific to the equipment, wherein the judgment module determines whether equipment adjacent to the abnormal equipment deviates from the range of normal sensor values ​​to track the path of abnormality diffusion, and the environmental factors specific to the equipment include the introduction price of the equipment, the number of sensor data generated by the equipment, the presence or absence of backup equipment capable of replacing the equipment, and the frequency of worker access by zone according to a preset radius centered on the location of the abnormal equipment, wherein the frequency of worker access by zone is defined as follows: a zone less than 50m formed in a concentric shape according to the radius distance centered on the abnormal equipment is designated as Zone 1, and a zone 50m or more This is achieved by an AI-based BIM smart maintenance system that divides an area of ​​less than 100m into a second zone and an area of ​​100m or more into a third zone, calculates based on frequency and distance weights of workers who have passed through the first zone, the second zone, and the third zone during a preset time, and the visualization unit is characterized by visually representing the abnormal diffusion path in a three-dimensional space according to the ranking information, and the control unit includes a verification module that, upon receiving a signal from the operator terminal, determines whether the signal complies with preset rules and calculates verification information, and a correction module that receives the verification information and corrects the control parameters of the abnormal equipment.

[0008] delete

[0009] delete

[0010] delete

[0011] delete Effects of the invention

[0012] According to the present invention, an artificial neural network model learns not only sensor data but also the location data of the equipment to autonomously derive a normal range of values ​​that reflects the environmental characteristics of each location. Through this, subtle abnormal signs that are difficult to detect using a simple threshold comparison method can be accurately identified at an early stage, and the reliability of the system can be ensured by minimizing false alarms.

[0013] In addition, by analyzing the correlation between abnormal equipment and adjacent equipment to track the diffusion path through which abnormal signs are transmitted and visualizing this in a three-dimensional space, it provides intuitive information to operators. This allows for the prevention of major accidents where the failure of a single piece of equipment leads to a chain shutdown of the entire system, and enables the effective confinement of the accident scope.

[0014] In addition, by correcting control parameters through a verification module that determines whether control signals received from the operator terminal comply with pre-set rules, malfunctions or secondary failures caused by human error can be fundamentally prevented regardless of the operator's proficiency. This provides an environment where even operators lacking specialized knowledge can safely respond to equipment anomalies.

[0015] In addition, it immediately provides operators with customized response guides based on the type of anomaly and significantly reduces the time required for initial response by sharing real-time operating status through a visualization unit. As a result, it maximizes the efficiency of equipment maintenance operations and reduces management costs.

[0016] Meanwhile, the effects of the present invention are not limited to those mentioned above, and various effects may be included within the scope obvious to a person skilled in the art from the contents described below. Brief explanation of the drawing

[0017] FIG. 1 illustrates an overall BIM-based smart maintenance system utilizing AI according to an embodiment of the present invention, and FIG. 2 illustrates the connections between the components of the detection unit of FIG. 1, and FIG. 3 illustrates the connections between the components of the control unit of FIG. 1, and Figure 4 illustrates real-time sensor data such as power, temperature, humidity, and flow rate, and Figure 5 illustrates the point focusing and evacuation route when a fire alarm occurs. Specific details for implementing the invention

[0018] Hereinafter, some embodiments of the present invention will be described in detail with reference to the exemplary drawings. It should be noted that in assigning reference numerals to the components of each drawing, the same components are given the same reference numeral whenever possible, even if they are shown in different drawings.

[0019] In addition, when describing embodiments of the present invention, if it is determined that a detailed description of related known configurations or functions would hinder understanding of the embodiments of the present invention, such detailed description is omitted.

[0020] In addition, terms such as first, second, A, B, (a), (b), etc., may be used when describing the components of the embodiments of the present invention. These terms are used merely to distinguish the components from other components, and the essence, order, or sequence of the components is not limited by the terms used.

[0021] Additionally, in the specification, the singular form includes the plural form unless specifically stated otherwise in the text. As used in the specification, “comprising” and “comprising” do not exclude the presence or addition of one or more other components in addition to the mentioned components.

[0022] Additionally, in describing embodiments of the present invention, each "part," "module," or "step" may be implemented through a processor and memory. The term processor should be broadly interpreted to include general-purpose processors, central processing units (CPUs), microprocessors, digital signal processors (DSPs), controllers, microcontrollers, state machines, and the like. In some environments, the term processor may refer to an application-specific integrated circuit (ASIC), a programmable logic device (PLD), a field programmable gate array (FPGA), and the like. The term processor may also refer to a combination of processing devices, such as, for example, a combination of a DSP and a microprocessor, a combination of multiple microprocessors, a combination of one or more microprocessors combined with a DSP core, or any other combination of such configurations.

[0023] Furthermore, memory should be broadly interpreted to include any electronic component capable of storing electronic information. Memory may also refer to various types of processor-readable media such as random access memory (RAM), read-only memory (ROM), non-volatile random access memory (NVRAM), programmable read-only memory (PROM), eraseable-programmable read-only memory (EPROM), electrically eraseable PROM (EEPROM), flash memory, magnetic or optical information storage devices, registers, etc. If a processor can read information from memory or write information to memory, memory is said to be in an electronic communication state with the processor, and each "part," "module," or "stage" may be implemented through a program or application based on the processor and memory in an electronic communication state.

[0025] Now, with reference to the attached drawings, a BIM-based smart maintenance system (100) utilizing AI according to one embodiment of the present invention will be described.

[0026] FIG. 1 illustrates an overall BIM-based smart maintenance system (100) utilizing AI according to an embodiment of the present invention, FIG. 2 illustrates the connections between the components of the detection unit (120) of FIG. 1, FIG. 3 illustrates the connections between the components of the control unit (150) of FIG. 1, FIG. 4 illustrates real-time sensor data such as power, temperature, humidity, and flow rate, FIG. 5 illustrates the focusing of the corresponding point and the evacuation route when a fire alarm occurs.

[0027] Referring to FIG. 1, a BIM-based smart maintenance system (100) utilizing AI according to one embodiment of the present invention is electrically connected to an operator terminal described later and may include a collection unit (110), a detection unit (120), a visualization unit (130), a guide unit (140), and a control unit (150).

[0028] First, the collection unit (110) collects data from multiple facilities and can be electrically connected to the detection unit (120) and the control unit (150).

[0029] The collection unit (110) can collect sensor data and location data from equipment installed within the facility on a BIM (Building Information Modeling, BIM) model.

[0030] Here, the term "equipment" may include mechanical / electrical devices included in piping, power, gas, and HVAC systems, and all digital objects that possess unique attribute information on a BIM model.

[0031] Here, sensor data refers to dynamic information converted into digital signals by measuring the operating status of equipment or the surrounding environment in real time. For example, it may include not only physical quantities such as temperature, humidity, pressure, and flow rate, but also electrical characteristic values ​​such as voltage, current, and leakage current, as well as information on the on / off status of the equipment.

[0032] Here, location data refers to the geometric location occupied by equipment within a facility, and may include absolute coordinates (such as GPS) or relative coordinate values ​​within the building, as well as floor and zone information defined in the BIM model.

[0033] Referring to FIG. 4, the collection unit (110) integrates and provides data in the form of a time-series dashboard by linking with heterogeneous sensor data and location data, such as power consumption, supply temperature, flow rate and humidity in the pipe, and can provide basic data for analyzing the correlation of each data (e.g., the relationship in which power consumption increases as the temperature rises). Through this, beyond simple monitoring of individual facilities, it is possible to determine the energy balance of the entire facility system based on location data in real time.

[0034] According to the above-described collection unit (110), by synchronizing and collecting sensor data and location data on a single platform, it has the effect of providing basic data for high-precision detection of abnormal signs and prediction of spatial diffusion paths that reflect the correlation between complex equipment within the facility.

[0035] Next, the detection unit (120) receives data from the collection unit (110) and detects whether there is an abnormality in the equipment using an artificial neural network, and can be electrically connected to the collection unit (110) and the visualization unit (130).

[0036] Referring to FIG. 2, the detection unit (120) may include a learning module (121) that learns a range of normal sensor values ​​(hereinafter, normal numerical range) according to the location of the equipment, a judgment module (122) that identifies abnormal equipment, and a ranking module (123) that calculates ranking information regarding maintenance priority.

[0037] First, the learning module (121) can learn the normal numerical range according to the location of the equipment by utilizing the sensor data and location data transmitted from the collection unit (110) together with the equipment's metadata as input to the artificial neural network.

[0038] Here, an artificial neural network refers to any form of machine learning algorithm system that learns collected data to extract specific patterns or perform predictions, and is not limited to the type of model such as CNN, RNN, LSTM, GNN, or the implementation method such as cloud or edge computing.

[0039] Here, metadata refers to static attribute information that is predefined in objects within a BIM model, and may include not only basic attribute information such as the manufacturer, model name, installation date, lifespan, and past maintenance history of the equipment, but also connection attribute information including vertical and horizontal connection relationships between equipment, and piping and wiring system information extracted from the topology of the BIM model.

[0040] Additionally, the learning module (121) can learn by processing location data, sensor data, and metadata into a computationally capable matrix form. Specifically, the learning module (121) can generate graph data composed of multiple nodes and edges by extracting vertical / horizontal connection relationships between facilities and topology information (topology) included in the metadata. At this time, the 3D location coordinates of the facilities included in the location data and the basic attribute information of the metadata (facility specifications, lifespan, etc.) can be quantified as static feature vectors of the nodes, and the physical connection relationships between facilities and the direction of energy flow can be quantified as an adjacency matrix including the weights of the edges. This can improve the precision of technical implementation by reflecting the actual physical system structure on the BIM model into the learning model, going beyond simple coordinate-centered analysis.

[0041] Next, the learning module (121) can map real-time sensor data of each facility collected during a preset time window to a dynamic feature vector of the corresponding node. At this time, multiple heterogeneous sensor data with different units and scales, such as mechanical physical quantities like temperature, humidity, pressure, and flow rate, and electrical characteristic values ​​like voltage and current, can be normalized according to preset standards and then merged into a single multi-dimensional feature vector for mapping.

[0042] Additionally, the artificial neural network of the learning module (121) may include an autoencoder structure that combines a Graph Convolutional Network (GCN) layer and a time series recurrent neural network (e.g., LSTM, GRU) layer. The Graph Convolutional Network layer can extract topological features based on metadata and location data by performing a weighted sum of the multidimensional feature vector of a specific facility and the multidimensional feature vectors of adjacent facilities physically connected thereto based on an adjacency matrix. These topological features can be input into the time series recurrent neural network layer to be encoded into a latent variable that reflects the temporal change pattern of the sensor data, and can be reconstructed into a heterogeneous sensor prediction value for each facility at that time point through a decoder.

[0043] Through this, not only changes in a single physical quantity (e.g., the effect of a change in flow rate of a specific pipe on the pressure of a downstream pipe) but also complex cross-correlations can be mathematically modeled, including the effect of an increase in mechanical load (temperature / pressure) of an air conditioning system on electrical characteristics (power consumption and overcurrent) based on the connection relationships between facilities defined in the metadata.

[0044] Additionally, the learning module (121) can optimize the parameters of the artificial neural network so that the residual (reconstruction error) between the actual observed multidimensional sensor data and the predicted value reconstructed through the decoder is minimized by using energy flow data collected during the normal operation of the facility as input.

[0045] Additionally, the learning module (121) can calculate the statistical distribution (e.g., mean and standard deviation) of the residuals output through an artificial neural network for newly input sensor data after the learning is completed. At this time, the learning module (121) can individually calculate a threshold, i.e., a 'normal value range,' for each of the multidimensional sensor data based on the standard deviation (σ) of the statistical distribution for each node (equipment) in order to detect complex abnormal energy usage patterns such as overcurrent, backflow, and leakage.

[0046] Next, the judgment module (122) can detect abnormal signs based on the normal numerical range calculated by the learning module (121) to identify abnormal equipment, and generate state classification information including the degree of abnormality, the type of abnormality, and unique identification information of the abnormal equipment.

[0047] Here, an abnormal sign refers to a precursor phenomenon in which real-time mapped sensor data deviates from a learned normal value range and may lead to a failure, and an abnormal facility may refer to a specific node object on the graph that requires inspection, correction, or control when such an abnormal sign is detected.

[0048] Here, the degree of the above refers to an indicator that quantifies the risk level of the equipment. Specifically, the judgment module (122) can calculate a statistical distance (e.g., Mahalanobis distance or Z-score) between the real-time residual derived from real-time sensor data passing through an artificial neural network (autoencoder) and the mean of the normal distribution calculated by the learning module (121). The judgment module (122) can calculate the magnitude of the deviation to create a step-by-step risk level indicator, such as classifying the calculated statistical distance (deviation value) as 'caution' if it is at a level of 1 to 2 times the standard deviation (σ) of the normal distribution, 'warning' if it is at a level of 2 to 3 times, and 'danger' if it exceeds 3 times.

[0049] Here, the type of anomaly refers to an item that classifies the cause of the anomaly and the nature of the necessary measures. Specifically, the judgment module (122) can identify the type of anomaly through feature contribution analysis, which quantifies the proportion of the reconstruction error of a specific physical quantity among the heterogeneous sensor data constituting the multidimensional feature vector in the total sum of residuals.

[0050] For example, the judgment module (122) can classify 1) the sum of residuals of the flow rate and pressure data of the pipe during a preset first time window as a 'state requiring physical emergency measures (leakage, etc.)' when the sum of residuals exceeds a preset first threshold, and

[0051] 2) If the rate of change of the time-series residual of the temperature data (slope) remains in a positive (+) direction during a pre-set second time window, and at the same time the residual change of the current data changes in tandem with a statistical correlation coefficient threshold (e.g., 0.7) or higher, this can be classified as a 'state requiring systemic parameter correction (overload, etc.),' and

[0052] 3) If the residuals of adjacent node facilities connected in the adjacency matrix are maintained within the normal numerical range, but a pattern is detected in which only the residual of a single physical quantity corresponding to a specific single facility deviates discontinuously by exceeding the second threshold within a pre-set third time window, this can be classified as a 'state suspected of facility aging and sensor error'.

[0053] Here, the unique identification information of an abnormal facility refers to the physical and logical attribute information of the specific abnormal facility identified in conjunction with object metadata on the BIM model. Specifically, the unique identification information may include the name, model name, and manufacturer of the facility corresponding to the node where anomalies were detected, as well as detailed specifications, installation date, and past maintenance history recorded in the metadata. It may also include a list of essential replacement parts and inspection tool information derived by matching the maintenance manual data of the facility, thereby guiding the operator to prepare necessary materials in advance for on-site inspections.

[0054] Additionally, the judgment module (122) can track the path of abnormal diffusion by comparing sensor data of adjacent facilities physically or logically connected to the abnormal facility in real time and analyzing whether the adjacent facilities sequentially deviate from the normal numerical range. For example, the judgment module (122) can identify the initial source of the abnormality and the future direction of propagation by utilizing the adjacency matrix of the graph constructed by the learning module (121) to detect a spatiotemporal sequence pattern in which pressure and flow rate data of adjacent nodes located downstream in the phase structure following the directionality of the edge sequentially deviate from the normal range with a time difference when a leak (abnormality) occurs at a specific pipe node.

[0055] Next, the ranking module (123) can generate ranking information that determines the ranking of targets to be prioritized for action among a plurality of abnormal facilities by combining the status classification information received from the judgment module (122) and environmental factors for each facility.

[0056] Here, environmental factors by facility refer to a composite indicator designed to quantify the impact on the operational continuity, economic value, and safety regulations of a facility in the event of a failure.

[0057] For example, environmental factors specific to each piece of equipment include the introduction price (e.g., a weight of 1.0 for 10 million KRW or more, 0.6 for 10 million KRW or more but less than 100 million KRW, and 0.2 for less than 10 million KRW), the number of sensor data generated by the equipment (e.g., a weight of 0.9 for multidimensional measurement equipment with 10 or more parameters, such as power quality analyzers or complex gas detectors, and 0.3 for general measurement equipment with 1 to 9 parameters, such as temperature / humidity sensors or flow meters), the ratio of cumulative operating time to the equipment's design life (e.g., a weight of 0.9 if the ratio is 90% or higher, indicating an imminent replacement cycle, and 0.1 if the ratio is less than 20%), the expected path of anomaly diffusion (e.g., a weight of 1.0 if fluid supply to the entire building is interrupted upon equipment failure, and 0.1 if it affects only specific equipment), the availability of backup equipment capable of replacing the relevant equipment (e.g., a weight of 1.0 if there is no backup, and 0.3 if immediate replacement is possible due to the availability of backup equipment), and the occurrence of defects. It may include the possibility of casualties (e.g., weighting factor 1.0 for gas barriers and high-voltage substations, and 0.1 for general lighting equipment and simple ventilation fans).

[0058] In addition, the above-mentioned environmental factors for each facility may further include the frequency of worker access by zone according to multiple radii set around the location of the abnormal facility. Specifically, the ranking module (123) may divide the area within 50m into a first zone, the area between 50m and 100m into a second zone, and the area exceeding 100m into a third zone in a concentric circle shape according to the radius distance centered on the abnormal facility, and may calculate frequency data of workers staying or passing through each zone during a preset time based on real-time location data or historical data. The ranking module (123) may assign a higher weight to the environmental factors as the frequency of worker access increases, and as the frequency of passing through the first zone closest to the abnormal facility increases compared to the frequency of passing through the third zone which is relatively far away, thereby preemptively raising the maintenance priority of facilities that are highly likely to cause casualties in the event of an explosion or leak.

[0059] Additionally, the ranking module (123) can calculate priorities as 'high', 'medium', and 'low' by inputting the degree of abnormality included in the ranking information and environmental factors for each facility into a pre-set weighting algorithm, and can set the relative priority higher as the number of environmental factors for each facility considered for the same priority increases. This can be utilized as a trigger parameter for determining work priority in the visualization unit (130), guide unit (140), and control unit (150).

[0060] In addition, the ranking information calculated from the ranking module (123) can be used as a key parameter to determine the display priority of the visualization unit (130), the scenario generation order of the guide unit (140), and the interlock operation order of the control unit (150).

[0061] The above-described detection unit (120) quantifies the risk level above based on the results of learning the correlation between location data, sensor data, and metadata extracted from the BIM model, and deeply identifies the root cause and the path of chain spread, thereby assisting the user of the operator terminal in quickly and intuitively determining the optimal response measure within the golden time at the beginning of the accident.

[0062] Next, the visualization unit (130) visualizes the operating status and type of abnormality of the equipment based on the detection result of the detection unit (120), and can be electrically connected to the detection unit (120), the guide unit (140), and the operator terminal.

[0063] The visualization unit (130) can visualize the operating status and types of abnormalities through a user interface (UI) or user experience (UX) environment.

[0064] Here, the operating status and type of abnormality may refer to the real-time operating status of the equipment and specific attributes of the problem that occurred. Specifically, it may include real-time energy flow data based on location data generated by the collection unit (110) and the type of abnormality classified by the detection unit (120) (physical measures, systemic correction, suspected aging / error, etc.).

[0065] In addition, the visualization unit (130) can provide a three-dimensional digital twin interface based on a BIM model.

[0066] Specifically, the visualization unit (130) can implement BIM model-based metadata (connection relationships between facilities and system information) as attribute values ​​on a 3D model and receive abnormal diffusion path information tracked by the judgment module (122), and can perform a 'diffusion path simulation' that dynamically shows the process of defect transmission over time. Through this, the operator can three-dimensionally identify the range of abnormal signs spreading along the physical connection structure based on facility attribute values.

[0067] Additionally, the visualization unit (130) can map the location coordinates of the abnormal equipment identified by the detection unit (120) to the corresponding object on the BIM model and display them in a highlighted manner, and can convert the calculated ranking information and the type of abnormality into visual variables to variably control and output visual parameters such as the color, scale, and blinking cycle of the corresponding equipment object.

[0068] For example, the priority of action based on ranking information can be differentiated and displayed by applying a red color and a fast flashing cycle of 10 times per second for a 'high' grade, applying an orange color and a medium speed flashing cycle of 5 times per second for a 'medium' grade, and a fixed yellow light for a 'low' grade.

[0069] For example, depending on the type of abnormality, if 'physical emergency measures' are required, the equipment object is indicated by a solid line; if 'systemic parameter correction' is required, the logical control path connected to the equipment is indicated by a dotted line; and if 'aging and sensor errors' are suspected, the object is indicated by a dashed line, allowing the user of the operator terminal to intuitively recognize the nature of the abnormality.

[0070] Additionally, the visualization unit (130) can visually represent the abnormal diffusion path tracked by the detection unit (120) in a BIM-based three-dimensional space. Specifically, the visualization unit (130) can sequentially flash the abnormal diffusion path along the pipe and wiring structure (location data) defined by the collection unit (110) to support the three-dimensional identification of the extent of the accident's impact by overlaying the reverse flow or overload state of the energy flow with dynamic graphic objects such as arrows.

[0071] Referring to FIG. 5, the visualization unit (130) can automatically focus on the corresponding point when a specific abnormality occurs, such as a fire alarm, and at the same time visualize and provide the shortest evacuation path or the location of the main shut-off valve based on BIM location data.

[0072] Additionally, the visualization unit (130) can transmit detailed analysis information to an operator terminal through various paths including at least one of a dashboard-type widget, a pop-up alarm window, an AR (augmented reality) overlay screen, or a mobile push notification interface.

[0073] Here, an operator terminal refers to an entity that monitors the system and exercises control authority, and may include an administrator's portable terminal, a workstation at a control center, as well as a management server that issues automatic control commands according to configured rules.

[0074] According to the visualization unit (130) described above, by synchronizing the status classification information of the detection unit (120) with BIM-based 3D object attributes and outputting it as variable visual parameters, the user of the operator terminal can immediately determine the location of an anomaly, the severity of the accident, and the nature of the measures even in a complex facility environment, thereby eliminating blind spots in management and significantly increasing the accuracy of the response.

[0075] Next, the guide unit (140) generates a response guide in natural language according to the type of abnormality of the equipment based on the detection result of the detection unit (120) and provides it to the operator terminal, and can be electrically connected to the visualization unit (130) and the operator terminal.

[0076] The guide unit (140) can generate an optimal response scenario by combining the status classification information derived from the detection unit (120) and the facility location and topology information on the BIM model. Here, the response guide may refer to a comprehensive action plan that goes beyond text instructions and includes parameter correction values ​​for facility control, physical movement paths for on-site measures, inspection priorities, and augmented reality (AR)-based visual instructions.

[0077] Additionally, the guide unit (140) can generate a top priority action notification to the operator terminal in conjunction with the flashing signal of the visualization unit (130) when the ranking information calculated by the detection unit (120) is at a 'high' level. At this time, if multiple abnormal facilities are identified simultaneously, the guide unit can provide a list of priority actions to be processed immediately to the operator terminal in real time by updating it, taking into account the relative ranking information of each facility.

[0078] In particular, the guide section (140) can generate three differentiated customized response scenarios depending on the above types.

[0079] First, if it is classified as requiring physical emergency measures, the guide unit (140) can suggest immediate physical solutions, such as shutting off the valve of the facility, and at the same time analyze the spatial data of the BIM model to calculate an optimal path for an operator (or field worker) to reach or evacuate to the facility. Referring to FIG. 5, the path is synchronized with the evacuation path displayed in the visualization unit (130) and may include a route that allows safe and rapid access to the site by avoiding obstacles or spreading danger zones (e.g., leak chain spread points).

[0080] Next, if it is classified as requiring systemic parameter correction, the guide unit (140) can calculate specific parameter correction recommendation values ​​that can optimize the operating rate or load condition of the equipment. For example, if the energy flow in a specific section is unstable, a scenario for adjusting the valve opening / closing rate or changing the pump RPM can be presented to the operator terminal, and upon receiving an approval signal from the operator terminal, the control unit (150) described later can be used to induce automatic correction.

[0081] Next, if it is classified as suspected of equipment aging or sensor error, the guide unit (140) can provide a corresponding guide for the calibration method of the sensor or information on the replacement cycle of the part. In addition, it can virtually correct the error data based on the location data and sensor data of adjacent normal equipment and deliver a technical guide to the operator terminal to maintain the continuity of system operation until the sensor is replaced.

[0082] Additionally, if the guide unit (140) determines that a detected abnormal pattern has an adverse effect on the energy balance of the entire facility, it can generate an Optimal Energy Flow proposal scenario. This may include strategies to bypass energy loss sections based on the piping / wiring topology structure in BIM, or load balancing the load of load-concentrating equipment identified by the visualization unit (130) to adjacent equipment. Through this, the operator terminal can make optimal decisions to reduce carbon emissions and operating costs for the entire facility.

[0083] In particular, as illustrated in FIG. 5, in a situation where evacuation and physical emergency measures are required simultaneously, such as in the occurrence of a fire, the guide unit (140) may specify the 'emergency recovery target equipment' that must be removed as a top priority among the equipment located on the calculated evacuation or movement path and include it in the response guide. Specifically, the guide unit (140) may calculate a priority score for removal of each piece of equipment by comprehensively combining the following criteria: i) ease of physical detachment of the equipment, ii) price of the equipment relative to mass, iii) possibility of insurance coverage in case of damage due to fire, etc., and iv) judgment criteria regarding the used price (residual value) in the market, as well as v) whether there is a risk of data loss because core process data is stored in the local memory of the equipment but backup (synchronization) to a cloud server has not been completed, vi) lead time required for procuring replacement parts and on-site reinstallation in case of loss of the equipment, and vii) whether the equipment contains chemical substances that may cause a secondary chain explosion or toxic gas generation upon exposure to fire. The guide unit (140) can guide field workers to minimize the loss of high-value assets and core data within the scope of ensuring human safety by designating equipment whose export priority score exceeds a preset threshold as the equipment subject to emergency recovery and visually highlighting it through an operator terminal.

[0084] The aforementioned guide section (140) integrates and provides situation-specific customized response scenarios, such as an optimal safe movement path for physical measures, recommended parameter correction values ​​for systemic optimization, or virtual correction guidelines for sensor errors, based on the state classification information calculated by the detection section (120), thereby significantly increasing the precision of the initial response to an accident regardless of the operator terminal user's proficiency and securing the golden time to minimize damage to facilities.

[0085] Next, the control unit (150) corrects the control parameters of the abnormal equipment and can be electrically connected to the collection unit (110) and the operator terminal.

[0086] Referring to FIGS. 1 and 3, the control unit (150) may include a verification module (151) that calculates verification information by determining whether control parameters comply with pre-set safety rules and process thresholds when receiving a signal from an operator terminal that has verified the scenario of the guide unit (140), a correction module (152) that corrects the control parameters of the equipment in real time by receiving the verification information, and a response module (153) that collects and responds to change information from outside the facility.

[0087] First, the verification module (151) is a module that calculates verification information by determining whether the correction plan complies with pre-set safety operation guidelines, physical process thresholds, equipment overload prevention rules, etc. when an approval signal from an operator terminal for a response scenario proposed by the guide unit (140) is input, and can be electrically connected to the correction module (152) and / or response module (153) described later.

[0088] Here, the signal transmitted from the operator terminal refers to any form of data signal input for the final determination or approval of system operation, and may include not only hardware button signals but also mobile approval tokens, voice commands, and confirmation feedback based on logical operations. For example, when a correction plan is derived in the guide unit (140) to increase the rotational speed (RPM) of a specific pump or adjust the valve opening / closing rate for system parameter correction, and an approval signal from the operator is received, the verification module (151) can numerically evaluate compliance with the correction plan by calculating a multidimensional constraint equation including the physical maximum allowable RPM of the pump, the current motor temperature, and the endurance pressure of the connecting pipe.

[0089] Additionally, the verification module (151) can pre-simulate the physical impact on adjacent facilities when applying correction parameters. For example, the verification module (151) can mathematically calculate the expected pressure change amount or the amount of overload transfer to adjacent electrical lines connected to the edge when the opening / closing rate (correction parameter) of a specific valve is drastically reduced by utilizing the adjacency matrix of graph data defined in the learning module (121).

[0090] Specifically, the verification module (151) can convert the fluctuation range of the control parameter (e.g., valve opening / closing rate reduced from 100% to 30%) into an initial perturbation vector of the corresponding node. Subsequently, the verification module (151) can perform matrix multiplication operations on the initial perturbation vector with the adjacency matrix representing the physical connectivity and directionality between facilities, and the edge weight matrix reflecting physical limit capacities such as the friction coefficient of the piping or the resistance component of the line. Through this message passing-based state transition operation, the verification module (151) can numerically quantify and calculate the expected pressure increase (ΔP) or expected current increase (ΔI) to be applied to each adjacent facility by simulating the process in which a change in physical quantity at a single node (specific valve) propagates sequentially to downstream nodes along the topology of the graph.

[0091] At this time, if the calculated expected pressure or load exceeds the process threshold of the downstream facility, the verification module (151) can perform a safety interlock function of rejecting the approval signal or automatically clipping it down to a safety parameter value within the threshold to prevent secondary accidents caused by misoperation of the operator terminal or system error.

[0092] Furthermore, the verification information produced by the verification module (151) can be generated in the form of a control packet that includes not only simple approval status (True / False) data, but also specific control values ​​of finally approved or down-adjusted parameters, a control delay time to prevent secondary accidents, and a sequential control sequence between multiple facilities. The correction module (152), which will be described later, receives the verification information in the form of a control packet transmitted from the verification module (151) and applies actual control signals to the control devices (e.g., PLC, DDC) of the field facilities according to the contained control sequence and control values, thereby enabling the correction of control parameters safely and precisely without collisions or side effects, even within facilities with complex phase structures.

[0093] Next, the correction module (152) receives verification information in the form of a control packet that is finally approved or safely clipped down from the verification module (151), and updates the control parameters of the abnormal equipment in real time through a field control device (PLC / DDC, etc.), thereby allowing the corresponding correction to be reflected in the sensor data collected by the collection unit (110), and can be electrically connected to the verification module (151) and / or the corresponding module (152).

[0094] Additionally, the correction module (152) directly changes the physical operating state of the field equipment through an electrical control signal, and thereby can act as a feedback mechanism to induce the sensor values ​​and equipment attribute information on the BIM model read in real time from the collection unit (110) to be corrected within a normal range and re-collected.

[0095] Here, control parameters are variables that determine the operating method or output performance of the equipment, and include the set temperature, pump rotational speed (RPM), valve opening / closing rate, and power supply status.

[0096] Additionally, the correction module (152) can perform a differentiated correction process based on instructions contained in the control packet according to the above types classified by the detection unit (120).

[0097] First, if physical emergency measures are required, the correction module (152) can apply an immediate electrical control signal to the field controller (PLC / DDC) based on the control packet that has passed the safety interlock procedure of the verification module (151) with the highest priority. For example, in the event of a leak or fire detection, the main valve can be shut off with the highest priority according to the control sequence of the control packet so that the flow sensor data input to the collection unit (110) is immediately reduced, and the control parameters can be updated by sequentially opening the valve of the adjacent bypass pipe or activating the emergency power to prevent secondary damage.

[0098] Next, if systemic parameter correction is required, the correction module (152) can precisely adjust the parameters of the equipment by reflecting the final control value, which ensures safety by simulating the physical impact on adjacent equipment through the verification module (151). For example, when a control signal to adjust the inverter frequency (Hz) or rotational speed (RPM) of a pump is transmitted to a field controller (PLC) to relieve overload in a specific section, the correction module (152) can gradually update the RPM parameter in compliance with the control delay time included in the control packet so that a sudden change in water pressure (water hammering, etc.) does not occur in the downstream piping, thereby allowing the energy balance data of the entire facility detected by the collection unit (110) to settle within the normal range.

[0099] Next, if equipment aging and sensor errors are suspected, the correction module (152) can ensure the continuity of system operation by modifying software data offset values ​​or temporarily correcting the data mapping logic of the collection unit (110) instead of performing physical control through the field controller (PLC).

[0100] Additionally, the correction module (152) can perform closed-loop control by reconfirming real-time sensor data fed back through the collection unit (110) immediately after performing parameter updates, and finally verifying whether the correction result has fully returned to the normal value range learned by the detection unit (120) (e.g., within the standard deviation of the normal distribution).

[0101] Next, the response module (153) is a module that calculates the total amount of available energy within the facility based on external fluctuation information and performs energy distribution optimization for each facility based on ranking information, and can be electrically connected to the verification module (151) and / or the correction module (152).

[0102] Here, external fluctuation information may refer to environmental, policy, and economic change factors entering from outside the facility, separate from any abnormal signs within the facility itself. Specifically, external fluctuation information may include recommendations for energy usage restrictions due to power supply emergencies, carbon emission allowances under the carbon emissions trading system, and information on energy unit price fluctuations resulting from sharp increases in oil and natural gas prices. Additionally, it may include operational schedule information, such as events held within the facility or changes in density in specific areas, as well as information on sudden weather changes, such as cold waves or heatwaves.

[0103] Here, energy distribution optimization refers to a process of reallocating resources to ensure the operational continuity of a facility when available energy resources are limited or the load is concentrated in a specific area due to the external fluctuation information. Specifically, the response module (153) can control the facility by referring to the ranking information derived from the ranking module (123) to prioritize the distribution of available energy to core facilities with relatively high priority to maintain normal operation, and can limit or temporarily cut off energy consumption for lower-level facilities with relatively low ranking information.

[0104] Additionally, the response module (153) can calculate an optimal energy flow path by utilizing the adjacency matrix of the graph data constructed in the learning module (121) to load-balance or divert the load of an energy-restricted zone to facilities in adjacent zones. The result calculated in this way is transmitted to the verification module (151) to undergo a preliminary simulation of the physical impact on adjacent facilities, and then can be reflected in real-time as a control parameter of the actual field facilities through the correction module (152).

[0105] According to the above-described control unit (150), by pre-simulating the physical impact on adjacent facilities to generate a control packet with a safety interlock applied, and then gradually and sequentially updating the control parameters of the field facilities based on this, and finally verifying the results through data fed back from the collection unit (110), there is an excellent effect of preventing secondary accidents caused by operator error or sudden control and simultaneously securing energy efficiency and operational safety of the entire facility with a complex phase structure.

[0107] Although it has been described above that all components constituting an embodiment of the present invention are combined or operate in combination, the present invention is not necessarily limited to such embodiments. That is, within the scope of the purpose of the present invention, all components may be selectively combined in one or more ways to operate.

[0108] Furthermore, terms such as "include," "compose," or "have" described above, unless specifically stated otherwise, mean that the relevant component may be inherent; thus, they should be interpreted as allowing for the inclusion of additional components rather than excluding them. All terms, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the present invention pertains, unless otherwise defined. Commonly used terms, such as those defined in advance, should be interpreted in accordance with their meaning in the context of the relevant technology and should not be interpreted in an ideal or overly formal sense unless explicitly defined in the present invention.

[0109] Furthermore, the above description is merely an illustrative explanation of the technical concept of the present invention, and those skilled in the art to which the present invention pertains will be able to make various modifications and variations within the scope of the essential characteristics of the present invention.

[0110] Accordingly, the embodiments disclosed in this invention are intended to illustrate, not limit, the technical concept of the invention, and the scope of the technical concept of the invention is not limited by these embodiments. The scope of protection of this invention shall be interpreted by the claims below, and all technical concepts within an equivalent scope shall be interpreted as being included within the scope of rights of this invention. Explanation of the symbols

[0111] 100: AI-powered BIM-based smart maintenance system 110 : Collection section 120 : Detection unit 121 : Learning Module 122 : Judgment Module 123 : Ranking Module 130 : Visualization section 140 : Guide section 150 : Control unit 151 : Verification Module 152 : Calibration Module

Claims

Claim 1 A collection unit that collects sensor data and location data from multiple facilities; a detection unit that inputs the sensor data, the location data, and metadata of the facilities into an artificial neural network model to detect abnormal signs and identify abnormal facilities; a visualization unit that visualizes the operating status and type of abnormality of the abnormal facilities; and a guide unit that provides a response guide according to the type of abnormality of the abnormal facilities to an operator terminal.The system includes a control unit that receives a signal from the operator terminal and corrects the control parameters of the abnormal equipment; the detection unit includes a learning module that inputs the sensor data, the location data, and metadata of the equipment into the artificial neural network model to learn the range of normal sensor values ​​according to the locations of the equipment; a judgment module that identifies the abnormal equipment based on the range of normal sensor values ​​and the sensor data to determine the type and degree of the abnormality; and a ranking module that generates ranking information regarding the abnormal equipment that requires priority action based on the type of abnormality, the degree of abnormality, and environmental factors specific to the equipment; the judgment module is characterized by determining whether equipment adjacent to the abnormal equipment deviates from the range of normal sensor values ​​to track the path of abnormality diffusion; the environmental factors specific to the equipment include the introduction price of the equipment, the number of sensor data generated by the equipment, the presence or absence of backup equipment capable of replacing the equipment, and the frequency of worker access by zone according to a preset radius centered on the location of the abnormal equipment; the frequency of worker access by zone is defined such that a zone less than 50m formed in a concentric shape according to the radius distance centered on the abnormal equipment is designated as the first zone, and a zone 50m or more A BIM-based smart maintenance system utilizing AI, wherein the system divides an area of ​​less than 100m into a second zone and an area of ​​100m or more into a third zone, calculates based on frequency and distance weights of workers who have passed through the first zone, the second zone, and the third zone during a preset time, and the visualization unit visually represents the abnormal diffusion path in a three-dimensional space according to the ranking information, and the control unit includes a verification module that calculates verification information by determining whether the signal complies with preset rules upon receiving a signal from the operator terminal, and a correction module that receives the verification information and corrects the control parameters of the abnormal equipment. Claim 2 delete Claim 3 delete Claim 4 delete Claim 5 delete

Images