Container terminal yard handling equipment decision analysis system based on artificial intelligence

Abstract

The application discloses a container terminal transport equipment decision analysis system based on artificial intelligence, belongs to the technical field of decision analysis, and comprises a data acquisition module for collecting and preprocessing data of each transport equipment in real time, a fault knowledge graph construction module for constructing and maintaining a transport equipment fault knowledge graph, a time sequence feature extraction module for extracting multi-modal fusion deep features, a transport equipment state evaluation module for constructing an equipment health baseline model and calculating a health degree score and dividing an evaluation grade, a fault prediction and hierarchical early warning module for performing abnormal detection and combining knowledge graph reasoning fault reasons and hierarchical early warning, and an intelligent scheduling decision module for optimizing scheduling with multiple targets of work efficiency, equipment health damage and load balance. The application adopts the above system, realizes integrated intelligent management of transport equipment state evaluation, fault early warning and scheduling decision, and effectively improves the operation reliability and work efficiency of the container terminal transport system.

Description

Technical Field

[0001] This invention relates to the field of decision analysis technology, and in particular to an artificial intelligence-based decision analysis system for container terminal handling equipment. Background Technology

[0002] Container terminal operations are a core component of port logistics, with large double-cantilever rail-mounted gantry cranes, end-to-end interchange cranes, and unmanned horizontal intelligent transport vehicles being key equipment in the operational chain. Currently, the operation and maintenance and scheduling system of container terminals faces two major pain points: 1. The scheduling strategy lacks robustness. Existing job scheduling focuses on job efficiency and shortest path as the sole optimization goal, without considering the real-time health status of equipment and potential failure risks. Once core equipment suddenly fails, it can easily cause blockage of the entire job chain. The scheduling scheme has weak anti-interference ability and low level of intelligence in task allocation, and cannot adapt to dynamically changing equipment status and job requirements.

[0003] 2. The level of intelligence in equipment operation and maintenance is low. Existing status monitoring relies heavily on single sensor threshold alarms, which cannot capture long-term degradation trends of equipment. Fault warnings are delayed, and fault diagnosis and knowledge accumulation depend on human experience, resulting in low efficiency and poor accuracy in fault tracing. At the same time, there is a lack of real-time status assessment for each transportation status. Summary of the Invention

[0004] The purpose of this invention is to provide an artificial intelligence-based decision analysis system for container terminal handling equipment to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, this invention provides an artificial intelligence-based decision analysis system for container terminal transport equipment, comprising a data acquisition module, a transport equipment fault knowledge graph construction module, a time-series feature extraction module, a transport equipment status assessment module, a fault prediction and graded early warning module, and an intelligent scheduling decision module, wherein: The data acquisition module is used to collect multi-source operating data, operational data and visual condition data of various transportation equipment in real time, and to preprocess the collected data. The fault knowledge graph construction module is used to extract fault knowledge of transportation equipment based on historical fault maintenance data and a large model in the field of container equipment, to construct and maintain a fault knowledge graph of transportation equipment. The temporal feature extraction module is used to extract multimodal fusion deep features of each transportation equipment; The status assessment module for transportation equipment is used to assess the status of transportation equipment based on the multimodal fusion depth features. The fault prediction and graded early warning module is used to perform anomaly analysis based on real-time data collected from various transportation equipment, reason about the causes of anomalies based on the fault knowledge graph, and provide graded early warnings. The intelligent scheduling and decision-making module allocates tasks based on the operational requirements, combined with the current assessment level, fault risk level, and current workload of each evacuation equipment.

[0006] Preferably, the data acquisition module specifically includes a multi-source data acquisition unit, a data cleaning unit, and a data fusion unit, wherein: The multi-source data acquisition unit collects multi-source operating data, operational data, and visual condition data from various transportation equipment in real time. The multi-source operating data includes vibration amplitude, equipment temperature, motor current, operating speed, load pressure, and braking parameters; the operational data includes workload, operation duration, container load, start-stop frequency, and task completion rate; and the visual condition data includes equipment component wear, loosening, misalignment, on-site operating environment, and equipment posture condition. The data cleaning unit performs data preprocessing on the collected data, including outlier removal, missing value imputation, and duplicate value removal. The data fusion unit uses the system's unified timestamp as a benchmark to perform modal alignment on multi-source operational data, job data, and visual condition data, forming a structured multimodal fusion dataset.

[0007] Preferably, the fault knowledge graph construction module includes a large model extraction unit and a knowledge graph iterative construction and update unit, wherein: The large model extraction unit extracts fault entities, entity relationships, and attributes based on historical fault maintenance data of various transportation equipment using a large model in the container equipment field, and outputs standardized triplet data. The knowledge graph iterative construction and update unit stores the extracted triple data in a structured manner through a graph database. At the same time, it sets up an incremental update mechanism to update the fault knowledge graph.

[0008] Preferably, the temporal feature extraction module includes a multimodal temporal feature extraction unit and a cross-modal feature fusion unit, wherein: The multimodal time-series feature extraction unit uses a one-dimensional convolutional neural network to extract the time-domain and frequency-domain features of vibration signals, a long short-term memory network to extract the trend and periodic features of temperature, pressure and current data, a Transformer encoder to extract the operation mode features and load change features of operation data, and a three-dimensional convolutional neural network to extract the equipment appearance change features of visual operating conditions data. The cross-modal feature fusion unit uses a cross-modal attention mechanism to align features of each modality and a gating fusion mechanism to adaptively calculate fusion weights based on the confidence and information content of each modality feature, thereby generating multimodal fusion deep features.

[0009] Preferably, the evacuation equipment status assessment module includes a lifecycle segmentation unit, an equipment health baseline construction unit, a real-time status comparison unit, and a comprehensive assessment level output unit, wherein: The life cycle is divided into three stages: break-in, stable operation, and degradation, based on the equipment's operating mechanism and failure rate curve characteristics. The equipment health baseline construction unit selects health status data segments from the historical operating data of the equipment at each stage, uses a Gaussian mixture model to model the multimodal fusion deep features under the health status, constructs health baseline models for each transportation equipment at different life cycle stages, and regularly performs incremental updates to the baseline models. The real-time status comparison unit compares the current multimodal fusion deep features with the health baseline model, calculates the probability density value of the current device status and maps it to a health score, and analyzes the degree of deviation of each modal feature from the baseline and the trend of health status change. The comprehensive assessment level output unit divides the equipment status into four levels: healthy, alert, warning, and critical, based on the health score, and generates a comprehensive assessment report that includes the health score, assessment level, modal deviation analysis, and trend prediction.

[0010] Preferably, the fault prediction and graded early warning module includes an anomaly detection unit, a fault knowledge reasoning unit, and a graded early warning unit, wherein: The anomaly detection unit uses two methods in parallel to detect equipment anomalies based on real-time collected data: statistical anomaly detection and time-series prediction anomaly detection. It comprehensively determines the anomaly state through a voting mechanism or a weighted scoring mechanism and classifies the detected anomalies into patterns. The fault knowledge reasoning unit maps the detected abnormal phenomena to the fault knowledge graph, and performs reasoning based on the fault knowledge to obtain the fault reasoning result; The graded early warning unit is used to classify the early warning level into four levels: alert, attention, warning, and emergency, based on the results of anomaly detection and fault reasoning. Different levels adopt different early warning methods and notification scopes, and disposal suggestions are pushed along with the early warning information.

[0011] Preferably, the fault knowledge reasoning unit specifically involves: mapping the detected abnormal phenomena to fault phenomenon entities in the fault knowledge graph; traversing and reasoning along the derivation relationship edges in the knowledge graph starting from the entity; sorting the candidate fault causes using a path-based sorting algorithm; and outputting a list of candidate fault causes, their confidence levels, and corresponding fault solutions.

[0012] Preferably, the intelligent scheduling decision module includes a task requirement analysis unit, a device status aggregation unit, and a multi-objective optimization scheduling unit, wherein: The task requirement analysis unit extracts the task type, task priority, task quantity, work time limit, work location and equipment type requirements from the work task sheet, decomposes complex tasks into multiple sub-tasks and establishes the dependencies between sub-tasks. The equipment status aggregation unit obtains the assessment level of each evacuation equipment from the evacuation equipment status assessment module, the fault risk level and warning level of each evacuation equipment from the fault prediction and graded early warning module, and obtains the operating status, task queue length and expected idle time of each evacuation equipment to construct the equipment status vector. The multi-objective optimization scheduling unit aims to minimize the total completion time of all tasks, minimize the equipment health impairment caused by task allocation, and minimize the task load variance among various transportation equipment. It uses equipment type constraints, location constraints, health constraints, and time constraints as constraints, and employs a non-dominated sorting genetic algorithm to solve the multi-objective optimization problem. The unit outputs the optimal scheduling scheme and sends it to the control system of each transportation equipment for execution.

[0013] Therefore, the above-mentioned decision analysis system for container terminal handling equipment based on artificial intelligence, adopted in this invention, has the following beneficial effects: (1) This invention constructs multimodal fusion deep features through the temporal feature extraction module, and combines the life cycle division and Gaussian mixture model health baseline in the evacuation equipment status assessment module to realize the health assessment of each evacuation equipment throughout its entire life cycle; (2) Anomaly detection of the transportation equipment is carried out through multi-source data, and the abnormal state is comprehensively judged by statistical anomaly detection and time-series prediction anomaly detection, which reduces the failure rate and false alarm rate and effectively overcomes the limitations of traditional single sensor threshold alarm. (3) By constructing a multi-objective optimization mechanism that links health and scheduling, the robustness and anti-interference ability of the scheduling strategy are enhanced, and the risk resistance and continuous operation stability of the terminal collection and distribution system under dynamic and complex working conditions are significantly improved.

[0014] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of an AI-based decision analysis system for container terminal transport equipment, according to an embodiment of the present invention. Figure 2 This is a system framework diagram of an AI-based container terminal transport equipment decision analysis system according to an embodiment of the present invention. Detailed Implementation

[0016] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0018] Example like Figure 1-2 As shown, this invention provides an artificial intelligence-based decision analysis system for container terminal transport equipment, including a data acquisition module, a transport equipment fault knowledge graph construction module, a time-series feature extraction module, a transport equipment status assessment module, a fault prediction and graded early warning module, and an intelligent scheduling decision module, wherein: The data acquisition module is used to collect multi-source operational data, business data, and visual condition data from various evacuation equipment in real time, and to preprocess the collected data. The evacuation equipment includes large double-cantilever rail-mounted cranes, end-interactive transfer cranes, and unmanned horizontal intelligent transport vehicles.

[0019] The fault knowledge graph construction module is used to extract fault knowledge of distribution equipment based on historical fault maintenance data and a large model in the container equipment domain, and to construct and maintain a fault knowledge graph of distribution equipment. The large model in the container equipment domain refers to a large model based on an open-source general-purpose language model, using historical fault inspection reports of various distribution equipment at container terminals, equipment operation and maintenance manuals, industry fault standards, and sensor anomaly annotation data as fine-tuning datasets, and employing the LoRA low-rank adaptation method to complete domain-specific fine-tuning.

[0020] The temporal feature extraction module is used to extract multimodal fusion deep features of each transportation equipment.

[0021] The evacuation equipment status assessment module is used to assess the status of evacuation equipment based on multimodal fusion depth features.

[0022] The fault prediction and graded early warning module is used to perform anomaly analysis based on real-time data collected from various transportation equipment, reason about the causes of anomalies based on a fault knowledge graph, and provide graded early warnings.

[0023] The intelligent scheduling and decision-making module allocates tasks based on the operational requirements, combined with the current assessment level, fault risk level, and current workload of each evacuation equipment.

[0024] In this embodiment, the data acquisition module specifically includes a multi-source data acquisition unit, a data cleaning unit, and a data fusion unit, wherein: The multi-source data acquisition unit collects multi-source operating data, operational data, and visual condition data of various transportation equipment in real time. The multi-source operating data includes vibration amplitude, equipment temperature, motor current, operating speed, load pressure, and braking parameters; the operational data includes workload, operation duration, container load, start-stop frequency, and task completion rate; and the visual condition data includes wear, loosening, and misalignment of equipment parts, on-site operating environment, and equipment posture condition.

[0025] The data cleaning unit performs data preprocessing on the collected data, including outlier removal, missing value imputation, duplicate value removal, and data standardization.

[0026] The data fusion unit uses the system's unified timestamp as a benchmark to perform modal alignment on multi-source operational data, job data, and visual condition data, forming a structured multimodal fusion dataset.

[0027] In this embodiment, the fault knowledge graph construction module includes a large model extraction unit and a knowledge graph iterative construction and update unit, wherein: The large-scale model extraction unit extracts fault entities, entity relationships, and attributes based on historical fault and maintenance data of various transportation equipment using a large-scale model in the container equipment domain. It outputs standardized triplet data, including equipment entity-occurrence-fault phenomenon, fault phenomenon-cause-abnormal operating condition, fault-corresponding-maintenance plan, and equipment wear-association-operating parameters. For different representations of the same entity, semantic similarity calculation is used for entity alignment and merging. Redundant relationships are deduplicated, and contradictory relationships are selected based on confidence level comparison.

[0028] The knowledge graph iterative construction and update unit stores the extracted triple data in a structured manner through a graph database. At the same time, an incremental update mechanism is set up to synchronize the newly added operation and maintenance records and fault data every month, automatically completing knowledge extraction and graph iteration to ensure the timeliness of knowledge.

[0029] In this embodiment, the temporal feature extraction module includes a multimodal temporal feature extraction unit and a cross-modal feature fusion unit, wherein: The multimodal temporal feature extraction unit uses a one-dimensional convolutional neural network to extract the temporal and frequency domain features of vibration signals. The 1D-CNN network structure includes: one input layer, three convolutional layers with kernel sizes of 7, 5, and 3, a max pooling layer, and a fully connected layer.

[0030] Long Short-Term Memory (LSTM) networks were used to extract trend and periodic features from temperature, pressure, and current data. The LSTM network structure consisted of an input layer, two LSTM layers, a dropout layer, and a fully connected layer. Each LSTM layer had 128 units, and the dropout rate of the dropout layer was 0.2.

[0031] A Transformer encoder is used to extract job pattern features and load variation features from job operation data. The Transformer encoder includes a position encoding layer, a multi-head self-attention layer, a feedforward neural network layer, a layer normalization layer, and a fully connected layer.

[0032] A 3D convolutional neural network is used to extract the appearance change features of equipment from visual working condition data. The 3D convolutional neural network uses 16 frames as a processing unit, and after passing through 5 layers of 3D convolution and 3 layers of 3D pooling, it outputs a 128-dimensional visual temporal feature vector through a fully connected layer.

[0033] The cross-modal feature fusion unit employs a cross-modal attention mechanism to align features across modalities. A gated fusion mechanism adaptively calculates fusion weights based on the confidence and information content of each modal feature, generating multimodal fusion deep features. Specifically, during feature alignment, vibration features are used as queries, and temperature / pressure / current features are used as keys and values ​​to calculate cross-modal attention weights, achieving alignment between vibration modes and other modes. The fusion weights are dynamically calculated through two fully connected layers.

[0034] In this embodiment, the evacuation equipment status assessment module includes a lifecycle segmentation unit, an equipment health baseline construction unit, a real-time status comparison unit, and a comprehensive assessment level output unit, wherein: The lifecycle of transportation equipment is divided into three phases based on its operating mechanism and failure rate curve characteristics: break-in phase, stable operation phase, and degradation phase. Break-in phase: The initial stage after the equipment is put into operation. During this stage, the failure rate is relatively high but shows a downward trend. The main failure types are installation and commissioning problems and initial wear of components. Stable operation phase: The normal operation phase. During this stage, the failure rate is low and relatively stable. The main failure type is random failure. Degradation phase: The later stage of equipment operation. During this stage, the failure rate gradually increases. The main failure types are fatigue, wear, aging, and other degradation-related failures.

[0035] Based on historical failure rate data and maintenance records of the equipment, a change point detection algorithm is used to automatically identify the boundary points of each stage. For new equipment lacking historical data, the average stage duration of equipment of the same model is used as the initial division basis, and is continuously corrected during operation.

[0036] The equipment health baseline construction unit selects health status data segments from historical operational data of the equipment at various stages as training samples for health baseline construction. A Gaussian mixture model is used to model the multimodal fusion deep features under health conditions, constructing health baseline models for each transportation equipment at different lifecycle stages, and periodically updating the baseline models incrementally. The health baseline model is as follows: ; In the formula, Represents multimodal fusion deep features in a healthy state. . For the first The mixing weights of Gaussian components, For the first The mean vector of each component Represented as the first The covariance matrix of each component, The number of Gaussian components.

[0037] The real-time status comparison unit compares the current multimodal fusion deep features with the health baseline model, calculates the probability density value of the device's current status and maps it to a health score, and analyzes the deviation of each modal feature from the baseline and the trend of health status changes. The comparison formula is as follows: ; The health score formula is: ; In the formula, Indicates the current modal characteristics. and To calibrate the parameters, fit them based on historical health or failure data.

[0038] The degree of deviation is calculated based on Mahalanobis distance, and the trend of change is shown. Using linear regression estimation The calculation of the short-term change slope, when When the value is less than -0.5, it is considered a rapid downward trend.

[0039] The comprehensive assessment level output unit divides the equipment status into four levels: healthy, alert, warning, and critical, based on the health score, and generates a comprehensive assessment report that includes the health score, assessment level, modal deviation analysis, and trend prediction.

[0040] The health score H∈[0,100] is divided into levels as follows: a health score ≥80 corresponds to Level I Healthy, 60≤health score<80 corresponds to Level II Attention, 40≤health score<60 corresponds to Level III Warning, and a health score<40 corresponds to Level IV Serious. When the health score shows a rapid downward trend, the assessment level is automatically raised by one level.

[0041] In this embodiment, the fault prediction and graded early warning module includes an anomaly detection unit, a fault knowledge reasoning unit, and a graded early warning unit, wherein: The anomaly detection unit, based on real-time collected data, employs both statistical anomaly detection and time-series prediction-based anomaly detection methods in parallel to detect equipment anomalies. It comprehensively determines the anomaly state through a voting mechanism or a weighted scoring mechanism and classifies the detected anomalies into patterns. Specifically: Based on statistical anomaly detection, the deviation of key parameters such as vibration, temperature, and pressure from historical normal values ​​is calculated. When the deviation exceeds the 3σ threshold, an anomaly marker is triggered.

[0042] Anomaly detection based on time-series prediction employs a time-series prediction model to perform multi-step advance prediction of key parameters. When the residual between the actual value and the predicted value exceeds a threshold, anomaly labeling is triggered. The time-series prediction model can be an LSTM prediction model.

[0043] The fault knowledge reasoning unit maps detected anomalies to fault phenomenon entities in the fault knowledge graph. Starting from the entity, it traverses and reasons along the derivation relationship edges in the knowledge graph. It uses a path-based sorting algorithm to sort the candidate fault causes and outputs a list of candidate fault causes, their confidence levels, and corresponding fault solutions.

[0044] The graded early warning unit is used to classify the early warning level into four levels: alert, attention, warning, and emergency, based on the results of anomaly detection and fault reasoning. Different levels adopt different early warning methods and notification scopes, and disposal suggestions are pushed along with the early warning information.

[0045] Warning (Level IV): A minor anomaly has been detected, but the specific cause of the fault has not been confirmed, or the confirmed cause of the fault has a minor impact. Warning method: The warning will be displayed on the system interface and will not trigger an alarm.

[0046] Attention (Level III Warning): A significant anomaly is detected, a possible cause of the fault is inferred with a confidence level ≥ 0.6, and the impact of the fault is limited to a single device. Warning method: Highlighted on the system interface and sent to device administrators via SMS.

[0047] Warning (Level II): A serious anomaly has been detected, and a fault cause with a high confidence level ≥ 0.8 has been deduced. The fault may cause equipment downtime or affect operational efficiency. Warning method: The system interface flashes an alarm, and a telephone voice notification is sent to equipment managers and dispatchers.

[0048] Emergency (Level I Warning): A critical anomaly is detected, such as safety device failure or impending breakage of a critical component. The exact cause of the failure is deduced, and the failure may lead to a safety accident or widespread operational disruption. Warning Method: The system issues a full-screen emergency alarm, simultaneously sending multi-channel emergency notifications to equipment managers, dispatchers, and safety managers, and automatically triggering equipment shutdown protection procedures.

[0049] In this embodiment, the intelligent scheduling decision module includes a task requirement parsing unit, a device status aggregation unit, and a multi-objective optimization scheduling unit, wherein: The task requirement analysis unit extracts the task type, task priority, task quantity, task time limit, task location and equipment type requirements from the task sheet, decomposes complex tasks into multiple sub-tasks and establishes the dependencies between sub-tasks.

[0050] The equipment status aggregation unit obtains the assessment level of each piece of transportation equipment from the transportation equipment status assessment module, the fault risk level and warning level of each piece of transportation equipment from the fault prediction and graded early warning module, and obtains the operating status, task queue length and expected idle time of each piece of transportation equipment to construct an equipment status vector. ; In the formula, The assessment levels are indicated as follows: 1 Healthy, 2 Caution, 3 Warning, 4 Serious. ∈{0,1,2,3,4} represents the fault risk level, where 0 indicates no risk and 4 indicates severe risk. Indicates the current job load. The length of the task queue. This represents the estimated free time.

[0051] The multi-objective optimization scheduling unit aims to minimize the total completion time of all tasks, minimize equipment health impairment caused by task allocation, and minimize the task load variance among various transportation equipment. It uses equipment type constraints, location constraints, health constraints, and time constraints as constraints, and employs a non-dominated sorting genetic algorithm to solve the multi-objective optimization problem. The optimal scheduling scheme is then output and distributed to the control systems of each transportation equipment for execution. The multi-objective optimization model is as follows: ; ; ; ; ; In the formula, For the task Completion time, Assigned to task The total number of TEUs (twenty-foot equivalent units) For equipment Execute the task The projected decline in health after the event is calculated based on a historical statistical regression model. Indicates task Assign to device As the first in its work sequence A sequence of positions.

[0052] Health constraints: Health scores of severe level and warning levels of emergency, warning and concern are not included in the allocation.

[0053] Location constraint: The task operation location is within the reach of the equipment.

[0054] Device type matching: =0 indicates that the device type does not meet the task requirements.

[0055] Time constraints: Execution is sequential, satisfying task release time and device availability time.

[0056] An electronic device includes a memory and a processor. The memory stores a computer program, and the processor invokes the computer program in the memory to implement the content of an artificial intelligence-based decision analysis system for container terminal transportation equipment.

[0057] A storage medium storing computer-executable instructions, which, when loaded and executed by a processor, realize the content of an artificial intelligence-based decision analysis system for container terminal handling equipment.

[0058] Therefore, the present invention adopts the above-mentioned decision analysis system for container terminal transportation equipment based on artificial intelligence, which realizes integrated intelligent management of transportation equipment status assessment, fault early warning and scheduling decision, effectively improving the operational reliability and efficiency of the container terminal transportation system.

[0059] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A decision analysis system for container terminal handling equipment based on artificial intelligence, characterized in that, It includes a data acquisition module, a knowledge graph construction module for transportation equipment faults, a time-series feature extraction module, a transportation equipment status assessment module, a fault prediction and graded early warning module, and an intelligent scheduling decision-making module, among which: The data acquisition module is used to collect multi-source operating data, operational data and visual condition data of various transportation equipment in real time, and to preprocess the collected data. The fault knowledge graph construction module is used to extract fault knowledge of transportation equipment based on historical fault maintenance data and a large model in the field of container equipment, to construct and maintain a fault knowledge graph of transportation equipment. The temporal feature extraction module is used to extract multimodal fusion deep features of each transportation equipment; The status assessment module for transportation equipment is used to assess the status of transportation equipment based on the multimodal fusion depth features. The fault prediction and graded early warning module is used to perform anomaly analysis based on real-time data collected from various transportation equipment, reason about the causes of anomalies based on the fault knowledge graph, and provide graded early warnings. The intelligent scheduling and decision-making module allocates tasks based on the operational requirements, combined with the current assessment level, fault risk level, and current workload of each evacuation equipment.

2. The decision analysis system for container terminal handling equipment based on artificial intelligence according to claim 1, characterized in that, The data acquisition module specifically includes a multi-source data acquisition unit, a data cleaning unit, and a data fusion unit, among which: The multi-source data acquisition unit collects multi-source operating data, operational data, and visual condition data from various transportation equipment in real time. The multi-source operating data includes vibration amplitude, equipment temperature, motor current, operating speed, load pressure, and braking parameters; the operational data includes workload, operation duration, container load, start-stop frequency, and task completion rate; and the visual condition data includes equipment component wear, loosening, misalignment, on-site operating environment, and equipment posture condition. The data cleaning unit performs data preprocessing on the collected data, including outlier removal, missing value imputation, and duplicate value removal. The data fusion unit uses the system's unified timestamp as a benchmark to perform modal alignment on multi-source operational data, job data, and visual condition data, forming a structured multimodal fusion dataset.

3. The decision analysis system for container terminal handling equipment based on artificial intelligence according to claim 2, characterized in that: The fault knowledge graph construction module includes a large model extraction unit and a knowledge graph iterative construction and update unit, wherein: The large model extraction unit extracts fault entities, entity relationships, and attributes based on historical fault maintenance data of various transportation equipment using a large model in the container equipment field, and outputs standardized triplet data. The knowledge graph iterative construction and update unit stores the extracted triple data in a structured manner through a graph database. At the same time, it sets up an incremental update mechanism to update the fault knowledge graph.

4. The decision analysis system for container terminal handling equipment based on artificial intelligence according to claim 2, characterized in that: The temporal feature extraction module includes a multimodal temporal feature extraction unit and a cross-modal feature fusion unit, wherein: The multimodal time-series feature extraction unit uses a one-dimensional convolutional neural network to extract the time-domain and frequency-domain features of vibration signals, a long short-term memory network to extract the trend and periodic features of temperature, pressure and current data, a Transformer encoder to extract the operation mode features and load change features of operation data, and a three-dimensional convolutional neural network to extract the equipment appearance change features of visual operating conditions data. The cross-modal feature fusion unit uses a cross-modal attention mechanism to align features of each modality and a gating fusion mechanism to adaptively calculate fusion weights based on the confidence and information content of each modality feature, thereby generating multimodal fusion deep features.

5. The decision analysis system for container terminal handling equipment based on artificial intelligence according to claim 4, characterized in that: The status assessment module for transportation equipment includes a lifecycle segmentation unit, an equipment health baseline construction unit, a real-time status comparison unit, and a comprehensive assessment level output unit, among which: The life cycle is divided into three stages: break-in, stable operation, and degradation, based on the equipment's operating mechanism and failure rate curve characteristics. The equipment health baseline construction unit selects health status data segments from the historical operating data of the equipment at each stage, uses a Gaussian mixture model to model the multimodal fusion deep features under the health status, constructs health baseline models for each transportation equipment at different life cycle stages, and regularly performs incremental updates to the baseline models. The real-time status comparison unit compares the current multimodal fusion deep features with the health baseline model, calculates the probability density value of the current device status and maps it to a health score, and analyzes the degree of deviation of each modal feature from the baseline and the trend of health status change. The comprehensive assessment level output unit divides the equipment status into four levels: healthy, alert, warning, and critical, based on the health score, and generates a comprehensive assessment report that includes the health score, assessment level, modal deviation analysis, and trend prediction.

6. The decision analysis system for container terminal handling equipment based on artificial intelligence according to claim 5, characterized in that, The fault prediction and graded early warning module includes an anomaly detection unit, a fault knowledge reasoning unit, and a graded early warning unit, wherein: The anomaly detection unit uses two methods in parallel to detect equipment anomalies based on real-time collected data: statistical anomaly detection and time-series prediction anomaly detection. It comprehensively determines the anomaly state through a voting mechanism or a weighted scoring mechanism and classifies the detected anomalies into patterns. The fault knowledge reasoning unit maps the detected abnormal phenomena to the fault knowledge graph, and performs reasoning based on the fault knowledge to obtain the fault reasoning result; The graded early warning unit is used to classify the early warning level into four levels: alert, attention, warning, and emergency, based on the results of anomaly detection and fault reasoning. Different levels adopt different early warning methods and notification scopes, and disposal suggestions are pushed along with the early warning information.

7. The decision analysis system for container terminal handling equipment based on artificial intelligence according to claim 6, characterized in that, The fault knowledge reasoning unit specifically maps detected abnormal phenomena to fault phenomenon entities in the fault knowledge graph. Starting from the entity, it traverses and reasons along the derivation relationship edges in the knowledge graph. It uses a path-based sorting algorithm to sort the candidate fault causes and outputs a list of candidate fault causes, their confidence levels, and corresponding fault solutions.

8. The decision analysis system for container terminal handling equipment based on artificial intelligence according to claim 7, characterized in that: The intelligent scheduling decision module includes a task requirement analysis unit, a device status aggregation unit, and a multi-objective optimization scheduling unit, wherein: The task requirement analysis unit extracts the task type, task priority, task quantity, work time limit, work location and equipment type requirements from the work task sheet, decomposes complex tasks into multiple sub-tasks and establishes the dependencies between sub-tasks. The equipment status aggregation unit obtains the assessment level of each evacuation equipment from the evacuation equipment status assessment module, the fault risk level and warning level of each evacuation equipment from the fault prediction and graded early warning module, and obtains the operating status, task queue length and expected idle time of each evacuation equipment to construct the equipment status vector. The multi-objective optimization scheduling unit aims to minimize the total completion time of all tasks, minimize the equipment health impairment caused by task allocation, and minimize the task load variance among various transportation equipment. It uses equipment type constraints, location constraints, health constraints, and time constraints as constraints, and employs a non-dominated sorting genetic algorithm to solve the multi-objective optimization problem. The unit outputs the optimal scheduling scheme and sends it to the control system of each transportation equipment for execution.

Images