A port equipment multi-working condition adaptive intelligent early warning method and system

Abstract

The present application belongs to the technical field of port equipment state monitoring, and relates to a port equipment multi-working condition adaptive intelligent early warning method and system. The system collects port equipment data through a data acquisition module, and performs preprocessing operation on the port equipment data to form a high signal-to-noise ratio data stream. A working condition identification module constructs a working condition identification model, and outputs the working condition type of the equipment and the corresponding time label. A dynamic threshold generation module generates a dynamic threshold value adapted to the current working condition. An abnormality early warning module identifies the current working condition type of the equipment, calls the corresponding dynamic threshold value, judges whether there is an abnormality in combination with the continuous change trend of the operating parameters, and outputs early warning information through a multi-channel mode. A threshold optimization module constructs a threshold optimization model, adjusts the weight parameters in the dynamic threshold value calculation, and performs iterative updating to realize adaptive optimization of the dynamic threshold value. The present application can effectively improve the early warning accuracy and the adaptability to complex working conditions and multi-scene operating conditions.

Description

Technical Field

[0001] This invention relates to the field of port equipment condition monitoring technology, specifically to a multi-condition adaptive intelligent early warning method and system for port equipment. Background Technology

[0002] Port equipment is the core support for port operations, and the stability of its operating status directly determines the port's production efficiency and operational safety. Therefore, it is crucial to achieve accurate and intelligent status monitoring and early warning.

[0003] In existing technologies, port equipment monitoring and early warning systems mostly adopt fixed thresholds or static model comparison methods, and the accuracy of early warning and adaptability to operating conditions need to be improved.

[0004] To this end, invention patent CN105737903A discloses an intelligent pre-diagnosis and intelligent early warning method for port machinery equipment faults, including: Step 1: Installing miniature current or voltage transformers, vibration sensors, and temperature sensors. Step 2: Setting the engineering parameters of the analog input module of the industrial controller. Step 3: Establishing a basic database by setting the equipment current, vibration, and temperature values ​​under different voltage and operating conditions during normal operation. Step 4: Reading the current, vibration, and temperature data of the equipment under different operating conditions in real time or at regular intervals. Step 5: Data that deviates from normal data or empirical values ​​after data sampling and analysis. Step 6: Maintenance personnel complete early on-site maintenance or repair based on the pre-diagnosis results. This method can pre-diagnose potential faults before key components of port machinery equipment fail or are damaged and shut down, and warn maintenance personnel whether the corresponding components need maintenance.

[0005] The above-mentioned technical solutions have made progress in using sensor data to achieve early warning, but the following technical problems still exist: (1) The warning threshold or judgment benchmark mainly relies on a pre-set static database and does not perform dynamic and refined threshold matching according to the specific working conditions of the equipment; (2) The impact of time dimension on equipment operation characteristics is not considered, and there is a lack of threshold adaptive adjustment mechanism based on time period, resulting in weak real-time scenario adaptability of early warning; (3) Lacking a closed-loop, continuous adaptive optimization mechanism, it is impossible to dynamically iterate and optimize the model based on the early warning feedback results and historical operating data, making it difficult to maintain high-precision early warning capabilities as the equipment life cycle evolves and operating conditions change.

[0006] In view of this, it is very necessary to provide a multi-condition adaptive intelligent early warning method and system for port equipment to solve the above-mentioned defects in the prior art. Summary of the Invention

[0007] The purpose of this invention is to solve the technical problems of low accuracy and poor adaptability to working conditions and scenarios in port equipment early warning due to static thresholds, lack of time dimension adaptation and closed-loop optimization mechanism. The invention provides a multi-working-condition adaptive intelligent early warning method and system for port equipment to solve the technical problems existing in the prior art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, the present invention provides a multi-condition adaptive intelligent early warning method for port equipment, comprising the following steps: Step S1: Data acquisition step, collecting port equipment data, and forming a high signal-to-noise ratio data stream after data preprocessing; Step S2: The step of working condition identification, which involves building a working condition identification model based on mechanistic features and data statistical features, and outputting the working condition type and time label in real time; Step S3: The step of generating dynamic thresholds involves calculating the initial threshold according to the working condition type, and then weighting and adjusting it according to the time period to obtain the dynamic threshold that adapts to the current operating state. Step S4: The abnormal warning step involves identifying the current operating condition and calling the corresponding threshold, judging the abnormality based on the continuous change trend, and pushing the warning information through multiple channels. Step S5: Optimize the dynamic threshold by building an optimization model, adjusting the threshold calculation weights, periodically iterating and optimizing the model, and adaptively updating the dynamic threshold.

[0009] Secondly, the present invention also provides a multi-condition adaptive intelligent early warning system for port equipment, comprising: The data acquisition module is used to collect port equipment data and perform preprocessing operations on the port equipment data to form a high signal-to-noise ratio data stream; The operating condition identification module constructs an operating condition identification model based on mechanistic features and data statistical features, analyzes the high signal-to-noise ratio data stream, and outputs the corresponding equipment operating condition type and time label corresponding to the equipment operating condition in real time. The dynamic threshold generation module is used to calculate the initial threshold based on the operating condition of the equipment, and to adjust the initial threshold by weighting it in combination with time period characteristics to generate a dynamic threshold that is adapted to the current operating condition. The anomaly warning module is used to identify the current operating condition of the equipment and call the corresponding dynamic threshold. It combines the continuous change trend of the operating parameters to determine whether there is an anomaly and outputs warning information through multiple channels. The threshold optimization module is used to build a threshold optimization model, adjust the weight parameters in the dynamic threshold calculation, and iteratively update the model according to a set period to achieve adaptive optimization of the dynamic threshold.

[0010] The modules work together to enable adaptive anomaly monitoring and intelligent early warning of port equipment under multiple operating conditions, thereby improving the accuracy of anomaly monitoring and adaptability to operating conditions.

[0011] The beneficial effects of this invention are as follows: This invention collects port equipment operating parameters, equipment operation commands, and position encoder data from all dimensions, and performs data preprocessing during the data formation process to reduce noise and invalid interference, improve data stability and reliability, and provide a high-quality data foundation for subsequent early warning judgments. This effectively improves the problem of low early warning accuracy caused by insufficient raw data quality.

[0012] Based on the mechanistic and statistical characteristics of port equipment, this invention distinguishes the operating status of the equipment in real time, outputs the corresponding operating condition type and introduces time tags, so that the early warning process can clearly distinguish different operating conditions and their time characteristics, thereby improving the adaptability to complex operating conditions and multi-scenario operating states, and avoiding misjudgments caused by using the same judgment standard for different operating conditions.

[0013] This invention calculates corresponding thresholds for different operating conditions and adjusts the thresholds by weighting them in conjunction with the operating time period. This allows the thresholds to change dynamically with changes in operating conditions and time evolution, overcoming the problem that traditional static thresholds cannot adapt to changes in equipment operating status, thereby improving the matching degree between anomaly identification results and actual operating status.

[0014] When making anomaly judgments, this invention calls the corresponding dynamic threshold based on the currently identified operating condition and performs a comprehensive analysis in conjunction with the continuous changing trend of operating parameters to achieve accurate identification and timely early warning of abnormal states, reduce occasional false alarms and missed alarms caused by single-point judgments, and improve the reliability and real-time performance of early warning results.

[0015] This invention continuously optimizes the threshold calculation process, adjusts relevant weights based on long-term changes in equipment operating data, and periodically updates the optimization model, enabling the threshold to continuously and adaptively adjust with changes in equipment operating status, forming a closed-loop optimization process. This solves the problem of existing early warning methods lacking adaptive updating and continuous optimization capabilities.

[0016] The method of the present invention can simultaneously take into account the differences in multiple operating conditions, time-varying characteristics, and long-term operational evolution patterns during the port equipment early warning process, thereby improving the accuracy of early warning and the adaptability to complex operating conditions and multiple scenarios.

[0017] Therefore, it is evident that the present invention has outstanding substantive features and significant progress compared with the prior art, and the beneficial effects of its implementation are also obvious. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0019] Figure 1 This is a flowchart of a multi-condition adaptive intelligent early warning method for port equipment; Figure 2 This is a schematic diagram of a multi-condition adaptive intelligent early warning system for port equipment. Detailed Implementation

[0020] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. The following embodiments are explanations of the present invention, but the present invention is not limited to the following implementation methods.

[0021] Example 1: like Figure 1 As shown in the figure, this embodiment provides a multi-condition adaptive intelligent early warning method for port equipment, which includes the following steps: Step S1: Data acquisition step, collecting port equipment data, and forming a high signal-to-noise ratio data stream after data preprocessing; Step S2: The step of working condition identification, which involves building a working condition identification model based on mechanistic features and data statistical features, and outputting the working condition type and time label in real time; Step S3: The step of generating dynamic thresholds involves calculating the initial threshold according to the working condition type, and then weighting and adjusting it according to the time period to obtain the dynamic threshold that adapts to the current operating state. Step S4: The abnormal warning step involves identifying the current operating condition and calling the corresponding threshold, judging the abnormality based on the continuous change trend, and pushing the warning information through multiple channels. Step S5: Optimize the dynamic threshold by building an optimization model, adjusting the threshold calculation weights, periodically iterating and optimizing the model, and adaptively updating the dynamic threshold.

[0022] In step S1: Port equipment data is collected through a distributed sensor network and connected to an edge gateway; the edge gateway performs data preprocessing operations and outputs a continuous and stable high signal-to-noise ratio data stream as input data for subsequent operating condition identification steps.

[0023] The port equipment data includes comprehensive operational data, equipment operation command data, and position encoder data. The comprehensive operational data includes lifting capacity, load current, vibration acceleration, rotational speed, vibration frequency, operating time, oil pressure, oil temperature, historical operational data, and real-time operational data. Lifting capacity is acquired using a high-precision tension sensor with a sampling frequency of 1Hz; load current is acquired using a Hall effect sensor with a sampling frequency of 1Hz; vibration acceleration is acquired using a triaxial accelerometer with a sampling frequency of 10kHz; and oil temperature is acquired using a thermocouple sensor with a sampling frequency of 1Hz. Historical operational data includes at least one complete operational cycle of valid data, along with equipment operation command data and position encoder data, reflecting the equipment's operational actions and displacement status.

[0024] The data preprocessing operations are as follows: After receiving port equipment data, the edge gateway performs unified access and parsing of port equipment data from different sources. It categorizes and manages the data according to data type and sampling frequency, ensuring that high-frequency and low-frequency data are processed according to their respective strategies. Noise reduction is performed using moving averages, low-pass filtering, or median filtering to weaken random noise and transient spikes. Simultaneously, missing or anomalously fluctuating data points are identified and removed to prevent outliers from interfering with subsequent analysis. The edge gateway performs time alignment processing on various data types based on a timestamp mechanism, mapping data from different sampling frequencies to a unified time axis to form data samples arranged in chronological order. For high-frequency data such as vibration acceleration, rotational speed, and vibration frequency, the edge gateway performs downsampling or feature preprocessing locally to reduce subsequent transmission and computational load. It then performs standardization processing, converting the data into a unified data structure and encapsulating and outputting a continuous high signal-to-noise ratio data stream according to a preset data model, providing stable and standardized data input for real-time analysis of subsequent operational condition identification models.

[0025] By utilizing distributed sensor networks and edge gateways to uniformly collect and preprocess multi-source operational data, the signal-to-noise ratio and temporal consistency of the collected data are effectively improved, and the impact of noise and abnormal fluctuations in the raw data on subsequent analysis is reduced, providing a reliable data foundation for the construction of operating condition identification models and dynamic threshold calculation.

[0026] In step S2: the high signal-to-noise ratio data stream is taken as input in chronological order, the mechanism features are extracted based on the mechanical structure and working principle of the equipment, and the working condition identification model is constructed by combining the data statistical features; the working condition mode of the equipment is automatically identified, and time dimension labels are introduced to record the time nodes of the working condition switching, and the corresponding equipment working condition type and the time label corresponding to the equipment working condition are output to provide time reference for subsequent threshold adjustment.

[0027] The operating modes include standby, low load, normal load, and high load, and support custom expansion of operating states (such as heavy load start-stop, no load movement, etc.) according to the type of port equipment (such as gantry crane, belt conveyor, quay crane) and actual operation requirements. Mechanistic characteristics are derived from the physical mechanism analysis of operating parameters based on the mechanical structure and working principle of port equipment. These characteristics reflect the essential relationships in the equipment's operating state, including load-speed correlation characteristics and vibration harmonic characteristics. The load-speed correlation characteristic is calculated by using the ratio of load current to equipment operating speed to determine the load-speed coupling coefficient. This describes the mechanical load characteristics and drive efficiency of the equipment under different loads, and is sensitive to distinguishing between low load, normal load, and high load conditions. Vibration harmonic characteristics are obtained by performing frequency domain analysis on the vibration acceleration signal to extract the fundamental frequency and harmonic amplitudes. The vibration acceleration signal is processed by Fast Fourier Transform or Wavelet Transform to obtain the fundamental frequency and corresponding harmonic amplitudes within each time window, reflecting the characteristics of the mechanical vibration of the equipment during operation, and is sensitive to load changes, mechanical wear, or fault conditions.

[0028] The mechanism feature extraction process is usually carried out in a sliding time window. Within each window, the load-speed correlation features and vibration harmonic features are calculated to generate a continuous feature vector, which provides physical mechanism support for working condition identification.

[0029] Data statistical characteristics are obtained through statistical calculations of equipment operating parameters over time series. They are used to describe the changing trends and fluctuation characteristics of these parameters, including the moving average, variance, and kurtosis. Moving average Variance represents the average level of a parameter within a certain time window, used to capture the overall trend of equipment operating status. The fluctuation range of the descriptive parameters reflects the stability of the operating load or vibration. The peak factor measures the degree of instantaneous spikes in the signal and is sensitive to sudden anomalies or transient impacts. Data statistical feature extraction is also performed using a sliding time window. The statistical calculation results within each window are used to form a feature vector, which is then combined with the mechanistic features to form a combined feature vector, which serves as the input for the subsequent operating condition identification model.

[0030] in, This represents the i-th sampled data point within the current sliding window, where n is the total number of data points within the sliding window. The window length can be set according to the sampling frequency and the expected feature smoothing requirements.

[0031] A working condition identification model is constructed based on a combination of mechanistic and statistical data feature vectors. This model employs the C4.5 decision tree algorithm, using the information gain ratio (IGR) as the node splitting criterion to evaluate the effectiveness of features in splitting classification nodes. The decision tree model can recursively partition and discriminate multi-dimensional features, outputting the current working condition mode of the equipment. It also supports extending to custom working conditions based on equipment type and operational requirements.

[0032] Historical equipment operating data is labeled into different operating condition categories to form a training dataset. During training, the operating condition recognition model selects the optimal split node by calculating the information gain ratio of each feature. After training, the accuracy of the operating condition recognition model is verified using test data. The information gain ratio is calculated using the following formula: ; Where Gain(A) is the information gain of feature A, representing the degree to which the uncertainty of the data is reduced after splitting the dataset using this feature, and SplitInfo(A) is the split information, representing the amount of information about the value distribution of feature A itself, used to normalize the information gain and avoid bias towards features with more values.

[0033] A higher information gain ratio indicates that the feature can significantly reduce the uncertainty of the dataset and has a reasonable value distribution, making it suitable as a basis for splitting decision tree nodes. In work condition identification, using the information gain ratio to select the optimal feature for splitting can effectively improve classification accuracy and ensure fair evaluation of multi-valued features, thereby achieving high-precision and low-latency work condition identification.

[0034] During the operation condition identification model discrimination process, after each model discrimination is completed, the timestamp of the corresponding time slice is bound to the discrimination operation condition type to generate a time dimension label. The time dimension label includes the start time node of the operation condition, the end time node of the operation condition, and the time period in which it is located. By comparing the operation condition types before and after in consecutive time slices, when the discrimination result is different from the operation condition of the previous time slice, it is determined that an operation condition switch has occurred, and the switch time node is automatically recorded to form a complete time dimension label.

[0035] The operating condition identification model can output the corresponding equipment operating condition type and the time dimension label corresponding to that operating condition, providing data basis for subsequent threshold generation and early warning.

[0036] In step S3: For the output equipment operating condition type, based on the equipment's historical operating data, statistical analysis and machine learning algorithms are used to identify the normal fluctuation range of each parameter under different operating conditions, and determine the initial threshold range of each parameter; a time period factor is introduced, and the period is set according to day, week, month, quarter, year. Each period can specify a specific time period. Combined with the equipment's historical operating data within the corresponding time period, the initial threshold range is weighted and optimized to generate dynamic thresholds under each operating condition, so as to realize the adaptive adjustment of thresholds for operating characteristics in different time periods.

[0037] For example, the time periods of the daily cycle can be set as 8:00-16:00 for the morning shift, 16:00-24:00 for the evening shift, and 0:00-8:00 for the night shift. The historical operating data of the equipment within the daily cycle time period shows that the vibration is higher at the beginning of the morning shift and the load is lower on weekends. The threshold range of the corresponding time period needs to be adjusted accordingly.

[0038] Dynamic thresholds include low reporting, low-low reporting, high reporting, high-high reporting thresholds, and automatic release thresholds. The dynamic threshold generation process follows the proximity principle, that is, the weight of recent historical data is higher than that of distant data, to ensure that the threshold matches the current operating status of the equipment.

[0039] The process of determining the initial threshold range is as follows: for each equipment operating condition type, based on the mean μ and standard deviation σ of the parameter, the normal fluctuation range of the parameter is determined using the 3σ principle, and the formula is: .

[0040] For example, the average vibration acceleration of a certain model of gantry crane under normal load is μ=0.85g, the standard deviation is σ=0.12g, and the initial threshold range of vibration acceleration is [0.49g, 1.21g].

[0041] Introducing dual-weighting coefficients to perform time-weighted optimization on the initial threshold can be described by the following formula: The dual-weighting coefficients include weights for recent data. and time period weight ; This is the ratio of the duration of recent data to the total duration of historical data.

[0042] Taking the normal load condition of the morning shift as an example, the weight of the time period is set to 1.1 due to the startup impact, while the weight of the time period for other time periods is set to 1.0. Substituting into the formula, the dynamic threshold of the normal load condition of the morning shift is [0.15g, 0.32g], which is further subdivided into low reporting and high reporting thresholds for different levels of anomaly detection.

[0043] By dynamically generating thresholds, the thresholds can be adapted to different operating conditions and time periods of the equipment, avoiding false alarms or missed alarms caused by fixed thresholds. By introducing time weights and the principle of prioritizing recent data, the current operating status and historical patterns of the equipment can be reflected, providing accurate threshold support for subsequent anomaly monitoring and early warning, and improving the accuracy of early warning and the reliability of system response.

[0044] In step S4: real-time collected port equipment full-dimensional operating parameter data, equipment operating condition type, and time label are used as input. According to the current equipment operating condition type, the corresponding dynamic threshold range is matched. The monitoring platform compares the real-time parameters with the dynamic threshold and performs point-by-point comparison and trend analysis on the real-time parameters. When the real-time parameters exceed the threshold range, a comprehensive judgment is made by introducing continuity criteria and abrupt change amplitude criteria. After filtering out short-term disturbance signals, an alarm is triggered, and an abnormal warning result of the corresponding level is generated. The abnormal warning result is pushed to the operation end through multiple channels such as SMS, mobile mini-program, and platform pop-up window. At the same time, the warning trigger time, current equipment operating condition type, abnormal parameter value, and other information are recorded. The operation end performs maintenance and records the operation and maintenance verification results and equipment maintenance records.

[0045] The continuity criterion is used to determine whether parameter anomalies have persistent characteristics. When real-time parameters exceed the dynamic threshold range of the corresponding equipment operating condition type multiple times in a row, the parameter status is considered to meet the abnormal triggering conditions, thus avoiding misjudgment caused by single sampling noise or instantaneous fluctuations.

[0046] The abrupt change magnitude criterion is used to measure the drasticness of parameter changes. Based on the continuity criterion, it further determines whether the parameter's change magnitude within a set time window exceeds a preset proportional threshold. Specifically, this is done by comparing the relative change between the current sampled value and the sampled value at the start of the time window; when the abrupt change is satisfied... At that time, the changes in the judgment parameters exhibit a sudden characteristic. Among them, t is the length of the time window (e.g., 5 seconds). This is the threshold for the magnitude of the mutation, used to distinguish between normal fluctuations and abnormal mutations.

[0047] By combining continuity and abrupt change criteria, short-term disturbance signals can be effectively filtered out. Disturbances such as power grid fluctuations and instantaneous impact loads typically exhibit short durations of exceeding limits, few instances of exceeding limits, or changes lacking continuity and suddenness. Their sampled data rarely simultaneously meet the conditions of multiple consecutive exceedances and abrupt changes reaching a threshold. Therefore, when parameters only show sporadic exceedances or insufficient variation, the system classifies them as short-term disturbances, does not generate abnormal warnings, and reduces the false alarm rate.

[0048] When real-time parameter changes simultaneously meet both the continuity criterion and the abrupt change magnitude criterion, an anomaly warning result is generated. The anomaly warning result includes true anomaly / false alarm / missed alarm, anomaly parameter name and corresponding value, current equipment operating condition type, timestamp, trigger threshold level, warning trigger time and duration, and an accompanying curve of the real-time parameter change within a preset time window. This is used for multi-channel warning push and historical record storage, providing a basis for operation and maintenance decisions and subsequent threshold optimization.

[0049] For example, during monitoring, the real-time collected operating parameters of port equipment across all dimensions are continuously compared with the dynamic thresholds corresponding to the current equipment operating condition type. When the same real-time parameter exceeds the corresponding high or low reporting threshold in three consecutive samples, and the change in the parameter reaches or exceeds a preset proportion of 30% within a five-second time window, the parameter change is determined to meet the abnormal triggering condition, triggering an alarm and generating an abnormal warning result. Taking the gantry crane vibration acceleration as an example, when the consecutive sampled values ​​of this parameter are 0.31g, 0.33g, and 0.32g, all exceeding the corresponding 0.30g high reporting threshold, an abnormal warning is triggered, and warning information is pushed to the operation terminal via SMS and mobile mini-program. At the same time, the corresponding equipment operating condition label and real-time parameter change curve are recorded for subsequent analysis and traceability.

[0050] By introducing continuous over-limit criteria and trend discrimination mechanisms, the probability of false alarms caused by instantaneous interference can be effectively reduced, and stable identification of abnormal equipment status can be achieved. At the same time, by combining multi-level thresholds to output different levels of early warning information, it is beneficial for the operation and maintenance end to quickly judge the severity of the anomaly and take targeted measures, thereby improving the accuracy and response efficiency of port equipment anomaly monitoring and providing reliable support for predictive maintenance.

[0051] In step S5: a warning feedback dataset containing warning results, operation and maintenance verification results, and equipment maintenance records is constructed; based on the warning feedback dataset, a threshold optimization model is constructed using a gradient boosting algorithm, and various weight coefficients involved in the dynamic threshold calculation process are updated and calculated; the model is iteratively trained according to a preset optimization cycle and convergence conditions, the updated weight parameters are output, and the weight parameters are fed back to the dynamic threshold generation process to generate new dynamic thresholds, thereby achieving adaptive updating of the threshold.

[0052] An optimization model is constructed using the gradient boosting algorithm, with the early warning feedback dataset as the training sample. The weight coefficients are adjusted using the gradient boosting algorithm, as shown in the formula: Where η is the learning rate, and L is the loss function based on the false positive rate and the false negative rate. The weighting coefficients are set before the update. A 30-day optimization cycle weighting coefficient is set, which includes historical data time weight, operating condition characteristic weight, and abnormal parameter weight. A convergence condition is set: when the false alarm rate fluctuation does not exceed 2% within two consecutive optimization cycles, the current weighting parameters remain unchanged, and the threshold optimization model is continuously iterated. When the equipment enters a new stage of its lifecycle, such as the aging period or when operating condition characteristics change significantly, such as when the load type is adjusted, the model automatically triggers an emergency optimization process to quickly adapt to the new operating state and continuously reduce the false alarm rate and missed alarm rate.

[0053] Step S5 avoids the accuracy decline caused by long-term fixed thresholds, improves the overall early warning accuracy, reduces the risk of unforeseen equipment downtime, and provides reliable support for the long-term stable operation and predictive maintenance of port equipment.

[0054] Example 2: like Figure 1 As shown in the figure, this embodiment provides a multi-condition adaptive intelligent early warning method for port equipment, which includes the following steps: Step S1: Data acquisition step, collecting port equipment data, and forming a high signal-to-noise ratio data stream after data preprocessing; Step S2: The step of working condition identification, which involves building a working condition identification model based on mechanistic features and data statistical features, and outputting the working condition type and time label in real time; Step S3: The step of generating dynamic thresholds involves calculating the initial threshold according to the working condition type, and then weighting and adjusting it according to the time period to obtain the dynamic threshold that adapts to the current operating state. Step S4: The abnormal warning step involves identifying the current operating condition and calling the corresponding threshold, judging the abnormality based on the continuous change trend, and pushing the warning information through multiple channels. Step S5: Optimize the dynamic threshold by building an optimization model, adjusting the threshold calculation weights, periodically iterating and optimizing the model, and adaptively updating the dynamic threshold.

[0055] In step S1: Port equipment data is collected through a distributed sensor network, real-time collection of all-dimensional operating data of port equipment, equipment operation commands and position encoder data, and the port equipment data is connected to the edge gateway; the edge gateway performs data preprocessing operations and outputs a continuous and stable high signal-to-noise ratio data stream as input data for subsequent operating condition identification steps.

[0056] The port equipment data includes comprehensive operational data, equipment operation command data, and position encoder data. The comprehensive operational data includes load current, lifting height, vibration acceleration along the X, Y, and Z axes, oil temperature, historical operational data, and real-time operational data. Load current is acquired via Hall effect sensors at a sampling frequency of 1Hz; lifting height is acquired via displacement encoders at a sampling frequency of 1Hz; trolley speed is acquired via speed sensors at a sampling frequency of 1Hz; vibration acceleration is acquired via three-axis accelerometers installed on key components such as the gantry crane motor, reducer, and boom at a sampling frequency of 10kHz; and oil temperature is acquired via thermocouple sensors at a sampling frequency of 1Hz. Historical operational data includes at least one year of complete operational data, covering different seasons and operational intensities. The equipment operation command data and position encoder data reflect the equipment's operational actions and displacement status.

[0057] After receiving port equipment data, the edge gateway performs unified access and parsing of port equipment data from different sources. It categorizes and manages the data according to data type and sampling frequency, and performs noise reduction processing, using methods such as moving average, low-pass filtering, or median filtering to weaken random noise and instantaneous spike interference. Simultaneously, it identifies and removes missing or anomalously fluctuating data points to prevent outliers from interfering with subsequent analysis. The edge gateway performs time alignment processing on various types of data based on a timestamp mechanism, mapping data with different sampling frequencies to a unified time axis, forming data samples arranged in chronological order. For high-frequency data such as vibration acceleration, rotational speed, and vibration frequency, the edge gateway performs downsampling or feature preprocessing locally to reduce subsequent transmission and computational load, performs standardization processing, converts it into a unified data structure, and encapsulates and outputs a continuous high signal-to-noise ratio data stream according to a preset data model, providing stable and standardized data input for the real-time analysis of subsequent operating condition identification models.

[0058] By utilizing distributed sensor networks and edge gateways to uniformly collect and preprocess multi-source operational data, the signal-to-noise ratio and temporal consistency of the collected data are effectively improved, and the impact of noise and abnormal fluctuations in the raw data on subsequent analysis is reduced, providing a reliable data foundation for the construction of operating condition identification models and dynamic threshold calculation.

[0059] In step S2: the high signal-to-noise ratio data stream is taken as input in chronological order, the mechanism features are extracted based on the mechanical structure and working principle of the equipment, and the working condition identification model is constructed by combining the data statistical features; the working condition mode of the equipment is automatically identified, and time dimension labels are introduced to record the time node and duration of the working condition switch, and the corresponding equipment working condition type and the time dimension label corresponding to the working condition are output to provide time reference for subsequent threshold adjustment.

[0060] The operating modes include standby, low load, normal load, and high load, and support custom expansion of operating states (such as heavy load start-stop, no load movement, etc.) according to the type of port equipment (such as gantry crane, belt conveyor, quay crane) and actual operation requirements. Mechanism characteristics: Selection of the coupling coefficient between load current and trolley The fundamental frequency and harmonic amplitude of the vibration acceleration signal. Statistical characteristics of the data are selected using a 5-second window, including the mean, variance, and peak factor.

[0061] A working condition identification model is constructed based on a combination of mechanistic and statistical data feature vectors. This model employs the C4.5 decision tree algorithm, using the information gain ratio (IGR) as the node splitting criterion to evaluate the effectiveness of features in splitting classification nodes. The decision tree model can recursively partition and discriminate multi-dimensional features, outputting the current working condition mode of the equipment. It also supports extending to custom working conditions based on equipment type and operational requirements.

[0062] The historical operating data of the equipment is labeled into different operating condition categories: when the load current is less than or equal to 5A, the equipment operating speed is 0, and this state lasts for no less than 30 seconds, the equipment is determined to be in standby mode; when the load current is greater than 5A but not more than 30A, and the operating speed is greater than 0 but not more than 0.5 meters per second, it is determined to be in low load mode; when the load current is greater than 30A but not more than 60A, and the operating speed is between 0.5 meters per second and 1.0 meter per second, it is determined to be in normal load mode; when the load current exceeds 60A and the operating speed is greater than 1.0 meter per second, it is determined to be in high load mode.

[0063] Based on this, historical operating data of the equipment was manually labeled. The labeled data was then divided into training samples (80%) and test samples (20%) to construct a working condition recognition model. Testing verified that the model achieved a working condition recognition accuracy of 96.8% on the test dataset, meeting the application requirements for automatic identification of port equipment operating conditions.

[0064] During training, the work condition recognition model selects the optimal split node by calculating the information gain ratio of each feature. After training, the accuracy of the work condition recognition model is verified using test data. The formula for calculating the information gain ratio is: ; Where Gain(A) is the information gain of feature A, representing the degree to which the uncertainty of the data is reduced after splitting the dataset using this feature, and SplitInfo(A) is the split information, representing the amount of information about the value distribution of feature A itself, used to normalize the information gain and avoid bias towards features with more values.

[0065] A higher information gain ratio indicates that the feature can significantly reduce the uncertainty of the dataset and has a reasonable value distribution, making it suitable as a basis for splitting decision tree nodes. In work condition identification, using the information gain ratio to select the optimal feature for splitting can effectively improve classification accuracy and ensure fair evaluation of multi-valued features, thereby achieving high-precision and low-latency work condition identification.

[0066] During the operation condition identification model discrimination process, after each model discrimination is completed, the timestamp of the corresponding time slice is bound to the discrimination operation condition type to generate a time dimension label. The time dimension label includes the start time node of the operation condition, the end time node of the operation condition, and the time period in which it is located. By comparing the operation condition types before and after in consecutive time slices, when the discrimination result is different from the operation condition of the previous time slice, it is determined that an operation condition switch has occurred, and the switch time node is automatically recorded to form a complete time dimension label.

[0067] Input the real-time operating data of the equipment into the trained operating condition recognition model, and output the operating condition type and the corresponding time dimension label, such as "normal load - early shift 8:30-9:15", with a recognition delay of less than 1 second.

[0068] In step S3: For the output equipment operating condition type, based on the equipment's historical operating data, statistical analysis and machine learning algorithms are used to identify the normal fluctuation range of each parameter under different operating conditions, and determine the initial threshold range of each parameter; a time period factor is introduced, and the period is set according to day, week, month, quarter, year. Each period can specify a specific time period. Combined with the equipment's historical operating data within the corresponding time period, the initial threshold range is weighted and optimized to generate dynamic thresholds under each operating condition, so as to realize the adaptive adjustment of thresholds for operating characteristics in different time periods.

[0069] Specifically, at least one year of historical equipment operation data is divided into four operating condition types, daily cycles, and weekly cycles. The operating condition types include standby, low load, normal load, and high load. The daily cycles are divided into morning shift 8:00–16:00, evening shift 16:00–24:00, and night shift 0:00–8:00. The weekly cycles are divided into weekdays and weekends, forming 12 subdivided historical equipment operation datasets. The 12 subdivided historical equipment operation datasets consist of 4 operating condition types × 3 time periods.

[0070] For each type of operating condition, the initial threshold range for each operating parameter is calculated based on the corresponding historical operating data. Taking the vibration acceleration X-axis parameter as an example, the normal fluctuation range of the parameter is determined using the 3σ principle based on the mean μ and standard deviation σ of the parameter under different operating conditions. Under standby conditions, the mean vibration acceleration is 0.2g and the standard deviation is 0.03g, corresponding to an initial threshold range of [0.11g, 0.29g]; under low load conditions, the mean is 0.55g and the standard deviation is 0.08g, corresponding to an initial threshold range of [0.31g, 0.79g]; under normal load conditions, the mean is 0.85g and the standard deviation is 0.12g, corresponding to an initial threshold range of [0.49g, 1.21g]; under high load conditions, the mean is 1.3g and the standard deviation is 0.17g, corresponding to an initial threshold range of [0.79g, 1.81g].

[0071] A time period factor is introduced to dynamically adjust the initial threshold range. The recent data duration is set to 90 days, and the total historical data duration is set to 365 days, resulting in a weight for the recent data. Based on the statistical results of equipment operating characteristics within different time periods, corresponding time period weights are assigned to each time period. For example, under normal load conditions during the early morning shift, the overall vibration acceleration level is higher due to the equipment startup shock. Therefore, the time period weight for this period is set to... The remaining time periods are set as Substituting recent data weights and time period weights into the dynamic threshold calculation formula, the initial threshold range for normal load early shift conditions is weighted and corrected, resulting in a dynamic threshold lower limit of 0.49g × 0.247 × 1.1 ≈ 0.13g and an upper limit of 1.21g × 0.247 × 1.1 ≈ 0.33g under this condition. Based on actual engineering conditions, the dynamic threshold range is adjusted to [0.15g, 0.32g]. Furthermore, a low alarm threshold of 0.25g, a high alarm threshold of 0.30g, and an automatic deactivation threshold of 0.20g are set to support multi-level alarm judgment in subsequent abnormal warning processes.

[0072] In step S4: real-time collected port equipment full-dimensional operating parameter data, equipment operating condition type, and time label are used as input. According to the current equipment operating condition type, the corresponding dynamic threshold range is matched. The monitoring platform compares the real-time parameters with the dynamic threshold and performs point-by-point comparison and trend analysis on the real-time parameters. When the real-time parameters exceed the threshold range, a comprehensive judgment is made by introducing continuity criteria and abrupt change amplitude criteria. After filtering out short-term disturbance signals, an alarm is triggered, and an abnormal warning result of the corresponding level is generated. The abnormal warning result is pushed to the operation end through multiple channels such as SMS, mobile mini-program, and platform pop-up window. At the same time, the warning trigger time, current equipment operating condition type, abnormal parameter value, and other information are recorded. The operation end performs maintenance and records the operation and maintenance verification results and equipment maintenance records.

[0073] Taking the vibration acceleration of the gantry crane as an example, the real-time collected full-dimensional operating parameters of the port equipment are uploaded to the monitoring platform every 100ms. The monitoring platform identifies the current normal load - early shift condition based on the real-time load current (45A) and velocity (0.8m / s), and calls the corresponding dynamic thresholds [0.15g, 0.32g] for comparison. When the vibration acceleration is detected to be 0.31g, 0.33g, and 0.32g for three consecutive sampling values, all exceeding the high alarm threshold of 0.30g, and the sudden change amplitude reaches 35% within 5 seconds, after eliminating instantaneous interference, a high alarm warning is triggered. The warning information is pushed to the operation terminal via SMS and mobile mini-program, and a pop-up window is displayed on the platform, recording the warning time (8:42:15), operating condition type, parameter values, and other information.

[0074] In step S5: a warning feedback dataset containing warning results, operation and maintenance verification results, and equipment maintenance records is constructed; based on the warning feedback dataset, a threshold optimization model is constructed using a gradient boosting algorithm, and various weight coefficients involved in the dynamic threshold calculation process are updated and calculated; the model is iteratively trained according to a preset optimization cycle and convergence conditions, the updated weight parameters are output, and the weight parameters are fed back to the dynamic threshold generation process to generate new dynamic thresholds, thereby achieving adaptive updating of the threshold.

[0075] Taking gantry crane vibration acceleration as an example, early warning information and maintenance feedback on gantry crane vibration acceleration were continuously recorded. A total of 120 early warning data points were collected over 30 days, including 28 true anomalies, 7 false alarms, and 3 missed alarms. An early warning feedback dataset was constructed. Using this dataset as training samples, the weight coefficients were adjusted using a gradient boosting algorithm. The formula is as follows: Where η = 0.03 is the learning rate, and L is the loss function based on the false positive rate and the false negative rate. These are the weighting coefficients before the update. After optimization, these are the weighting coefficients for the normal load conditions during the early morning shift. Adjusted to 0.26, The version was adjusted to 1.08, and the dynamic thresholds were updated to [0.16g, 0.31g]. After continuous monitoring for 30 days following the optimization, it was found that the early warning accuracy improved from 92% to 95% under normal early shift load conditions, the false alarm rate decreased from 5.8% to 3.2%, and the missed alarm rate decreased from 2.5% to 1.8%, meeting the port operation and maintenance requirements.

[0076] Example 3: like Figure 2 As shown in the figure, this embodiment provides a port equipment multi-condition adaptive intelligent early warning system, including: Data acquisition module 1 continuously monitors the operation of port equipment through a distributed sensor network, synchronously acquiring port equipment data according to a preset sampling frequency, and performing time alignment and unified encapsulation on data from different sampling frequencies. Port equipment data includes full-dimensional operational data of the port equipment, equipment operation commands, and displacement information fed back from the position encoder. After acquisition, the edge gateway performs data preprocessing to reduce the impact of electromagnetic interference, transient shocks, and other factors on data quality, outputting a continuous and stable high signal-to-noise ratio data stream in chronological order. This ensures data integrity and reliability, providing high-quality input for condition identification and threshold generation, thereby improving the accuracy and real-time performance of the overall monitoring system.

[0077] The operating condition identification module 2 receives a high signal-to-noise ratio data stream, calculates its mechanical characteristics based on the equipment's mechanical structure and working principle, and statistically analyzes the distribution characteristics of each parameter within a time window to extract statistical features. These extracted mechanistic and statistical features are used to construct an operating condition identification model, perform discrimination calculations, and output the operating condition type corresponding to the equipment within the current time period. The operating condition identification module generates a time label for each type of operating condition to identify its start and end times, and outputs the corresponding equipment operating condition type and its associated time label as the result. This provides a reliable basis for dynamic threshold generation and anomaly warning, reducing misjudgments and delays.

[0078] The dynamic threshold generation module 3, based on the corresponding operating condition type and time stamp of the equipment, and using historical operating data, calculates the normal fluctuation range of parameters using statistical analysis methods to determine the initial threshold interval for each operating condition. A time cycle factor is introduced to map equipment operating time to daily, weekly, monthly, or quarterly cycles, further subdivided into specific work periods. Combining historical operating data within this time period, the initial threshold interval is weighted and optimized to generate a dynamic threshold adapted to the current operating condition type and time stamp. This module can adjust the threshold in real time according to the equipment operating condition and time cycle, improving threshold matching and sensitivity, thereby reducing false alarm and false negative rates.

[0079] The anomaly early warning module 4 continuously receives equipment operating parameter data during real-time operation of port equipment. Based on the operating condition identification results, it determines the current operating status of the equipment, retrieves the dynamic threshold range for the corresponding operating condition, compares real-time parameters point by point, and analyzes parameter change trends over time. When parameters exceed the threshold range, it introduces continuity and abrupt change criteria to comprehensively judge the abnormal state, distinguishing between persistent anomalies and short-term disturbances. Confirmed abnormal events are assigned corresponding warning levels and sent to the maintenance end through multiple channels such as SMS, mobile mini-programs, and platform pop-ups, while simultaneously generating warning records. This module achieves real-time and accurate anomaly identification and multi-channel notification, improving maintenance response speed and ensuring safe equipment operation.

[0080] The threshold optimization module 5 constructs an early warning feedback dataset and uses it to train a threshold optimization model. By adjusting the weight parameters involved in the dynamic threshold calculation, it achieves adaptive optimization of the threshold generation model. The optimization process is executed automatically according to a preset cycle, and the model's stability is judged based on changes in the false alarm rate and false negative rate. When equipment enters a new stage of its life cycle or its operating characteristics change significantly, the optimization frequency is accelerated to achieve rapid updates of the dynamic threshold. This module continuously improves the adaptability and accuracy of the dynamic threshold, effectively reduces the false alarm rate and false negative rate, and ensures the long-term stability and reliability of the anomaly early warning system.

[0081] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. The methods disclosed in the embodiments are described simply because they correspond to the systems disclosed in the embodiments; relevant details can be found in the method section.

[0082] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0083] In the embodiments provided by this invention, it should be understood that the disclosed systems, methods, and approaches can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between systems or units may be electrical, mechanical, or other forms.

[0084] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0085] In addition, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit.

[0086] Similarly, in the various embodiments of the present invention, each processing unit can be integrated into a functional module, or each processing unit can exist physically, or two or more processing units can be integrated into a functional module.

[0087] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0088] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0089] The above-disclosed embodiments are merely preferred embodiments of the present invention, but the present invention is not limited thereto. Any non-creative variations that can be conceived by those skilled in the art, as well as any improvements and modifications made without departing from the principles of the present invention, should fall within the protection scope of the present invention.

Claims

1. A multi-condition adaptive intelligent early warning method for port equipment, characterized in that, Includes the following steps: Step S1: Data acquisition step, collecting port equipment data, and forming a high signal-to-noise ratio data stream after data preprocessing; Step S2: The step of working condition identification, which involves constructing a working condition identification model based on mechanistic features and data statistical features, and outputting the working condition type and time label in real time; Step S3: The step of generating dynamic thresholds involves calculating the initial threshold according to the working condition type, and then weighting and adjusting it based on the time period to obtain a dynamic threshold that adapts to the current operating state. Step S4: The abnormal warning step involves identifying the current operating condition and calling the corresponding threshold, judging the abnormality based on the continuous change trend, and pushing the warning information through multiple channels. Step S5: Optimize the dynamic threshold by building an optimization model, adjusting the threshold calculation weights, periodically iterating and optimizing the model, and adaptively updating the dynamic threshold.

2. The multi-condition adaptive intelligent early warning method for port equipment according to claim 1, characterized in that, In step S1: Port equipment data is collected through a distributed sensor network and connected to an edge gateway; the edge gateway performs data preprocessing and outputs a high signal-to-noise ratio data stream.

3. A multi-condition adaptive intelligent early warning method for port equipment according to claim 1 or 2, characterized in that, The port equipment data includes comprehensive operational data, equipment operation command data, and position encoder data. The comprehensive operational data includes lifting capacity, load current, vibration acceleration, rotational speed, vibration frequency, operating time, oil pressure, oil temperature, historical operational data, and real-time operational data. Lifting capacity is acquired using a high-precision tension sensor with a sampling frequency of 1Hz; load current is acquired using a Hall effect sensor with a sampling frequency of 1Hz; vibration acceleration is acquired using a triaxial accelerometer with a sampling frequency of 10kHz; and oil temperature is acquired using a thermocouple sensor with a sampling frequency of 1Hz. Historical operational data includes at least one complete operational cycle of valid data. The equipment operation command data and position encoder data reflect the equipment's operational actions and displacement status.

4. The multi-condition adaptive intelligent early warning method for port equipment according to claim 3, characterized in that, In step S2: the high signal-to-noise ratio data stream is input in chronological order; mechanism features are extracted based on the equipment's mechanical structure and working principle; and a working condition identification model is constructed by combining data statistical features. The working condition mode of the equipment is automatically identified, and time dimension labels are introduced to record the time nodes of working condition switching. The corresponding equipment working condition type and the time label corresponding to the equipment working condition are output. The working condition modes include standby, low load, normal load, and high load, and support custom expansion of the operating status according to the port equipment type and actual operation requirements.

5. The multi-condition adaptive intelligent early warning method for port equipment according to claim 4, characterized in that, The mechanistic characteristics include load-speed correlation characteristics and vibration harmonic characteristics. The load-speed correlation characteristics are calculated by using the ratio of load current to equipment operating speed to determine the load-speed coupling coefficient. The description of the mechanical load characteristics and drive efficiency of the equipment under different loads; the vibration harmonic characteristics are obtained by frequency domain analysis of the vibration acceleration signal to extract the fundamental frequency and harmonic amplitude; the data statistical characteristics are obtained by statistical calculation of the equipment operating parameters over time series, including the moving mean, variance and peak factor. A working condition identification model is constructed based on a combination of mechanistic and statistical data feature vectors. This model employs the C4.5 decision tree algorithm, using the information gain ratio as the node splitting criterion. The information gain ratio is calculated using the following formula: ; During the working condition identification model discrimination process, after each model discrimination is completed, the timestamp of the corresponding time slice is bound to the discriminated working condition type to generate a time dimension label. The time dimension label includes the working condition start time node, the working condition end time node, and the time period in which it is located. By comparing the working condition types before and after in consecutive time slices, when the discrimination result is different from the working condition of the previous time slice, it is determined that a working condition switch has occurred, and the switch time node is automatically recorded.

6. The multi-condition adaptive intelligent early warning method for port equipment according to claim 5, characterized in that, In step S3: For the output equipment operating condition type, based on historical equipment operating data, statistical analysis and machine learning algorithms are used to identify the normal fluctuation range of each parameter under different operating conditions, and determine the initial threshold range of each parameter; a time period factor is introduced, and periods are set according to daily, weekly, monthly, quarterly, and annual periods. Each period can specify a specific time period. Combined with the historical equipment operating data within the corresponding time period, the initial threshold range is weighted and optimized to generate dynamic thresholds for each operating condition. Dynamic thresholds include low reporting, low-low reporting, high reporting, high-high reporting thresholds, and automatic release thresholds; the dynamic threshold generation process follows the proximity principle. The process of determining the initial threshold range is as follows: for each equipment operating condition type, based on the mean μ and standard deviation σ of the parameter, the normal fluctuation range of the parameter is determined using the 3σ principle, and the formula is: Introducing dual-weight coefficients to perform time-weighted optimization on the initial threshold can be described by the following formula: The dual-weighting coefficients include weights for recent data. and time period weight ; This is the ratio of the duration of recent data to the total duration of historical data.

7. The multi-condition adaptive intelligent early warning method for port equipment according to claim 6, characterized in that, In step S4: the real-time collected port equipment full-dimensional operation parameter data, equipment operating condition type and time label are used as input, and the corresponding dynamic threshold range is matched according to the current equipment operating condition type. The monitoring platform compares the real-time parameters with the dynamic threshold and performs point-by-point comparison and trend analysis on the real-time parameters. When real-time parameters exceed the threshold range, a comprehensive judgment is made by introducing continuity criteria and abrupt change amplitude criteria. After filtering out short-term disturbance signals, an alarm is triggered, generating an abnormal warning result of the corresponding level. The abnormal warning result is pushed to the operation end through multiple channels. At the same time, the warning trigger time, the current equipment operating condition type, and the abnormal parameter value information are recorded. The operation end performs maintenance and records the operation and maintenance verification results and equipment maintenance records.

8. The multi-condition adaptive intelligent early warning method for port equipment according to claim 7, characterized in that, In step S5: a warning feedback dataset containing warning results, operation and maintenance verification results, and equipment maintenance records is constructed; Based on the early warning feedback dataset, a threshold optimization model is constructed using the gradient boosting algorithm. Various weight coefficients involved in the dynamic threshold calculation process are updated and calculated. The model is iteratively trained according to the preset optimization cycle and convergence conditions, and the updated weight parameters are output. The weight parameters are then fed back to the dynamic threshold generation process to generate new dynamic thresholds.

9. A multi-condition adaptive intelligent early warning system for port equipment, characterized in that, include: Data acquisition module (1), working condition identification module (2), dynamic threshold generation module (3), anomaly warning module (4), threshold optimization module (5); The data acquisition module (1) acquires port equipment data and performs preprocessing operations on the port equipment data to form a high signal-to-noise ratio data stream; The working condition identification module (2) constructs a working condition identification model based on mechanism features and data statistical features, analyzes the high signal-to-noise ratio data stream, and outputs the corresponding equipment working condition type and the time label corresponding to the equipment working condition in real time. The dynamic threshold generation module (3) calculates the initial threshold according to the working condition type of the equipment, and adjusts the initial threshold by weighting it in combination with the time period characteristics to generate a dynamic threshold that is adapted to the current working condition. The abnormal warning module (4) is used to identify the current operating condition of the equipment and call the corresponding dynamic threshold, combine the continuous change trend of the operating parameters to determine whether there is an abnormality, and output the warning information through a multi-channel method; The threshold optimization module (5) is used to construct a threshold optimization model, adjust the weight parameters in the dynamic threshold calculation, and iteratively update the model according to a set period to achieve adaptive optimization of the dynamic threshold.

10. A port equipment multi-condition adaptive intelligent early warning system according to claim 9, characterized in that, The data acquisition module (1) continuously monitors the operation of port equipment through a distributed sensor network, synchronously acquires port equipment data according to a preset sampling frequency, and performs time alignment and unified encapsulation on data with different sampling frequencies; the port equipment data includes all-dimensional operation data of port equipment, equipment operation instructions and displacement information fed back by the position encoder. After the acquisition is completed, the edge gateway performs data preprocessing operation on the port equipment data and outputs a high signal-to-noise ratio data stream. The working condition identification module (2) receives a high signal-to-noise ratio data stream, calculates the mechanical characteristics of the equipment based on its mechanical structure and working principle, and performs statistical analysis on the distribution characteristics of each parameter within the time window. It extracts data statistical characteristics, and uses the extracted mechanical characteristics and data statistical characteristics to construct a working condition identification model, perform discrimination calculations, output the working condition type of the equipment within the current time period, generate time tags for each type of working condition to identify the start and end times of the working condition, and outputs the corresponding equipment working condition type and the time tag corresponding to the equipment working condition as the result. The dynamic threshold generation module (3) calculates the normal fluctuation range of parameters based on the working condition type and time tag corresponding to the equipment and the historical operating data of the equipment, determines the initial threshold interval under each working condition, introduces the time cycle factor, maps the equipment operating time to the daily cycle, weekly cycle, monthly cycle or quarterly cycle, and subdivides it into specific working periods. Combined with the historical operating data of the equipment in this period, the initial threshold interval is weighted and optimized to generate a dynamic threshold that is adapted to the current working condition type and time tag. The abnormal early warning module (4) continuously receives equipment operating parameter data during the real-time operation of port equipment, determines the current operating status of the equipment based on the operating condition identification result, retrieves the dynamic threshold range under the corresponding operating condition, compares the real-time parameters point by point, and analyzes the parameter change trend in the time dimension. When the parameters exceed the threshold range, the continuity criterion and the mutation amplitude criterion are introduced to make a comprehensive judgment on the abnormal state. The confirmed abnormal events are assigned the corresponding early warning level and sent to the operation and maintenance end through multiple channels, while generating early warning records. The threshold optimization module (5) constructs an early warning feedback dataset, uses the early warning feedback dataset to train a threshold optimization model, and adjusts the weight parameters involved in the dynamic threshold calculation to achieve adaptive optimization of the threshold generation model. The optimization process is automatically executed according to a preset cycle, and the model stability is judged based on the changes in false alarm rate and false alarm rate. When the equipment enters a new stage of its life cycle or the operating characteristics change significantly, the optimization frequency is accelerated to achieve rapid updating of the dynamic threshold.

Images