Method and system for monitoring abnormal pressure of fuel oil common rail pipe of marine main engine

Abstract

The invention discloses a method and a system for monitoring pressure abnormity of a fuel oil common rail pipe of a marine main engine. The method comprises the following steps: acquiring historical operation parameter data and constructing a data set; the historical operation parameter data comprises historical host load data, historical host rotating speed data and historical fuel oil common rail pressure data; historical operation parameter data are preprocessed, an SVDD model is constructed and trained based on the historical operation parameter data, the model obtains a historical normal data hyper-sphere in a three-dimensional feature space, and the center and the radius of the hyper-sphere are obtained; similarly, real-time operation parameter data collected in real time are preprocessed and then input into the model, the distance between the real-time operation parameter data and the center of the hypersphere is obtained, and the value of the distance is compared with the value of the radius of the hypersphere in historical normal data; and when the numerical value of the distance is large, it is judged that the real-time operation parameter data is within an abnormal range, the fuel oil common rail pressure of the ship main engine is in an abnormal state, and an alarm signal is triggered. The method has the effect of low error rate.

Description

[0001] This application relates to the field of ship pressure monitoring, and in particular to a method and system for monitoring abnormal pressure in the common rail of ship main engine fuel. Background Technology

[0002] Marine main engine fuel common rail systems operate under high pressure, and their pressure stability directly affects the main engine's performance and safety. Traditional monitoring methods mainly rely on manual periodic inspections or preset fixed threshold alarm mechanisms. Manual periodic inspections suffer from low efficiency and poor real-time performance, making it difficult to continuously capture dynamic pressure changes. While fixed threshold alarms can identify obvious out-of-limit anomalies, the set values ​​are usually based on experience or theoretical calculations, making it difficult to adapt to the complex and ever-changing operating conditions of the main engine, easily leading to false alarms or missed alarms. In particular, minor anomalies in the early stages may go undetected because they do not exceed the normal threshold range, resulting in the accumulation of potential fault hazards. Existing technologies lack the ability to dynamically analyze the multidimensional correlation between main engine load, speed, and fuel common rail pressure, and cannot establish safety boundaries that adaptively adjust according to operating conditions. Therefore, there is an urgent need for an intelligent monitoring solution that does not require prior fault data and can identify abnormal states of multidimensional operating parameters in real time. Summary of the Invention

[0003] To improve the problems of high error rate and low monitoring accuracy of traditional monitoring methods, this application provides a method and system for monitoring abnormal pressure in the common rail of marine main engine fuel.

[0004] In the first aspect, the method for monitoring abnormal pressure in the common rail fuel line of a ship's main engine provided in this application adopts the following technical solution: S1, acquiring historical operating parameter data of the ship's main engine and constructing a historical operating parameter dataset; The historical operating parameter data includes historical main engine load data, historical main engine speed data, and historical fuel common rail pressure data; S2. Preprocess the historical operating parameter data to obtain preprocessed historical operating parameter data; S3. Construct and train an SVDD model based on the preprocessed historical operating parameter data. The SVDD model obtains a historical normal data hypersphere in the three-dimensional feature space and obtains the three-dimensional feature position and radius of the center of the historical normal data hypersphere. The volume of the historical normal data hypersphere is defined as the distribution range of the historical operating parameter data in the three-dimensional feature space; S4. Preprocess the real-time collected operating parameter data according to the steps in S2, and input the preprocessed real-time operating parameter data into the SVDD model. Calculate the distance between the real-time operating parameter data and the three-dimensional feature position of the center of the historical normal data hypersphere, and compare the magnitude of the distance between the real-time operating parameter data and the center of the historical normal data hypersphere with the magnitude of the radius of the historical normal data hypersphere. The real-time operating parameter data includes real-time main engine load data, real-time main engine speed data, and real-time fuel common rail pressure data; S5. When the distance between the real-time operating parameter data and the center of the three-dimensional feature position of the historical normal data hypersphere is greater than the radius of the historical normal data hypersphere, it is determined that the real-time operating parameter data is within the abnormal range, the common rail pressure of the ship's main engine fuel is in an abnormal state, and an alarm signal is triggered.

[0005] By adopting the above technical solution, this method constructs a historical operating parameter dataset by acquiring historical operating parameter data of the ship's main engine. All historical operating parameter data is preprocessed uniformly. Based on this, an SVDD model is constructed and trained to form a hypersphere of historical normal data in a three-dimensional feature space to define the normal distribution range of historical operating parameter data. Real-time operating parameter data is then subjected to the same preprocessing and input into the trained SVDD model. Anomalies in the ship's main engine fuel common rail pressure are detected and alarms are triggered by comparing the distance between the center of the real-time operating parameter data and the center of the historical normal data hypersphere with the radius of the hypersphere. This simplifies the monitoring process of ship's main engine fuel common rail pressure, eliminating the need for manual monitoring. Using historical operating parameter data to train the SVDD model ensures high accuracy and reliability of anomaly monitoring. Three-dimensional feature space modeling fully integrates historical operating parameter data, improving the monitoring efficiency of main engine fuel common rail pressure anomalies; the preprocessing step improves the quality of historical operating parameter data, reducing the risk of false alarms. Automated real-time monitoring provides timely warnings of potential faults to prevent damage to main engine equipment and extend the service life of the main engine, ultimately reducing maintenance costs, operational burden, and downtime losses.

[0006] Optionally, step S5 further includes determining that the real-time operating parameter data is within the normal range and the ship's main engine fuel common rail pressure is in a normal state when the distance between the real-time operating parameter data and the center of the historical normal data hypersphere is less than or equal to the radius of the historical normal data hypersphere, adding the real-time operating parameter data to the historical operating parameter dataset, and updating the SVDD model based on this.

[0007] By adopting the above technical solution, real-time operating parameter data within the normal range is continuously expanded into the historical operating parameter dataset. This allows the hypersphere boundary of the historical normal data to dynamically adjust with equipment aging and changes in operating conditions, avoiding misjudgments caused by the time-deterioration of historical operating parameter data in traditional static models. This significantly improves the monitoring adaptability throughout the lifecycle. The new normal historical operating parameter dataset drives the re-solution and updating of the center and radius of the historical normal data hypersphere, simultaneously optimizing the boundary of the historical operating parameter dataset. This ensures that the data distribution range for anomaly detection always matches the current optimal operating state, reducing the false alarm rate caused by operating condition fluctuations. Furthermore, only rigorously verified normal real-time operating parameter data is received, preventing abnormal real-time operating parameter data from contaminating the historical operating parameter dataset. This incremental update model effectively reduces the computational load of full retraining, significantly reducing hardware resource consumption while ensuring real-time monitoring.

[0008] Optionally, the preprocessing of the historical operating parameter data includes removing outliers, filtering noisy data, and smoothing the data to obtain cleaned historical operating parameter data; then, the values ​​of the cleaned historical operating parameter data are uniformly mapped to the same normalization range through normalization processing to obtain normalized historical operating parameter data.

[0009] By employing the aforementioned technical solution, historical operating parameter data is specifically cleaned through outlier removal, noise filtering, and smoothing. This effectively eliminates the impact of sensor interference and instantaneous fluctuations on historical operating parameter data, ensuring the reliability of the core trends in the historical operating parameter data. The cleaned historical operating parameter data is then uniformly normalized to eliminate differences in dimensions and numerical magnitudes between data points, allowing multidimensional features to achieve a balanced representation in the three-dimensional feature space. This solution significantly improves the stability and generalization ability of subsequent SVDD model training. The cleaning step ensures the authenticity of historical operating parameter data, preventing outliers from distorting the boundaries of normal historical data beyond the spherical boundary. The normalization operation balances the feature contribution, enabling the SVDD model to accurately capture the dynamic coupling relationship between main engine load, main engine speed, and common rail fuel pressure data, thereby improving the sensitivity and accuracy of real-time anomaly detection and reducing the risk of false alarms and missed alarms from the source of operating parameter data.

[0010] Optionally, step S3 includes selecting a radial basis function kernel function to construct the SVDD model; inputting the preprocessed historical running parameter data into the SVDD model; training the SVDD model to obtain a historical normal data hypersphere in the three-dimensional feature space; and calculating and solving the three-dimensional feature position and radius of the center of the historical normal data hypersphere, so that all historical running parameter data in the historical running parameter dataset that are within the normal range are located inside or on the boundary of the historical normal data hypersphere.

[0011] By adopting the above technical solution, a radial basis function (RBF) kernel function is selected to construct an SVDD model. Preprocessed historical operating parameter data is input into the model for training, forming a historical normal data hypersphere in a three-dimensional feature space, and solving for the three-dimensional feature position and radius of its center. This solution significantly improves the performance of ship main engine fuel common rail pressure monitoring. The RBF kernel function effectively maps nonlinear data relationships, enhancing the fusion expression of main engine load data, main engine speed data, and fuel common rail pressure data in the three-dimensional feature space, enabling the constructed SVDD model to accurately fit complex operating modes. This historical normal data hypersphere defines the boundary of the distribution range of normal historical operating parameter data in the three-dimensional feature space, maximizing the inclusiveness of the normal historical operating parameter data distribution and avoiding overgeneralization or boundary ambiguity. The training process optimizes the solution for the center and radius of the historical normal data hypersphere, ensuring the stability and reliability of the distance calculation between the real-time operating parameter data and the three-dimensional feature position of the center of the historical normal data hypersphere when performing real-time operating parameter data anomaly judgment. Overall, it enhances the sensitivity and accuracy of abnormal pressure monitoring, reduces the risk of false alarms and missed alarms, and reduces the complexity of model training and computational resource consumption, providing efficient support for automated monitoring systems, helping to prevent host failures and extend equipment life.

[0012] Optionally, based on historical host load data, multiple operating condition intervals are divided, and the correlation weight between historical fuel common rail pressure data and historical host speed data in each interval is calculated. Based on the correlation weight, the historical operating parameter data in the three-dimensional feature space is weighted and scaled to generate a weighted feature vector. The weighted feature vector is used to train the SVDD model to construct a hypersphere of historical normal data.

[0013] By adopting the above technical solution, the SVDD model is trained by dividing the historical main engine load data into operating condition intervals and calculating the weights of the correlation between pressure and speed. This is done after weighted scaling of the historical operating parameter data in the three-dimensional feature space. The division of operating condition intervals accurately identifies the operating state characteristics under different load conditions. The correlation weight calculation quantifies the dynamic coupling strength between pressure and speed, and the weighted scaling operation strengthens the weights of key correlation parameters in the three-dimensional feature space. This solution significantly optimizes the boundary construction of the hypersphere for historical normal data. The weighted eigenvectors enable the SVDD model to focus on the essential correlation patterns between historical fuel common rail pressure data and historical main engine speed data within different operating condition intervals, suppressing unnecessary fluctuation interference and enhancing the accuracy of identifying multi-operating condition coupling relationships in the three-dimensional feature space. This improves the adaptability and robustness of monitoring abnormal ship main engine fuel common rail pressure.

[0014] Optionally, step S6 is also included: when an alarm signal is triggered, real-time common rail vibration spectrum data is acquired simultaneously; if the amplitude of the vibration spectrum at the characteristic frequency exceeds the adaptive threshold, then the fuel common rail pressure is confirmed to be abnormal.

[0015] By adopting the above technical solution, real-time common rail vibration spectrum data is simultaneously collected upon alarm triggering. Anomaly confirmation is then achieved by comparing the characteristic frequency amplitude with an adaptive threshold. This dual-channel verification system integrates real-time operating parameter data anomaly alarms with vibration spectrum characteristics and frequency amplitude analysis. Leveraging the high sensitivity of vibration signals to mechanical conditions, it effectively distinguishes between genuine common rail pressure anomalies and potential environmental interference false alarms from the SVDD model, thus improving alarm confidence. The adaptive threshold adjusts the tolerance of the characteristic frequency amplitude in real time according to different operating conditions, avoiding the failure of fixed thresholds under changes in engine speed and load. This ensures the applicability and accuracy of the vibration spectrum judgment method across the entire operating range. Simultaneously, characteristic frequency amplitudes exceeding the adaptive threshold directly correlate with the common rail mechanical vibration characteristics, providing a physical attribution basis for common rail pressure anomalies. This assists in quickly locating potential fault points, preventing system cascading damage caused by common rail runaway, significantly reducing false alarm rates and improving fault diagnosis efficiency.

[0016] Optionally, the historical operating parameter dataset obtained in step S1 is divided into a historical normal operating parameter dataset and a historical abnormal operating parameter dataset; the historical normal operating parameter dataset is further divided into a training set and a test subset; the test subset and the historical abnormal operating parameter dataset are combined to generate a test set; wherein, the training set is used to train the SVDD model to construct a historical normal data hypersphere; the test set is used to input into the trained SVDD model to calculate its distance from the three-dimensional feature position of the center of the historical normal data hypersphere, and based on the comparison result of this distance and the radius of the historical normal data hypersphere, the judgment ability of the SVDD model is tested, and the historical normal data hypersphere constructed by the SVDD model is optimized and improved according to the test results.

[0017] By adopting the above technical solution, the accuracy of the hypersphere boundary in historical normal data is significantly optimized. The test set comprehensively covers both normal and abnormal states, ensuring the comprehensiveness of model testing. A feedback mechanism based on the comparison of the distance and radius of the three-dimensional feature positions of the hypersphere center between the test set data and historical normal data drives iterative optimization of the hypersphere center and radius in historical normal data, improving the tightness of the boundary's representation of the load-speed-pressure coupling relationship. Ultimately, this enhances the reliability and adaptability of the anomaly monitoring system, reduces the risk of misjudgment caused by variable operating condition interference, and thus improves the robustness and early warning accuracy of common rail pressure anomaly monitoring for ship main engine fuel.

[0018] Secondly, this application provides a system for monitoring abnormal pressure in the common rail of marine main engine fuel, including a data acquisition module, a data preprocessing module, a model training module, a real-time monitoring module, and an alarm module; The data acquisition module is used to acquire historical operating parameter data of the ship's main engine and construct a historical operating parameter dataset; The data preprocessing module is used to clean historical fuel common rail pressure data separately and then preprocess historical operating parameter data uniformly. The model training module is used to build and train the SVDD model based on preprocessed historical running parameter data, and to update the trained SVDD model after real-time running parameter data is added to the historical running parameter dataset; The real-time monitoring module is used to collect real-time operating parameter data, process the real-time operating parameter data according to the preprocessing steps of the data preprocessing module, and input the preprocessed real-time operating parameter data into the SVDD model; calculate the distance between the real-time operating parameter data and the three-dimensional feature position of the center of the hypersphere of historical normal data, and compare the relationship between the distance value and the radius value of the hypersphere of historical normal data. The alarm module is used to trigger an alarm signal when the distance value of the three-dimensional feature position of the center of the sphere in historical normal data is greater than or equal to the radius value.

[0019] Understandably, the system for monitoring abnormal pressure in the common rail of marine main engine fuel provided in the second aspect above is used to execute the method provided in this application. Therefore, the beneficial effects it can achieve can be referred to the beneficial effects in the corresponding method, and will not be repeated here.

[0020] In summary, this application includes at least one of the following beneficial technical effects: 1. By employing a method of constructing an SVDD model based on historical operating parameter data and creating a hypersphere of historical normal data in a three-dimensional feature space, the system can accurately delineate the boundary between normal operating parameter data and abnormal operating parameter data under multi-dimensional features of main engine load, main engine speed, and fuel common rail pressure. This effectively solves the problem of false alarms and missed alarms caused by the excessive reliance on a single parameter in traditional threshold monitoring methods, thereby achieving real-time high-precision monitoring of abnormal fuel common rail pressure of ship main engines and significantly improving the efficiency of automated anomaly monitoring.

[0021] 2. Due to the adoption of a dynamic optimization mechanism that automatically updates the historical operating parameter dataset and retrains the SVDD model when the real-time operating parameter data is within the normal range, the system can adaptively learn changes in the main engine's operating status. This effectively solves the model degradation problem caused by equipment aging or environmental drift during long-term monitoring, thereby ensuring the continuous accuracy and strong adaptability of the ship's main engine fuel common rail pressure anomaly monitoring and reducing manual maintenance costs.

[0022] 3. By employing separate cleaning of historical fuel common rail pressure data followed by unified normalization preprocessing of historical operating parameter data, and weighted feature scaling technology based on the correlation weight of operating condition intervals, combined with a secondary confirmation step for alarm signals using vibration spectrum data, the system enhances the robustness and reliability of identifying abnormalities in the ship's main engine fuel common rail pressure. It effectively eliminates the impact of noise interference and uneven distribution of operating parameter data, thereby providing multi-level fault protection for the ship's main engine and ensuring operational stability and safety. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating the method for monitoring abnormal pressure in the common rail fuel line of a ship's main engine, as provided in Embodiment 1 of this application. Figure 2 This is another flowchart illustrating the method for monitoring abnormal pressure in the common rail fuel line of a ship's main engine, as provided in Embodiment 2 of this application. Figure 3 This is a schematic diagram of the system for monitoring abnormal pressure in the common rail fuel line of a ship's main engine, provided in an embodiment of this application. Figure 4 This is a computer device architecture diagram provided in the embodiments of this application. Detailed Implementation

[0025] The following is in conjunction with the appendix Figure 1 —4 provides further detailed description of this application.

[0026] This application discloses a method and system for monitoring abnormal pressure in the common rail fuel line of a ship's main engine.

[0027] Reference Figure 1 A method for monitoring abnormal pressure in the common rail fuel line of a ship's main engine includes S1: acquiring historical operating parameter data of the ship's main engine and constructing a historical operating parameter dataset; wherein the historical operating parameter data includes historical main engine load data, historical main engine speed data, and historical common rail fuel pressure data.

[0028] Among them, historical main engine load data represents the mechanical power output parameters of the ship's main engine during normal operation, reflecting the main engine load level. Historical main engine speed data represents the crankshaft rotation speed parameters, indicating operational stability. Historical fuel common rail pressure data represents the high-pressure pipeline pressure parameters in the fuel supply system, used to monitor fuel injection status.

[0029] Specifically, historical main engine load data, historical main engine speed data, and historical common rail fuel pressure data are first retrieved from the main engine's memory. This data is collected by load sensors, speed sensors, and pressure sensors installed on the main engine and stored in its memory. The acquisition process ensures that the historical operating parameter data covers various typical operating conditions, such as different speeds, load variations, and stable cruise states, to guarantee the representativeness and comprehensiveness of the historical operating parameter dataset. Subsequently, the collected historical main engine load data, historical main engine speed data, and historical common rail fuel pressure data are aligned by timestamp and organized into structured three-dimensional data sets. Each three-dimensional data set consists of a corresponding set of main engine load data, main engine speed data, and common rail fuel pressure data. These three-dimensional data sets are integrated to form a historical operating parameter dataset. This dataset only includes historical operating parameter data during normal main engine operation, excluding records of historical operating parameter data outside the normal operating range, ensuring that the historical operating parameter dataset only reflects the standard operating mode. This serves as the basis for subsequent SVDD model training and monitoring of abnormal ship main engine common rail fuel pressure.

[0030] S2. Preprocess the historical operating parameter data to obtain preprocessed historical operating parameter data.

[0031] Individual cleaning includes identifying and removing outliers caused by pressure sensor malfunctions or electromagnetic interference; filtering high-frequency noise data to reduce random fluctuations; and applying data smoothing methods such as median filtering or moving average filtering to stabilize the historical operating parameter data sequence, generating cleaned historical operating parameter data. Unified preprocessing performs a standardization transformation on all historical operating parameter data, covering cleaned historical main engine load data, historical main engine speed data, and historical common rail fuel pressure data.

[0032] Specifically, firstly, outliers in the historical operating parameter data are removed, noisy data is filtered out, and smoothing is performed to obtain cleaned historical operating parameter data. Then, the values ​​of the cleaned historical operating parameter data are linearly scaled to the 0-1 range through normalization to eliminate dimensional differences, resulting in normalized historical operating parameter data. The normalization method used is min-max normalization, and the formula includes... ,in For the normalized historical operating parameter data, To input the historical running parameter data for the formula, The maximum value in the historical operating parameter data. This represents the minimum value among the historical operating parameter data. The calculation process involves using the above normalization formula to normalize each preprocessed historical host load data, historical host speed data, and historical fuel common rail pressure data. Based on this, all historical operating parameter data within the historical operating parameter dataset are mapped to the range [0,1], ensuring a consistent distribution of historical operating parameter data and facilitating subsequent modeling calculations. Ultimately, the preprocessed historical operating parameter data serves as the input for training the SVDD model.

[0033] S3. Construct and train an SVDD model based on the preprocessed historical operating parameter data. The SVDD model obtains a historical normal data hypersphere in the three-dimensional feature space and obtains the three-dimensional feature position and radius of the center of the historical normal data hypersphere. The volume of the historical normal data hypersphere is defined as the distribution range of the historical operating parameter data in the three-dimensional feature space.

[0034] The preprocessed historical operating parameter data refers to the historical main engine load data, historical main engine speed data, and historical fuel common rail pressure data preprocessed in step S2. The SVDD (Support Vector Data Description) model represents the support vector data description algorithm, used to define the distribution range boundary of historical operating parameter data in a three-dimensional feature space. The historical normal data hypersphere represents a spherical decision boundary, with its center's three-dimensional feature position being the coordinate vector of the hypersphere's center, and its radius being the scalar distance from the hypersphere's boundary to its center.

[0035] Specifically, firstly, an SVDD model framework is constructed based on the preprocessed historical operating parameter dataset. A radial basis function (RBF) kernel is selected as the kernel function of the SVDD model to handle the nonlinear characteristics of the historical operating parameter data. Then, the preprocessed historical operating parameter data is input into the SVDD model as training data. An optimization algorithm is used to solve for the center three-dimensional feature position and radius value of the historical normal data hypersphere. The solution process aims to minimize the volume of this historical normal data hypersphere while constraining all training data points, i.e., historical operating parameter data, to be located inside or on the boundary of the historical normal data hypersphere. After model training, the SVDD model obtains the three-dimensional feature position and radius value of the center of the historical normal data hypersphere. These data collectively define the normal distribution range of the historical operating parameter data in the three-dimensional feature space, which is used for subsequent real-time monitoring of abnormal common rail pressure of ship main engine fuel.

[0036] S4. Preprocess the real-time operating parameter data collected in real time according to the steps in S2, and input the preprocessed real-time operating parameter data into the SVDD model. Calculate the distance between the real-time operating parameter data and the three-dimensional feature position of the center of the historical normal data hypersphere, and compare the magnitude of the distance between the real-time operating parameter data and the center of the historical normal data hypersphere with the magnitude of the radius of the historical normal data hypersphere. The real-time operating parameter data includes real-time main engine load data, real-time main engine speed data, and real-time fuel common rail pressure data.

[0037] Among them, real-time main engine load data represents the current output mechanical power parameters of the ship's main engine. Real-time main engine speed data represents the current rotational speed parameters of the main engine crankshaft. Real-time fuel common rail pressure data represents the current pressure parameters of the high-pressure pipelines in the fuel supply system. Distance refers to the Euclidean distance between the real-time operating parameter data point and the three-dimensional feature position of the center of the historical normal data supersphere, representing the positional deviation of the data point in the three-dimensional feature space.

[0038] Specifically, firstly, real-time host load data, real-time host speed data, and real-time common rail fuel pressure data are acquired in real time through various sensors. Then, following the preprocessing method in S2, the real-time operating parameter data is first cleaned, and then all real-time operating parameter data undergoes a unified minimum-maximum normalization operation, linearly mapping the real-time operating parameter data to the [0,1] range to generate preprocessed real-time operating parameter data. The preprocessed real-time operating parameter data is then input into a trained SVDD model; the trained SVDD model automatically calculates the distance from each real-time operating parameter data point to the three-dimensional feature position of the hypersphere center of the historical normal data. This calculation process uses vector operations in the three-dimensional feature space to obtain the distance value. Then, the distance from each calculated real-time operating parameter data point to the center of the historical normal data hypersphere's three-dimensional feature position is compared with the radius of the historical normal data hypersphere. The comparison process uses numerical comparator logic. If the distance from the real-time operating parameter data point to the center of the historical normal data hypersphere's three-dimensional feature position is less than or equal to the radius of the historical normal data hypersphere, it indicates that the data point is located inside or on the boundary of the historical normal data hypersphere, and the real-time operating parameter data is within the normal range. If the distance from the real-time operating parameter data point to the center of the historical normal data hypersphere's three-dimensional feature position is greater than the radius of the historical normal data hypersphere, it indicates that the data point is outside the boundary of the historical normal data hypersphere, and the real-time operating parameter data is within the abnormal range. This comparison result is directly passed to subsequent judgment steps. It should be noted that the preprocessing step must ensure that the scale range of the real-time operating parameter data is consistent with that of the historical operating parameter data to avoid the units affecting the calculation of the distances mentioned above. The input to the SVDD model requires a three-dimensional feature vector form; therefore, the real-time operating parameter data is strictly organized according to three dimensions: main engine load, main engine speed, and common rail fuel pressure.

[0039] S5. When the distance between the real-time operating parameter data and the center of the three-dimensional feature position of the historical normal data hypersphere is greater than the radius of the historical normal data hypersphere, it is determined that the real-time operating parameter data is within the abnormal range, the common rail pressure of the ship's main engine fuel is in an abnormal state, and an alarm signal is triggered.

[0040] Specifically, when the distance from the data point containing the real-time operating parameter data to the three-dimensional feature position of the center of the historical normal data supersphere is greater than the radius of the historical normal data supersphere, it indicates that the data point is outside the boundary of the historical normal data supersphere, and the fuel common rail pressure is determined to be in an abnormal state. An alarm signal is immediately triggered, which includes an abnormal state timestamp, the coordinates of the data point containing the abnormal real-time operating parameter data, and the degree of abnormality (i.e., the degree of deviation from the normal state). An audible and visual warning is issued through the ship's alarm device, and the abnormal event is recorded. It should be noted that the judgment condition "the distance from the data point containing the real-time operating parameter data to the three-dimensional feature position of the center of the historical normal data supersphere is greater than the radius of the historical normal data supersphere" strictly corresponds to the mathematical definition of the data point containing the abnormal real-time operating parameter data in the three-dimensional feature space, ensuring that the alarm is triggered only when the data point containing the real-time operating parameter data is within the abnormal range. Simultaneously, the alarm signal triggering will interrupt the process of updating the real-time operating parameter data into the historical operating parameter dataset, preventing contamination of the normal historical operating parameter dataset. Meanwhile, the radius of the hypersphere in historical normal data is used as a dynamic boundary threshold. Its value is obtained by training with normal historical operating parameter data, eliminating the need to manually set a fixed empirical threshold, thereby eliminating the risk of missed or false alarms caused by improper threshold setting in traditional methods.

[0041] Specifically, when the distance from the data point containing the real-time operating parameter data to the three-dimensional feature position of the center of the historical normal data hypersphere is less than or equal to the radius of the historical normal data hypersphere, it indicates that the data point containing the real-time operating parameter data is located inside or on the boundary of the historical normal data hypersphere. Therefore, the fuel common rail pressure is determined to be in a normal state, and the data update process is initiated. First, the real-time operating parameter data determined to be within the normal range is timestamped and appended to the historical operating parameter dataset. Then, the model update mechanism is triggered. Using the complete historical operating parameter dataset containing the newly added real-time operating parameter data within the normal range as the new model input, the SVDD model construction and training process in step S3 is re-executed to solve for the new three-dimensional feature position and radius value of the center of the historical normal data hypersphere, thereby dynamically optimizing the boundary range of the distribution of normal historical operating parameter data in the three-dimensional feature space. It should be noted that this data update mechanism ensures that the historical operating parameter dataset continues to expand, enhancing the adaptability of the SVDD model to changes in the ship's main engine operating conditions. The SVDD model update adopts a full retraining strategy, recalculating the hypersphere of the historical normal data based on the updated historical operating parameter dataset, enabling the SVDD model to learn the latest normal operating parameter data. During the update process, the center and radius values ​​of the historical normal data hypersphere will adaptively change with the changes in the historical operating parameter dataset. This adaptive adjustment mechanism effectively avoids the risk of misjudgment caused by the long-term drift of the host's operating state, while maintaining the real-time nature of anomaly monitoring.

[0042] The following is a description of another embodiment of the method provided in this implementation. Please refer to... Figure 2 This is another flowchart illustrating the method for monitoring abnormal pressure in the common rail fuel line of a ship's main engine, as described in this application.

[0043] Based on the method for monitoring abnormal pressure in the common rail fuel line of a ship's main engine as described in Embodiment 1, this Embodiment 2 adds some specific implementation methods.

[0044] In this embodiment, step S203 involves constructing and training an SVDD model based on preprocessed historical operating parameter data. The SVDD model obtains a historical normal data hypersphere in a three-dimensional feature space and obtains the three-dimensional feature position and radius of the center of the historical normal data hypersphere. The volume of the historical normal data hypersphere is defined as the distribution range of the historical operating parameter data in the three-dimensional feature space.

[0045] Specifically, in step S203, during the construction of the SVDD model, the historical OEM load data range is first divided into multiple continuous operating condition intervals based on the historical OEM load data. Each operating condition interval represents a specific OEM operating load segment. Then, within each operating condition interval, the correlation weight between historical fuel common rail pressure data and historical OEM speed data is calculated using Pearson correlation coefficient analysis. This correlation weight quantifies the coupling strength between historical fuel common rail pressure data and historical OEM speed data under the corresponding operating condition. The calculation process is as follows: Let the divided m continuous operating condition intervals be... Where K = 1, 2, m, and m is a positive integer (all instances of K in the following text have this meaning and will not be elaborated further), extract n sets of historical operating parameter data within m consecutive operating condition intervals, and obtain the historical main engine speed data sequence R = {r 1， r 2， r n} and historical common rail pressure data sequence P={p 1， p 2， p n Therefore, it can be done through the formula. ,in This represents the historical average engine speed. ; This represents the historical average common rail pressure for fuel. ; The Pearson correlation coefficient ranges from -1 to 1, and a larger absolute value indicates a stronger linear correlation. Then... Mapped to scaling weights in 3D feature space ,in Based on the weights, For weak correlations, the default weight is usually set to... Based on the aforementioned correlation weights, a scaling factor matrix for the three-dimensional feature space is generated. This matrix is ​​then used to perform a weighted scaling transformation on the historical operating parameter data. This results in the historical main engine load data, historical main engine speed data, and historical common rail fuel pressure data being adjusted according to the relative importance of their correlation under different operating conditions, forming a weighted feature vector set. The weighted feature vectors are then used to train the SVDD model, which in turn drives the SVDD model to construct a hypersphere of historical normal data with the smallest volume in the weighted feature space. The boundary of this historical normal data hypersphere adapts to the three-dimensional feature correlation strength under different operating conditions, forming a distribution range that precisely encompasses the normal historical operating parameter data.

[0046] Step S205: When the distance between the real-time operating parameter data and the center of the three-dimensional feature position of the historical normal data hypersphere is greater than the radius of the historical normal data hypersphere, it is determined that the real-time operating parameter data is within the abnormal range, the common rail pressure of the ship's main engine fuel is in an abnormal state, and an alarm signal is triggered.

[0047] It also includes step S206: when an alarm signal is triggered, real-time common rail vibration spectrum data is acquired simultaneously; if the amplitude of the vibration spectrum at the characteristic frequency exceeds the adaptive threshold, then the fuel common rail pressure is confirmed to be abnormal.

[0048] The alarm signal refers to the alarm signal triggered in step S205 based on the abnormal fuel common rail pressure determination. Real-time common rail vibration spectrum data represents the frequency domain data generated by fast Fourier transform of the vibration time-domain signal collected by vibration sensors installed on the fuel common rail. The characteristic frequency refers to the fundamental frequency and its integer multiples of harmonic frequencies directly associated with the ship's current real-time engine speed data. The adaptive threshold represents the boundary value of the abnormal range dynamically calculated based on historical normal common rail vibration statistical characteristics. The amplitude refers to the quantized value of the energy intensity of the vibration spectrum at the frequency point corresponding to the characteristic frequency.

[0049] Specifically, at the instant an abnormal alarm signal is generated, a high-speed acquisition command is simultaneously sent to the vibration sensor to obtain the raw vibration signal within the current time window. The time-domain signal is then converted to a frequency-domain spectrum using a Fast Fourier Transform algorithm, and the theoretical value of the characteristic frequency is calculated based on real-time main engine speed data. The adaptive threshold is set using a dynamic baseline method, which involves continuously analyzing the statistical distribution of characteristic frequency amplitudes during historical normal operation and multiplying the upper quartile of this distribution by a safety factor as the dynamic threshold boundary. Then, the actual amplitude of the current spectrum at the characteristic frequency is extracted. If the actual amplitude of any characteristic frequency exceeds the corresponding adaptive threshold, an abnormality confirmation command is generated and marked as a mechanical fault level alarm. If the actual amplitude of all characteristic frequencies does not exceed the corresponding adaptive threshold, a downgrade command is sent to the alarm module, downgrading the original ship main engine fuel common rail pressure abnormality alarm to a low-confidence alarm. Simultaneously, a current data snapshot, including current real-time operating parameter data and the vibration spectrum, is frozen and stored in a manual review queue for final confirmation.

[0050] It should be noted that the synchronous acquisition mechanism ensures strict alignment between vibration data and abnormal fuel common rail pressure, preventing erroneous correlations caused by changes in operating conditions. The calculation of characteristic frequencies strictly follows the crankshaft rotation dynamics of the main engine; the fundamental frequency equals the real-time main engine speed divided by the time conversion factor, and harmonic frequencies are extended in integer multiples. The safety factor of the adaptive threshold is automatically adjusted based on the dispersion of historical operating parameter data, suppressing random fluctuation interference while ensuring monitoring sensitivity. This multi-source verification mechanism effectively distinguishes between real mechanical faults and single-point failure scenarios such as sensor drift, improving the reliability of the alarm system.

[0051] In this embodiment, the historical operating parameter dataset obtained in step S101 is divided into a historical normal operating parameter dataset and a historical abnormal operating parameter dataset; the historical normal operating parameter dataset is further divided into a training set and a test subset; the test subset and the historical abnormal operating parameter dataset are combined to generate a test set; wherein, the training set is used to train the SVDD model to construct a historical normal data hypersphere; the test set is used to input into the trained SVDD model to calculate the distance between its distance and the three-dimensional feature position of the center of the historical normal data hypersphere, and based on the comparison result of the distance and the radius of the historical normal data hypersphere, the judgment ability of the SVDD model is tested, and the historical normal data hypersphere constructed by the SVDD model is optimized and improved according to the test results.

[0052] The historical normal operation parameter dataset refers to the set of historical normal operation parameters collected by the ship's main engine under fault-free conditions. The historical abnormal operation parameter dataset refers to the set of historical abnormal operation parameters collected by the ship's main engine under known faults or abnormal operating conditions. The training set is a subset of the historical normal operation parameter dataset used to train the SVDD model; the test subset is a reserved subset of normal data in the historical normal operation parameter dataset that was not used for training; the test set is a set of evaluation data samples composed of the test subset and the historical abnormal operation parameter dataset.

[0053] Specifically, the system first loads the historical operating parameter dataset and, based on the abnormal timestamps marked in the ship's main engine operation logs, separates the historical operating parameter dataset into a historical normal operating parameter dataset and a historical abnormal operating parameter dataset. Then, it performs a random partitioning operation on the historical normal operating parameter dataset, using the majority portion as the training set to construct a historical normal data hypersphere in the SVDD model training process, and the remaining portion as the test subset. Finally, the test subset is combined with the complete historical abnormal operating parameter dataset to form a test set, which contains both historical normal operating parameter samples and known abnormal operating parameter samples.

[0054] It should be noted that the training set is specifically used to solve for the 3D feature position and radius parameters of the center of the historical normal data hypersphere; the test set is used as input to the trained SVDD model to calculate the Euclidean distance from each sample point to the center of the historical normal data hypersphere, and the classification results are generated by comparing the distance values ​​with the radius values ​​of the historical normal data hypersphere. By statistically analyzing the misclassification rate of historical normal operating parameter samples and the detection rate of known abnormal operating parameter samples, the ability of the SVDD model to balance the inclusion of normal data and the exclusion of abnormal data is evaluated. Based on this, the kernel function parameters or data preprocessing strategies are adjusted to optimize the boundary range of the historical normal data hypersphere, ensuring that the historical normal data hypersphere contains as much normal historical operating parameter data as possible, while minimizing the sphere volume to exclude potentially abnormal historical operating parameter data, forming a closed-loop mechanism for model iteration and improvement.

[0055] Reference Figure 3 This is a schematic diagram of a system for monitoring abnormal pressure in the common rail fuel line of a ship's main engine, as described in this application.

[0056] The system for monitoring abnormal pressure in the common rail fuel line of a ship's main engine includes a data acquisition module 1, a data preprocessing module 2, a model training module 3, a real-time monitoring module 4, and an alarm module 5.

[0057] The modules are interconnected via a control bus to form the aforementioned system.

[0058] Data acquisition module 1 is used to acquire historical operating parameter data of the ship's main engine and construct a historical operating parameter dataset.

[0059] The hardware implementation of the data acquisition module 1 is to connect the communication interface card and the data acquisition card that controls the ship's main engine. It receives historical operating parameter data from the main engine load sensor, main engine speed sensor and fuel common rail pressure sensor deployed in the real-time data acquisition unit 41 through the control bus, and writes it into the embedded memory to build a historical operating parameter dataset.

[0060] Data preprocessing module 2 is used to clean historical fuel common rail pressure data separately and then preprocess historical operating parameter data uniformly.

[0061] Model training module 3 is used to build and train the SVDD model based on preprocessed historical running parameter data, and to update the trained SVDD model after real-time running parameter data is added to the historical running parameter dataset.

[0062] The real-time monitoring module 4 is used to collect real-time operating parameter data, process the real-time operating parameter data according to the preprocessing steps of the data preprocessing module 2, and input the preprocessed real-time operating parameter data into the SVDD model; calculate the distance between the real-time operating parameter data and the three-dimensional feature position of the center of the hypersphere of historical normal data, and compare the magnitude of the distance value with the magnitude of the hypersphere radius value of historical normal data.

[0063] The real-time monitoring module 4 consists of a real-time data acquisition unit 41 and a distance calculation unit 42. The real-time data acquisition unit 41 acquires the real-time sensor data of each deployed sensor and reuses the cleaning and normalization process of the preprocessing module. The distance calculation unit 42 calls the historical normal data hypersphere parameters stored in the model training module 3 to perform Euclidean distance calculation and radius comparison and outputs an alarm signal.

[0064] Alarm module 5 is used to trigger an alarm signal when the distance value of the three-dimensional feature position of the center of the sphere in historical normal data is greater than or equal to the radius value.

[0065] The alarm module 5 integrates a relay output unit 51 and an audible and visual alarm unit 52. When the real-time monitoring module 4 outputs an alarm signal, it triggers the relay output unit 51 to connect the alarm circuit and drives the audible and visual alarm unit 52 to perform an audible and visual alarm.

[0066] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0067] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows. Figure 4 As shown. The computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the computer device contains data related to a method for monitoring abnormal pressure in the common rail of marine main engine fuel. The network interface of the computer device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for monitoring abnormal pressure in the common rail of marine main engine fuel. The electronic device includes a processor, memory, and network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the electronic device stores data. The network interface of the electronic device is used for communication with other external terminals or servers via a network connection. In some embodiments, the network interface may be a wired network interface; in other embodiments, it may be a wireless network interface. When the computer program is executed by a processor, it implements the methods described in the embodiments of this application.

[0068] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0069] Unless otherwise defined, the technical or scientific terms used in this application shall have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms "first," "second," "third," and similar terms used in this application specification and claims do not indicate any order, quantity, or importance, but are merely used to distinguish different components. The terms "an" or "a" and similar terms do not indicate a quantity limitation, but rather indicate the presence of at least one. The terms "comprising" or "including" and similar terms mean that the elements or objects preceding "comprising" or "including" encompass the elements or objects listed following "comprising" or "including" and their equivalents, and do not exclude other elements or objects. "Above," "below," "left," "right," etc., are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0070] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for monitoring abnormal pressure in the common rail of marine main engine fuel, characterized in that, include: S1. Obtain historical operating parameter data of the ship's main engine and construct a historical operating parameter dataset; The historical operating parameter data includes historical main engine load data, historical main engine speed data, and historical fuel common rail pressure data; S2. Preprocess the historical operating parameter data to obtain preprocessed historical operating parameter data; S3. Construct and train an SVDD model based on the preprocessed historical operating parameter data. The SVDD model obtains a historical normal data hypersphere in the three-dimensional feature space and obtains the three-dimensional feature position and radius of the center of the historical normal data hypersphere. The volume of the historical normal data hypersphere is defined as the distribution range of the historical operating parameter data in the three-dimensional feature space; S4. Preprocess the real-time collected operating parameter data according to the steps in S2, and input the preprocessed real-time operating parameter data into the SVDD model. Calculate the distance between the real-time operating parameter data and the three-dimensional feature position of the center of the historical normal data hypersphere, and compare the magnitude of the distance between the real-time operating parameter data and the center of the historical normal data hypersphere with the magnitude of the radius of the historical normal data hypersphere. The real-time operating parameter data includes real-time main engine load data, real-time main engine speed data, and real-time fuel common rail pressure data; S5. When the distance between the real-time operating parameter data and the center of the three-dimensional feature position of the historical normal data hypersphere is greater than the radius of the historical normal data hypersphere, it is determined that the real-time operating parameter data is within the abnormal range, the common rail pressure of the ship's main engine fuel is in an abnormal state, and an alarm signal is triggered.

2. The method for monitoring abnormal pressure in the common rail fuel line of a ship's main engine according to claim 1, characterized in that, Step S5 further includes determining that the real-time operating parameter data is within the normal range and the ship's main engine fuel common rail pressure is in a normal state when the distance between the real-time operating parameter data and the center of the historical normal data hypersphere is less than or equal to the radius of the historical normal data hypersphere, adding the real-time operating parameter data to the historical operating parameter dataset, and updating the SVDD model based on this.

3. The method for monitoring abnormal pressure in the common rail of marine main engine fuel according to claim 1, characterized in that: The preprocessing of the historical operating parameter data includes removing outliers, filtering noisy data, and smoothing the data to obtain cleaned historical operating parameter data. Then, the values ​​of the cleaned historical operating parameter data are mapped to the same normalized range through normalization processing to obtain the normalized historical operating parameter data.

4. The method for monitoring abnormal pressure in the common rail of marine main engine fuel according to claim 1, characterized in that: Step S3 includes selecting a radial basis kernel function to construct the SVDD model; inputting the preprocessed historical running parameter data into the SVDD model; training the SVDD model to obtain the historical normal data hypersphere in the three-dimensional feature space; and calculating and solving the three-dimensional feature position and radius of the center of the historical normal data hypersphere, so that all historical running parameter data in the historical running parameter dataset that are within the normal range are located inside or on the boundary of the historical normal data hypersphere.

5. The method for monitoring abnormal pressure in the common rail of marine main engine fuel according to claim 4, characterized in that: Based on historical host load data, multiple operating condition intervals are divided, and the correlation weight between historical fuel common rail pressure data and historical host speed data in each interval is calculated. Based on the correlation weight, the historical operating parameter data in the three-dimensional feature space is weighted and scaled to generate a weighted feature vector. The weighted feature vector is used to train the SVDD model and construct a hypersphere of historical normal data.

6. The method for monitoring abnormal pressure in the common rail fuel line of a ship's main engine according to claim 1, characterized in that: It also includes step S6: when an alarm signal is triggered, real-time common rail vibration spectrum data is acquired simultaneously; if the amplitude of the vibration spectrum at the characteristic frequency exceeds the adaptive threshold, then the fuel common rail pressure is confirmed to be abnormal.

7. The method for monitoring abnormal pressure in the common rail fuel line of a ship's main engine according to claim 1, characterized in that: The historical operating parameter dataset obtained in step S1 is divided into a historical normal operating parameter dataset and a historical abnormal operating parameter dataset. The historical normal operation parameter dataset is divided into a training set and a test subset. The test subset is combined with the historical abnormal operation parameter dataset to generate a test set. The training set is used to train the SVDD model to construct a historical normal data hypersphere. The test set is used to input into the trained SVDD model to calculate the distance between the model and the three-dimensional feature position of the center of the historical normal data hypersphere. Based on the comparison between this distance and the radius of the historical normal data hypersphere, the judgment ability of the SVDD model is tested. The historical normal data hypersphere constructed by the SVDD model is optimized and improved based on the test results.

8. A system for monitoring abnormal pressure in the common rail of marine main engine fuel, characterized in that: It includes a data acquisition module, a data preprocessing module, a model training module, a real-time monitoring module, and an alarm module; The data acquisition module is used to acquire historical operating parameter data of the ship's main engine and construct a historical operating parameter dataset; The data preprocessing module is used to uniformly preprocess historical operating parameter data; The model training module is used to build and train the SVDD model based on preprocessed historical running parameter data, and to update the trained SVDD model after real-time running parameter data is added to the historical running parameter dataset; The real-time monitoring module is used to collect real-time operating parameter data, process the real-time operating parameter data according to the preprocessing steps of the data preprocessing module, and input the preprocessed real-time operating parameter data into the SVDD model; calculate the distance between the real-time operating parameter data and the three-dimensional feature position of the center of the hypersphere of historical normal data, and compare the relationship between the distance value and the radius value of the hypersphere of historical normal data. The alarm module is used to trigger an alarm signal when the distance value of the three-dimensional feature position of the center of the sphere in historical normal data is greater than or equal to the radius value.

Images