A remote fault diagnosis method and system for a ballast water system

Abstract

The application relates to the technical field of industrial equipment fault diagnosis, and discloses a ballast water system remote fault diagnosis method and system, the method comprising the following steps: acquiring time series of multi-dimensional parameters such as pressure, flow rate and temperature; performing standardization and principal component analysis dimension reduction on the time series to obtain low-dimensional characteristic sequences; locking fluctuation intervals of the low-dimensional characteristics through sliding window analysis; constructing a cumulative distance matrix and searching for an optimal matching path for the pressure and flow rate sequences in the intervals to calculate a time series correlation strength coefficient; and finally determining a fault mode in combination with the coefficient and the change trend of each sequence. Through the technical path combining principal component analysis and dynamic time warping, the application realizes accurate identification and diagnosis of early and complex fault modes of the ballast water system, and significantly improves the reliability and safety of system operation.

Description

Technical Field

[0001] This invention relates to the field of industrial equipment fault diagnosis technology, and in particular to a remote fault diagnosis method and system for ballast water systems. Background Technology

[0002] Ballast water systems are critical equipment for ensuring the safety and balance of ship navigation, and their operational reliability directly affects the seaworthiness and operational efficiency of the vessel. As the shipping industry moves towards intelligent and unmanned operations, accurate and real-time remote fault diagnosis of core equipment such as ballast water systems has become an important technical support for improving ship management and safety assurance capabilities.

[0003] Currently, a typical method for ballast water system fault diagnosis is based on Multivariate Statistical Process Control (MSPC), such as using Principal Component Analysis (PCA) to construct a statistical model of the system during normal operation. Specifically, this method first collects a large number of multidimensional operating parameters (such as pressure, flow rate, and temperature) of the system under healthy conditions, and then uses the PCA model to project these parameters into a low-dimensional feature space, defining control limits representing normal operation (such as control limits for Hotelling's T² and SPE statistics). In actual monitoring, the parameters collected in real time are also projected onto this model. Once the statistics at the projected point exceed the preset control limits, the system will trigger a fault alarm. However, the aforementioned existing technical solutions based on statistical models have inherent technical limitations in terms of diagnostic accuracy and timeliness. While this method can effectively determine whether a system fault has occurred, its diagnostic capabilities regarding "what kind of fault occurred" and "the root cause of the fault" are very limited. It combines multiple parameters with clear physical meanings, such as pressure and flow rate, into one or a few statistical indicators lacking clear physical meaning. This makes it difficult to deduce the specific fault mode (e.g., whether it's pump impeller wear or valve sealing surface damage) from these abstract statistics when a fault occurs. Maintenance personnel often only receive a general "deviation from normal state" alarm, without obtaining precise fault location information. This severely delays subsequent maintenance decisions and reduces the practical value of remote diagnostics.

[0004] Therefore, existing technologies suffer from the technical problem of losing physical meaning due to feature mixing, making it difficult to achieve accurate fault mode recognition. Summary of the Invention

[0005] This invention provides a remote fault diagnosis method and system for ballast water systems to achieve accurate identification of specific fault modes under complex operating conditions.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a remote fault diagnosis method for a ballast water system, comprising: Obtain multi-dimensional operating parameters of the ship's ballast water system, including time series of pressure, flow, and temperature parameters. The time series of the multidimensional operating parameters are subjected to zero-mean-unit-variance standardization to obtain a standardized multidimensional time series. Principal component analysis and dimensionality reduction are performed on the standardized multidimensional time series to obtain a low-dimensional feature vector sequence. A sliding window analysis is performed on the low-dimensional feature vector sequence to determine the feature fluctuation range where the amplitude fluctuation exceeds a preset fluctuation threshold; Extract the time series of pressure and flow parameters within the characteristic fluctuation range, and construct the cumulative distance matrix between them based on the time series; The optimal matching path is searched on the cumulative distance matrix, and the time series of the pressure parameter and flow parameter are aligned and correlated based on the optimal matching path to calculate the time series correlation strength coefficient. The failure modes of the ballast water system are obtained based on the respective changing trends of the time-series correlation strength coefficient, the pressure parameter, and the flow rate parameter within the characteristic fluctuation range.

[0007] Preferably, the step of performing zero-mean-unit-variance standardization on the time series of the multidimensional operating parameters to obtain a standardized multidimensional time series includes: For the time series of pressure, flow rate, and temperature among the multidimensional operating parameters, calculate their respective mean and standard deviation; The calculated mean and standard deviation are used to perform numerical normalization on the time series, mapping the original values ​​to a standard interval, thereby obtaining a standardized multidimensional time series.

[0008] Preferably, the step of performing principal component analysis and dimensionality reduction on the standardized multidimensional time series to obtain a low-dimensional feature vector sequence includes: Based on the pressure, flow rate, and temperature time series in the standardized multidimensional time series, the covariance matrix among these three parameters is calculated. Perform eigenvalue decomposition on the covariance matrix to obtain a set of eigenvalues ​​and a corresponding set of eigenvectors; The cumulative contribution rate of each principal component is calculated based on the eigenvalues, and based on the comparison result of the cumulative contribution rate and the preset threshold, k eigenvectors corresponding to the first k principal components to be retained are selected from the eigenvector set. A dimensionality reduction transformation matrix is ​​constructed using the eigenvectors, and the dimensionality reduction transformation matrix is ​​applied to project the standardized multidimensional time series to obtain a low-dimensional eigenvector sequence.

[0009] Preferably, the step of performing sliding window analysis on the low-dimensional feature vector sequence to determine the feature fluctuation range where the amplitude fluctuation exceeds a preset fluctuation threshold includes: Based on the low-dimensional feature vector sequence, the variance explained ratio at each time point is calculated to obtain the variance explained ratio time series. A fixed-length sliding window is used to scan the variance explanation ratio time series, and the difference between the maximum and minimum variance explanation ratios within the sliding window is calculated as the amplitude fluctuation. When the amplitude fluctuation exceeds a preset fluctuation threshold, the time range covered by the current sliding window is determined as the characteristic fluctuation interval.

[0010] Preferably, the step of extracting the time series of pressure and flow parameters within the characteristic fluctuation range, and constructing a cumulative distance matrix between them based on the time series, includes: The time series of pressure and flow parameters within the characteristic fluctuation range are length aligned and numerically normalized to obtain standardized pressure and flow feature vectors. Calculate the Euclidean distance between the standardized pressure feature vector and the flow feature vector at each time point, and construct the cumulative distance matrix between them.

[0011] Preferably, the step of searching for an optimal matching path on the cumulative distance matrix and aligning and quantifying the correlation of the time series of the pressure and flow parameters based on the optimal matching path to calculate the time series correlation strength coefficient includes: In the cumulative distance matrix, the optimal matching path with the minimum total cost is determined by backtracking through dynamic programming. Based on the optimal matching path, path-guided resampling is performed to obtain two aligned time series of the same length. The aligned time series is divided into multiple time periods, and the Pearson correlation coefficient for each time period is calculated. Finally, the time series correlation strength coefficient is obtained by weighted average algorithm.

[0012] Preferably, obtaining the failure mode of the ballast water system based on the respective changing trends of the time-series correlation strength coefficient, the pressure parameter, and the flow rate parameter within the characteristic fluctuation range includes: The temporal correlation strength coefficient is compared with a preset first threshold and a second threshold, wherein the first threshold is less than the second threshold; When the time-series correlation strength coefficient is less than the first threshold and the time series of the pressure parameter shows a downward trend, the fault mode is judged as pump impeller wear fault. When the time-series correlation strength coefficient is greater than the second threshold and the time series of the flow parameters exhibits periodic pulsation, the fault mode is judged as valve sealing surface damage fault. When the temporal correlation strength coefficient is between the first threshold and the second threshold, the numerical distance between the temporal correlation strength coefficient and the first threshold and the second threshold is calculated respectively, and the weight for comprehensive evaluation is determined based on the numerical distance to finally determine the fault mode.

[0013] Secondly, the present invention provides a remote fault diagnosis system for a ballast water system, comprising: The parameter acquisition module is used to acquire multi-dimensional operating parameters of the ship's ballast water system, including time series of pressure parameters, flow parameters, and temperature parameters. The standardization module is used to perform zero-mean-unit-variance standardization on the time series of the multidimensional operating parameters to obtain a standardized multidimensional time series. The feature reduction module is used to perform principal component analysis and dimensionality reduction on the standardized multidimensional time series to obtain a low-dimensional feature vector sequence. The fluctuation detection module is used to perform sliding window analysis on the low-dimensional feature vector sequence to determine the feature fluctuation range in which the amplitude fluctuation exceeds a preset fluctuation threshold. The cost matrix construction module is used to extract the time series of pressure parameters and flow parameters within the characteristic fluctuation range, and construct the cumulative distance matrix between the two based on the time series. The correlation strength quantization module is used to search for the optimal matching path on the cumulative distance matrix, and to align and quantify the correlation of the time series of the pressure parameter and flow parameter based on the optimal matching path, and to calculate the time series correlation strength coefficient. The fault diagnosis module is used to obtain the fault mode of the ballast water system based on the respective changing trends of the time-series correlation strength coefficient, the pressure parameter, and the flow rate parameter within the characteristic fluctuation range.

[0014] Thirdly, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the remote fault diagnosis method for ballast water systems described in any one of the above.

[0015] Fourthly, the present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the remote fault diagnosis method for ballast water systems described in any one of the above.

[0016] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention first performs principal component analysis (PCA) to reduce the dimensionality of the multidimensional operating parameters, and then uses sliding window analysis to lock the fluctuation range of the low-dimensional features. Finally, it returns to the original physical parameter space corresponding to the range for in-depth analysis. This two-stage strategy of "coarse first and then fine" and "from abstract to concrete" can use the efficiency of dimensionality reduction analysis to quickly locate the potential fault time range. At the same time, by performing subsequent refined diagnosis in the original parameter space, the clear physical interpretability of parameters such as pressure and flow is preserved, thereby realizing the rapid locking of the fault occurrence time and the accurate tracing of the root cause of the fault.

[0017] (2) This invention introduces an optimal matching path search based on Dynamic Time Warping (DTW) to handle the dynamic correlation between core parameters under unsteady conditions. By constructing a cumulative distance matrix between pressure and flow time series and using dynamic programming to search for the optimal matching path, this invention can "bend" the time axis to find the best correspondence between two time series with dynamic delays and morphological differences. Based on this, this invention can calculate a "time series correlation strength coefficient" that is not limited by linear assumptions and better reflects the inherent physical correlation between the two, providing a robust and reliable quantitative indicator for subsequent fault mode judgment.

[0018] (3) This invention constructs a set of composite fault decision rules based on a combination of quantitative correlation and physical change trends to improve the accuracy of fault mode identification. This invention compares the time-series correlation strength coefficient calculated in the previous step with a preset threshold to perform preliminary fault classification. Then, within each classification, it further combines the pressure and flow parameters within the abnormal interval with their respective physically significant change trends (such as monotonically decreasing or periodic pulsation) for logical judgment. This decision-making method, which combines data-driven quantitative indicators with mechanism-based physical trends, can effectively distinguish fault modes that are similar in a single dimension but have different underlying causes, thereby achieving high-precision identification of specific fault types. Attached Figure Description

[0019] Figure 1 This is a schematic flowchart of a remote fault diagnosis method for a ballast water system provided in the first embodiment of the present invention; Figure 2This is a schematic diagram of a remote fault diagnosis system for a ballast water system provided in the second embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Reference Figure 1 The first embodiment of the present invention provides a remote fault diagnosis method for a ballast water system, comprising the following steps: S11, Obtain multi-dimensional operating parameters of the ship's ballast water system, including time series of pressure parameters, flow parameters, and temperature parameters; S12, perform zero-mean-unit-variance standardization on the time series of the multidimensional operating parameters to obtain a standardized multidimensional time series; S13, perform principal component analysis and dimensionality reduction on the standardized multidimensional time series to obtain a low-dimensional feature vector sequence; S14, Perform sliding window analysis on the low-dimensional feature vector sequence to determine the feature fluctuation range in which the amplitude fluctuation exceeds the preset fluctuation threshold; S15, extract the time series of pressure parameters and flow parameters within the characteristic fluctuation range, and construct the cumulative distance matrix between them based on the time series; S16, search for the optimal matching path on the cumulative distance matrix, and align and quantify the time series of the pressure parameter and flow parameter based on the optimal matching path to calculate the time series correlation strength coefficient. S17. Based on the respective changing trends of the time-series correlation strength coefficient, the pressure parameter, and the flow rate parameter within the characteristic fluctuation range, the failure mode of the ballast water system is obtained.

[0022] In step S11, multi-dimensional operating parameters of the ship's ballast water system are obtained, including time series of pressure parameters, flow parameters, and temperature parameters.

[0023] Specifically, the process begins by simultaneously acquiring time-series signals of pressure (e.g., 0 to 1.6 MPa), flow rate (e.g., 0 to 500 m³ / h), and temperature using sensor arrays deployed at key nodes of the ballast water system (e.g., inlets, main pipelines, and pump houses). These raw analog signals are then converted from analog to digital via a data acquisition card at a uniform sampling frequency (e.g., 1000 Hz) and rigorously aligned according to millisecond-accurate timestamps to form a time-synchronized multidimensional data matrix. To filter out electromagnetic or mechanical vibration noise that sensors are susceptible to in dynamic navigation environments, this step further employs a Kalman filter algorithm to process the aligned numerical sequence. Through an iterative process of prediction and updating, the Kalman filter optimally estimates the true state of the signal, effectively smoothing the data. Finally, the filtered time-series data is organized to obtain a high-fidelity raw multidimensional operating parameter time series containing pressure, flow rate, and temperature parameters, laying the data foundation for subsequent standardization and feature analysis.

[0024] In step S12, the time series of the multidimensional operating parameters is subjected to zero-mean-unit variance standardization to obtain a standardized multidimensional time series.

[0025] Specifically, this step aims to eliminate dimensional differences between multidimensional operating parameters caused by variations in physical units and numerical ranges, providing a fair and comparable data foundation for subsequent principal component analysis. In a preferred embodiment, this process is performed independently for each time series, such as pressure, flow rate, and temperature.

[0026] First, for each time series (e.g., a pressure parameter series), calculate the arithmetic mean of the series within the current analysis time window. and standard deviation For example, if the original pressure value within a time window ranges from 0.8 to 1.2 MPa, its mean can be calculated. The standard deviation is 1.0 MPa. The value is 0.1 MPa. Then, using the calculated mean... and standard deviation For each raw value in this time series The transformation is performed using the zero-mean-unit-variance standardization (Z-scoreNormalization) formula: in, This represents the original measurement value at the i-th time point in the time series. This represents the arithmetic mean of the time series over the entire analysis window. This represents the standard deviation of the time series over the entire analysis window. This represents the new value at time point i after standardization. The core principle of this formula is to map the original values ​​onto a new distribution, such that the mean of the new sequence is 0 and the variance is 1. Through this process, the aforementioned pressure values ​​will be mapped to a standard numerical range, for example, -2 to +2. The same calculation process is performed on the flow rate and temperature time series.

[0027] In addition, an outlier detection and correction step is included before standardization. For example, when a pressure value (e.g., 3.5 MPa) is detected to be far beyond its physical range or normal fluctuation range, a weighted average of the values ​​before and after it can be used to correct it, ensuring the continuity and reliability of the data series and avoiding the contamination of subsequent statistical calculations (such as the mean and standard deviation) by extreme outliers. Through the above processing, a standardized multidimensional time series with eliminated dimensional influences and a more stable data distribution is finally obtained.

[0028] In step S13, principal component analysis and dimensionality reduction are performed on the standardized multidimensional time series to obtain a low-dimensional feature vector sequence.

[0029] It should be noted that, in the first embodiment of the present invention, this processing step specifically includes steps S131 to S134: S131, Based on the pressure, flow rate, and temperature time series in the standardized multidimensional time series, the covariance matrix among these three is calculated; S132, Perform eigenvalue decomposition on the covariance matrix to obtain the set of eigenvalues ​​and the corresponding set of eigenvectors; S133, calculate the cumulative contribution rate of each principal component based on the feature value, and select the k feature vectors corresponding to the first k principal components to be retained from the feature vector set based on the comparison result of the cumulative contribution rate and the preset threshold. S134, construct a dimension reduction transformation matrix using the k feature vectors, and project the standardized multidimensional time series onto the matrix using the dimension reduction transformation matrix to obtain a low-dimensional feature vector sequence.

[0030] In step S131, the covariance matrix among pressure, flow rate, and temperature time series in the standardized multidimensional time series is calculated. Specifically, this step aims to quantify the linear correlation between different operating parameters. In a preferred embodiment, a 3x3 covariance matrix is ​​constructed based on the standardized multidimensional time series obtained in step S12. The matrix is ​​constructed as follows: the diagonal elements (row i, column i) represent the variance of the i-th parameter itself, reflecting the fluctuation range of that parameter; while the off-diagonal elements (row i, column j) represent the covariance between the i-th and j-th parameters, reflecting the cooperative relationship between these two parameters. For example, under normal operating conditions, the covariance between pressure and flow rate may be 0.65, showing a strong positive correlation, while the covariance between temperature and pressure may be only 0.23, showing a weak correlation.

[0031] In step S132, eigenvalue decomposition is performed on the covariance matrix to obtain a set of eigenvalues ​​and a corresponding set of eigenvectors. Specifically, this step aims to find the main direction of data variation (principal components). In a preferred embodiment, the Jacobi eigenvalue algorithm is used to perform eigenvalue decomposition on the covariance matrix obtained in the previous step. This algorithm iteratively approaches the off-diagonal elements to zero through a series of planar rotation transformations, thereby diagonalizing the covariance matrix. The convergence threshold of the iteration can be preset, for example, 0.001. The iteration stops when the change in eigenvalues ​​calculated in two adjacent iterations is less than this threshold. The result of the decomposition is: a set containing three eigenvalues ​​(e.g., {2.45, 1.32, 0.78}), and a set containing three corresponding, mutually orthogonal eigenvectors. The magnitude of each eigenvalue represents the variance of the data in the direction of the corresponding eigenvector. To ensure the strict orthogonality of the feature vectors, the Gram-Schmidt process can be used for post-processing verification and correction.

[0032] In step S133, the cumulative contribution rate of each principal component is calculated based on the eigenvalues. Based on the comparison between the cumulative contribution rate and a preset threshold, k eigenvectors corresponding to the top k principal components to be retained are selected from the eigenvector set. Specifically, this step aims to determine the target dimension k after dimensionality reduction, achieving a balance between preserving the original data information to the greatest extent and achieving effective dimensionality reduction. First, the variance contribution rate of each principal component is calculated. This contribution rate is obtained by dividing the eigenvalue corresponding to the principal component by the sum of all eigenvalues; it precisely quantifies the percentage of the total variance of the original data that the principal component can explain. Then, the eigenvalues ​​are arranged in descending order, and their variance contribution rates are accumulated sequentially to obtain the cumulative contribution rate. This cumulative contribution rate is compared with a preset contribution rate threshold, which is a value recognized in the art for achieving a balance between information preservation and dimensionality reduction, for example, it can be set to 85% or 95%. The minimum number of principal components that allow the cumulative contribution rate to first reach or exceed this threshold is selected; this number is k. For example, if the cumulative contribution rate of the first two principal components has reached 92.8% (more than 85%), then the number of principal components to be retained, k, is determined to be 2.

[0033] In step S134, a dimensionality reduction transformation matrix is ​​constructed using k eigenvectors, and this matrix is ​​applied to project the standardized multidimensional time series, thereby obtaining a low-dimensional eigenvector sequence. Specifically, this step performs the final dimensionality reduction operation. First, the first k eigenvectors determined in the previous step are selected from the eigenvector set obtained in step S132. These k eigenvectors are then arranged in columns to construct a... The dimension reduction transformation matrix of dimension 1. Then, the dimension 1 obtained in step S12 is... The standardized multidimensional time series matrix (where n is the length of the time series), and the... The dimension reduction transformation matrix is ​​used for matrix multiplication. The mathematical essence of this projection operation is to linearly transform the original 3D data points and project them onto a k-dimensional subspace composed of the first k principal components. The result of the operation is a dimensionless subspace. A new matrix is ​​generated, where each row of the matrix represents the low-dimensional feature vector of the original multidimensional parameters at the corresponding time point. Finally, these low-dimensional feature vectors are arranged in chronological order to obtain the final sequence of low-dimensional feature vectors.

[0034] In step S14, a sliding window analysis is performed on the low-dimensional feature vector sequence to determine the feature fluctuation range where the amplitude fluctuation exceeds the preset fluctuation threshold.

[0035] It should be noted that, in the first embodiment of the present invention, this processing step specifically includes steps S141 to S143: S141, Based on the low-dimensional feature vector sequence, calculate the variance explained ratio at each time point to obtain the variance explained ratio time series; S142, a fixed-length sliding window is used to scan the variance explanation ratio time series, and the difference between the maximum and minimum variance explanation ratios within the sliding window is calculated as the amplitude fluctuation. S143, when the amplitude fluctuation exceeds the preset fluctuation threshold, the time range covered by the current sliding window is determined as the characteristic fluctuation interval.

[0036] In step S141, the variance explained proportion at each time point is calculated based on the low-dimensional eigenvector sequence, resulting in a variance explained proportion time series. Specifically, this step aims to transform a multi-dimensional eigenvector sequence into a one-dimensional time series that quantifies the degree to which the system "deviations from the norm" at each time point. In a preferred embodiment, for each time point in the low-dimensional eigenvector sequence, its variance explained proportion is calculated as the squared Euclidean norm of the low-dimensional eigenvector at that time point (i.e., the sum of the squares of the vector components). Since the data has been projected onto the principal component space centered at the origin in step S13, the magnitude of this squared norm directly reflects the degree to which the system state deviates from its statistical center at that time, i.e., its contribution to the overall variance at that time. By performing this calculation on all time points, a one-dimensional variance explained proportion time series can be obtained; for example, a series containing 600 time points may have a value range between 0.15 and 0.89.

[0037] In step S142, a fixed-length sliding window is used to scan the variance explained ratio time series, and the difference between the maximum and minimum variance explained ratios within the sliding window is calculated as the amplitude fluctuation. Specifically, this step aims to detect local drastic changes in the variance explained ratio time series. In a preferred embodiment, a fixed-length sliding window is used to traverse the variance explained ratio time series obtained in the previous step, with each time step shift. The length L of this window is a key parameter, for example, it can be set to 30 sampling points. The length L is set based on the duration of a typical fault transient process in a ballast water system, ensuring that the window is long enough to completely capture an abnormal event, while also being short enough to ensure real-time detection. At each window position, this step finds the maximum and minimum values ​​among all 30 values ​​within the window and calculates the difference between them, which is defined as the "amplitude fluctuation" of the current window.

[0038] In step S143, when the amplitude fluctuation exceeds the preset fluctuation threshold, the time range covered by the current sliding window is determined as the characteristic fluctuation interval.

[0039] Specifically, this step is the final decision-making stage for determining the abnormal range. The amplitude fluctuation calculated in the previous step is compared with a preset fluctuation threshold, for example, 0.3. This fluctuation threshold is set by applying this method to a large amount of historical normal operation data, calculating the statistical distribution of amplitude fluctuations under normal operating conditions, and taking, for example, the 99th percentile value. This ensures that the threshold can effectively distinguish between random fluctuations under normal operating conditions and potential fault anomalies. If the amplitude fluctuation calculated in the current window (e.g., 0.37) is greater than the preset threshold (0.3), it is determined that the system has undergone a significant state change within this time period. At this time, the time range covered by the current sliding window is defined by its start and end timestamps (e.g., from 180.0 seconds to 183.0 seconds), and this range is determined as a characteristic fluctuation range and passed to the subsequent refined diagnostic module.

[0040] In step S15, the time series of pressure parameters and flow parameters within the characteristic fluctuation range are extracted, and the cumulative distance matrix between the two is constructed based on the time series.

[0041] It should be noted that, in the first embodiment of the present invention, this processing step specifically includes steps S151 to S153: S151, preprocess the time series of pressure and flow parameters within the characteristic fluctuation range to obtain standardized pressure feature vectors and flow feature vectors; S152, calculate the local cost between all time point pairs of the standardized pressure feature vector and flow feature vector, and construct the local cost matrix; S153, based on the local cost matrix, the cumulative distance matrix is ​​calculated through dynamic programming.

[0042] In step S151, the time series of pressure and flow parameters within the characteristic fluctuation range are preprocessed to obtain standardized pressure and flow feature vectors. Specifically, this step aims to preprocess the original time series used for refined diagnostics to ensure their comparability. This process includes: Length Alignment: Due to slight differences in sampling rates between different sensors, the extracted pressure and flow time series within the same characteristic fluctuation range may not have completely consistent lengths (e.g., one with 1000 data points and the other with 998). To address this issue, in a preferred embodiment, a cubic spline interpolation algorithm is used to resample the shorter time series, making its length completely consistent with the longer series.

[0043] Numerical normalization: To eliminate the significant dimensional differences between pressure (megapascals) and flow rate (cubic meters per hour), this step performs min-max normalization on both length-aligned time series. This method uses a linear transformation to map the values ​​of each series to a uniform, for example, interval.

[0044] After these two processing steps, two standardized pressure feature vectors and flow feature vectors with equal length and comparable numerical ranges are finally obtained.

[0045] In step S152, the local cost between all time point pairs of the standardized pressure feature vector and flow feature vector is calculated, and a local cost matrix is ​​constructed. Specifically, this step is the first step of dynamic time warping analysis, that is, calculating the "cost" of matching all possible point pairs between two time series. In a preferred embodiment, the process is as follows: First, create a... A two-dimensional matrix is ​​constructed, where M and N are the lengths of the standardized pressure and flow feature vectors, respectively. This matrix is ​​called the Local Cost Matrix or Distance Matrix. Then, the value of each element C(i,j) in this matrix is ​​calculated. This value is defined as the numerical value of the pressure feature vector at the i-th time point. The value at the j-th time point of the flow feature vector The square Euclidean distance between them, i.e. For example, if , Then the value of element C(50,50) in the local cost matrix is ​​(0.6-0.2)² = 0.16. By traversing all combinations of i and j (from 1 to M / N) and filling the entire matrix, a basic matrix recording the matching cost between all point pairs is obtained.

[0046] In step S153, the cumulative distance matrix is ​​calculated using dynamic programming based on the local cost matrix.

[0047] Specifically, this step is the core of dynamic time warping analysis, aiming to find a path from the bottom left corner to the top right corner of the matrix that minimizes the total cost. In a preferred embodiment, this process is implemented using dynamic programming, generating a cumulative distance matrix D of the same size as the local cost matrix. The value of each element D(i,j) of this cumulative distance matrix D represents the cumulative cost of the optimal path (i.e., the path with the minimum cost) from the starting point (1,1) to the current point (i,j).

[0048] The matrix filling process includes two stages: boundary initialization and recursive calculation. Boundary condition initialization: First, initialize the starting point: ,in It is the first element of the local cost matrix calculated in the previous step.

[0049] Then, initialize the first row: for all other elements in the first row. Its cumulative cost can only be accumulated from its left-hand neighbor, therefore .

[0050] Similarly, initialize the first column: for all other elements in the first column Its cumulative cost can only be accumulated from its lower neighbors, therefore .

[0051] Recursive computation: After boundary initialization, the computation of all other internal elements D(i,j) in the matrix (where i>1 and j>1) follows a recursive relation: Where C(i,j) is the cost at the corresponding position in the local cost matrix. This formula means that the optimal path to point (i,j) must pass through the point with the minimum cumulative cost among its three adjacent predecessor points (i-1,j), (i,j-1), or (i-1,j-1).

[0052] By first initializing the boundary and then filling the entire interior of matrix D row by row and column by column according to the recursive relationship, the final matrix D is the cumulative distance matrix containing the optimal cumulative cost of all possible paths. The last element D(M,N) of this matrix is ​​the shortest normalized distance between the two time series.

[0053] In step S16, an optimal matching path is searched on the cumulative distance matrix, and the time series of pressure and flow parameters are aligned and their correlation quantified based on the optimal matching path to calculate the time series correlation strength coefficient. It should be noted that, in the first embodiment of the present invention, this processing step specifically includes steps S161 to S163: S161, On the cumulative distance matrix, the optimal matching path with the minimum total cost is determined by backtracking search through dynamic programming; S162, Based on the optimal matching path, perform path-guided resampling to generate two aligned time series of the same length; S163, the aligned time series is divided into multiple time periods, and the Pearson correlation coefficient of each time period is calculated. Finally, the time series correlation strength coefficient is obtained by weighted average algorithm.

[0054] In step S161, the optimal matching path with the minimum total cost is determined by backtracking through dynamic programming on the cumulative distance matrix. Specifically, this step aims to trace back to the path with the minimum total cost from the cumulative distance matrix generated in step S15, which records the cumulative costs of all paths. In a preferred embodiment, this process is implemented through dynamic programming backtracking. The algorithm starts from the end point of the cumulative distance matrix (i.e., the upper right element D(M,N)) and searches backward to the starting point (the lower left element D(1,1)). At each step, the algorithm checks the three possible predecessor neighbor points of the current point (i,j)—(i-1,j), (i,j-1), and (i-1,j-1)—and selects the point with the smallest cumulative distance value as the previous point on the path. This process is repeated until the starting point is reached. The coordinate sequence of the points found in this way is the best time alignment path with the lowest mathematical total cost connecting the two time series, i.e., the optimal matching path. In addition, to prevent ill-conditioned matching (i.e., matching too many different time points from one time point), a Sakoe-Chiba Band can be introduced as a constraint, which restricts the entire path search to a band of fixed width (e.g., 10 time points) near the main diagonal of the matrix.

[0055] In step S162, path-guided resampling is performed based on the optimal matching path to generate two aligned time series of equal length. Specifically, this step aims to utilize the optimal matching path found in the previous step to generate two new time series that are fully aligned in time and can be directly compared for correlation. In a preferred embodiment, the process is as follows: First, define a "normalized time axis". The length K of this time axis is the total number of points contained in the optimal matching path found in the previous step. This optimal matching path itself can be represented as a sequence containing K coordinate pairs: Each element , representing the first of the original pressure sequence The point and the original flow sequence The optimal match is then achieved for each point. Path-guided resampling is then performed based on this path W to generate two new aligned sequences, each of length K. Generate a new pressure sequence P': the k-th element of this sequence Its value is taken from the original pressure sequence P, from the path point The specified number The value of each element, i.e. .

[0056] Generate a new flow sequence F': the k-th element of this sequence Its value is taken from the original flow sequence F, from the path point The specified number The value of each element, i.e. .

[0057] By traversing all K points in the optimal matching path W, we finally obtain two aligned time series with exactly the same length and each time point being the best match.

[0058] In step S163, the aligned time series is divided into multiple time periods, and the Pearson correlation coefficient for each time period is calculated. Finally, a weighted average algorithm is used to obtain the time series correlation strength coefficient. Specifically, this step aims to quantify the correlation between the two aligned time series. In a preferred embodiment, to capture local changes in correlation, the standardized sequence pairs of the same length obtained in the previous step are first divided into multiple equal-length, non-overlapping time periods (e.g., 10 time periods). Then, for each time period, the Pearson correlation coefficient between the pressure and flow series within that period is calculated independently. The value range of this coefficient is [-1, 1], which can accurately quantify the degree of linear correlation between the two series within that time period. Finally, to obtain a single indicator that can represent the overall correlation strength, a weighted average algorithm is used to combine the Pearson correlation coefficients of all time periods. The weights can be set according to the data quality or volatility of each time period; for example, more stable data segments are given higher weights (e.g., 1.2), and segments with drastic fluctuations are given lower weights (e.g., 0.8). The final value obtained by weighted averaging (e.g., 0.88) is the time-series correlation strength coefficient that can comprehensively reflect the dynamic correlation between pressure and flow parameters throughout the entire abnormal range.

[0059] In step S17, the failure modes of the ballast water system are obtained based on the respective changing trends of the time-series correlation strength coefficient, pressure parameters, and flow parameters within the characteristic fluctuation range. It should be noted that, in the first embodiment of the present invention, this processing step specifically includes steps S171 to S174: S171, compare the temporal correlation strength coefficient with a preset first threshold and a second threshold, wherein the first threshold is less than the second threshold; S172, when the time series correlation strength coefficient is less than the first threshold and the time series of pressure parameters shows a downward trend, the fault mode is judged as pump body impeller wear fault. S173, when the time sequence correlation strength coefficient is greater than the second threshold and the time series of flow parameters shows periodic pulsation, the fault mode is judged as valve sealing surface damage fault. S174, When the time-series correlation strength coefficient is between the first threshold and the second threshold, calculate the numerical distance between the time-series correlation strength coefficient and the first threshold and the second threshold respectively, and determine the weight for comprehensive evaluation based on the numerical distance to finally determine the failure mode.

[0060] In step S171, the temporal correlation strength coefficient is compared with a preset first threshold and a second threshold, wherein the first threshold is less than the second threshold. Specifically, this step is the initial decision-making stage for fault mode classification. In a preferred embodiment, two key correlation strength thresholds are preset: a first threshold and a second threshold. For example, the first threshold can be set to 0.6 and the second threshold to 0.8. These two thresholds are determined by running steps S11 to S16 of the present invention on a historical dataset containing a large number of known fault types (such as impeller wear, valve damage, etc.), calculating the statistical distribution law of the temporal correlation strength coefficient under different fault modes, and finding the decision boundary that can best distinguish these types of faults. In this step, the temporal correlation strength coefficient value obtained in the previous step is compared with these two thresholds, and they are divided into three intervals: "low correlation" (less than 0.6), "high correlation" (greater than 0.8), and "medium correlation" (between 0.6 and 0.8).

[0061] In step S172, when the time series correlation strength coefficient is less than the first threshold and the time series of pressure parameters shows a downward trend, the fault mode is judged as pump body impeller wear fault.

[0062] Specifically, when the coefficient is less than a first threshold (e.g., 0.6), it indicates a deterioration in the synchronicity of pressure and flow, which aligns with the physical mechanism of pump performance degradation. In this case, this step further analyzes the trend of the pressure time series within the characteristic fluctuation range. In a preferred embodiment, linear regression analysis is used to quantify the trend. This involves performing a least-squares linear fit on the pressure series and calculating the slope of the fitted line. If the slope remains negative and its absolute value exceeds a preset rate-of-decrease threshold (e.g., 0.02 MPa / s), the pressure series is determined to exhibit a significant downward trend. When both the "low correlation" and "pressure drop" conditions are met simultaneously, the failure mode is identified as pump impeller wear failure.

[0063] In step S173, when the time-series correlation strength coefficient is greater than the second threshold and the time series of the flow parameters exhibits periodic pulsation, the fault mode is determined to be valve sealing surface damage fault. Specifically, when the coefficient is greater than the second threshold (e.g., 0.8), it indicates that the pressure and flow are still highly synchronized, but the system has an anomaly, which usually points to local disturbances in the flow path. At this time, this step performs periodic analysis on the flow time series within the characteristic fluctuation range. In a preferred embodiment, a Fast Fourier Transform (FFT) is used to transform it to the frequency domain. By analyzing its spectrum, if there is a significant energy peak at a specific frequency point (e.g., 1.2 Hz), the amplitude of which exceeds a multiple (e.g., 3 times) of the surrounding background noise baseline, the flow series is determined to exhibit periodic pulsation. When both the conditions of "high correlation" and "periodic pulsation of flow" are met simultaneously, the fault mode is determined to be valve sealing surface damage fault.

[0064] In step S174, when the time-series correlation strength coefficient is between a first threshold and a second threshold, the numerical distance between the time-series correlation strength coefficient and the first and second thresholds is calculated respectively, and the weight used for comprehensive evaluation is determined based on the numerical distance to ultimately determine the fault mode. Specifically, when the coefficient is between the first threshold... Second threshold When the value is between 0.72 and 0.82, it indicates that the fault characteristics are unclear, and it may be a compound fault or an early fault. In this case, a weighted decision-making method based on fuzzy logic is used in this step. The detailed process is as follows: First, the membership weight of coefficient C for the two typical failure modes is calculated. This is determined by calculating the normalized distance of C to two thresholds. For the "pump body impeller wear" failure (corresponding to low correlation), its weight is... The calculation method is to measure the high correlation threshold of C. How far? The specific formula is: .

[0065] For the "valve sealing surface damage" fault (corresponding to high correlation), its weight The calculation method is to measure the distance of C from the low correlation threshold. How far? The specific formula is: For example, when C=0.72, Th1=0.6, and Th2=0.8, the calculation yields: The sum of these two weights is 1, clearly quantifying that the current state is more inclined towards "valve damage".

[0066] Then, this step performs feature analyses for two types of faults in parallel: analyzing whether the pressure sequence has a downward trend as in step S172, and analyzing whether the flow sequence has periodic fluctuations as in step S173. Each of these analyses will output a Boolean result (yes / no, i.e., 1 or 0).

[0067] Finally, a weighted composite evaluation is performed to determine the final failure mode. This step calculates the final scores for both types of failures: Ultimately, the fault mode with the higher score will be identified as the primary fault, while the fault mode with the lower score (if the score is not zero) will be identified as the secondary fault. For example, if a flow pulsation is detected at C=0.72 (presence = 1), but no significant pressure drop is detected (absence = 0), then... Therefore, the final diagnosis was: the main fault was "damage to the valve sealing surface." This method achieved a well-reasoned, layered diagnosis even in ambiguous situations.

[0068] Reference Figure 2 The second embodiment of the present invention provides a remote fault diagnosis system for a ballast water system, comprising: The parameter acquisition module is used to acquire multi-dimensional operating parameters of the ship's ballast water system, including time series of pressure parameters, flow parameters, and temperature parameters. The standardization module is used to perform zero-mean-unit-variance standardization on the time series of the multidimensional operating parameters to obtain a standardized multidimensional time series. The feature reduction module is used to perform principal component analysis and dimensionality reduction on the standardized multidimensional time series to obtain a low-dimensional feature vector sequence. The fluctuation detection module is used to perform sliding window analysis on the low-dimensional feature vector sequence to determine the feature fluctuation range in which the amplitude fluctuation exceeds a preset fluctuation threshold. The cost matrix construction module is used to extract the time series of pressure parameters and flow parameters within the characteristic fluctuation range, and construct the cumulative distance matrix between the two based on the time series. The correlation strength quantization module is used to search for the optimal matching path on the cumulative distance matrix, and to align and quantify the correlation of the time series of the pressure parameter and flow parameter based on the optimal matching path, and to calculate the time series correlation strength coefficient. The fault diagnosis module is used to obtain the fault mode of the ballast water system based on the respective changing trends of the time-series correlation strength coefficient, the pressure parameter, and the flow rate parameter within the characteristic fluctuation range.

[0069] It should be noted that the remote fault diagnosis system for ballast water system provided in this embodiment of the invention is used to execute all the process steps of the remote fault diagnosis method for ballast water system in the above embodiment. The working principle and beneficial effects of the two are one-to-one, so they will not be described again.

[0070] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a feature reduction program. When the processor executes the computer program, it implements the steps in the various embodiments of the remote fault diagnosis method for ballast water systems described above, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, such as the fluctuation detection module.

[0071] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0072] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0073] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0074] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0075] Wherein, if the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0076] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and 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 modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0077] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A remote fault diagnosis method for a ballast water system, characterized in that, include: Obtain multi-dimensional operating parameters of the ship's ballast water system, including time series of pressure, flow, and temperature parameters. The time series of the multidimensional operating parameters are subjected to zero-mean-unit-variance standardization to obtain a standardized multidimensional time series. Principal component analysis and dimensionality reduction are performed on the standardized multidimensional time series to obtain a low-dimensional feature vector sequence. A sliding window analysis is performed on the low-dimensional feature vector sequence to determine the feature fluctuation range where the amplitude fluctuation exceeds a preset fluctuation threshold; Extract the time series of pressure and flow parameters within the characteristic fluctuation range, and construct the cumulative distance matrix between them based on the time series; The optimal matching path is searched on the cumulative distance matrix, and the time series of the pressure parameter and flow parameter are aligned and correlated based on the optimal matching path to calculate the time series correlation strength coefficient. Based on the respective trends of the time-series correlation strength coefficient, the pressure parameter, and the flow rate parameter within the characteristic fluctuation range, the failure modes of the ballast water system are obtained, including: The temporal correlation strength coefficient is compared with a preset first threshold and a second threshold, wherein the first threshold is less than the second threshold; When the time-series correlation strength coefficient is less than the first threshold and the time series of the pressure parameter shows a downward trend, the fault mode is judged as pump impeller wear fault. When the time-series correlation strength coefficient is greater than the second threshold and the time series of the flow parameters exhibits periodic pulsation, the fault mode is judged as valve sealing surface damage fault. When the time-series correlation strength coefficient is between the first threshold Second threshold When the time-series correlation strength coefficient C is between the first and second thresholds, the normalized distances are calculated and used as the membership weights for the two typical failure modes; for pump impeller wear failure, the weights are... The calculation method is as follows: For valve sealing surface damage faults, its weight The calculation method is as follows: Then, feature analyses for the two types of faults are performed in parallel: analyzing whether the pressure sequence shows a downward trend; analyzing whether the flow sequence shows periodic fluctuations; each analysis outputs a Boolean value result, 1 if a Boolean value exists, and 0 if no Boolean value exists; finally, a weighted comprehensive evaluation is performed to determine the final fault mode, and the final scores for the two faults are calculated: ； ； Ultimately, the fault mode with the higher final score between the two fault types will be identified as the primary fault; while the one with the lower score but not zero will be identified as the secondary fault mode.

2. The remote fault diagnosis method for a ballast water system as described in claim 1, characterized in that, The process of performing zero-mean-unit-variance standardization on the time series of the multidimensional operating parameters to obtain a standardized multidimensional time series includes: For the time series of pressure, flow rate, and temperature in the multidimensional operating parameters, calculate their respective mean and standard deviation; The calculated mean and standard deviation are used to perform numerical normalization on the time series, mapping the original values ​​to a standard interval, thereby obtaining a standardized multidimensional time series.

3. The remote fault diagnosis method for a ballast water system as described in claim 1, characterized in that, The step of performing principal component analysis and dimensionality reduction on the standardized multidimensional time series to obtain a low-dimensional feature vector sequence includes: Based on the pressure, flow rate, and temperature time series in the standardized multidimensional time series, the covariance matrix among these three parameters is calculated. Perform eigenvalue decomposition on the covariance matrix to obtain a set of eigenvalues ​​and a corresponding set of eigenvectors; The cumulative contribution rate of each principal component is calculated based on the eigenvalues, and based on the comparison result of the cumulative contribution rate and the preset threshold, k eigenvectors corresponding to the first k principal components to be retained are selected from the eigenvector set. A dimensionality reduction transformation matrix is ​​constructed using the eigenvectors, and the dimensionality reduction transformation matrix is ​​applied to project the standardized multidimensional time series to obtain a low-dimensional eigenvector sequence.

4. The remote fault diagnosis method for a ballast water system as described in claim 1, characterized in that, The step of performing sliding window analysis on the low-dimensional feature vector sequence to determine the feature fluctuation range where the amplitude fluctuation exceeds a preset fluctuation threshold includes: Based on the low-dimensional feature vector sequence, the variance explained ratio at each time point is calculated to obtain the variance explained ratio time series. A fixed-length sliding window is used to scan the variance explanation ratio time series, and the difference between the maximum and minimum variance explanation ratios within the sliding window is calculated as the amplitude fluctuation. When the amplitude fluctuation exceeds a preset fluctuation threshold, the time range covered by the current sliding window is determined as the characteristic fluctuation interval.

5. The remote fault diagnosis method for a ballast water system as described in claim 1, characterized in that, The step of extracting the time series of pressure and flow parameters within the characteristic fluctuation range, and constructing a cumulative distance matrix between them based on the time series, includes: The time series of pressure and flow parameters within the characteristic fluctuation range are length aligned and numerically normalized to obtain standardized pressure and flow feature vectors. Calculate the Euclidean distance between the standardized pressure feature vector and the flow feature vector at each time point, and construct the cumulative distance matrix between them.

6. The remote fault diagnosis method for a ballast water system as described in claim 1, characterized in that, The process of searching for the optimal matching path on the cumulative distance matrix, aligning and quantifying the correlation of the time series of the pressure and flow parameters based on the optimal matching path, and calculating the time series correlation strength coefficient includes: In the cumulative distance matrix, the optimal matching path with the minimum total cost is determined by backtracking through dynamic programming. Based on the optimal matching path, path-guided resampling is performed to obtain two aligned time series of the same length. The aligned time series is divided into multiple time periods, and the Pearson correlation coefficient for each time period is calculated. Finally, the time series correlation strength coefficient is obtained by weighted average algorithm.

7. A remote fault diagnosis system for a ballast water system, characterized in that, A method for implementing remote fault diagnosis of a ballast water system as described in any one of claims 1 to 6, comprising: The parameter acquisition module is used to acquire multi-dimensional operating parameters of the ship's ballast water system, including time series of pressure parameters, flow parameters, and temperature parameters. The standardization module is used to perform zero-mean-unit-variance standardization on the time series of the multidimensional operating parameters to obtain a standardized multidimensional time series. The feature reduction module is used to perform principal component analysis and dimensionality reduction on the standardized multidimensional time series to obtain a low-dimensional feature vector sequence. The fluctuation detection module is used to perform sliding window analysis on the low-dimensional feature vector sequence to determine the feature fluctuation range in which the amplitude fluctuation exceeds a preset fluctuation threshold. The cost matrix construction module is used to extract the time series of pressure parameters and flow parameters within the characteristic fluctuation range, and construct the cumulative distance matrix between the two based on the time series. The correlation strength quantization module is used to search for the optimal matching path on the cumulative distance matrix, and to align and quantify the correlation of the time series of the pressure parameter and flow parameter based on the optimal matching path, and to calculate the time series correlation strength coefficient. The fault diagnosis module is used to obtain the fault mode of the ballast water system based on the respective changing trends of the time-series correlation strength coefficient, the pressure parameter, and the flow rate parameter within the characteristic fluctuation range.

8. An electronic device, characterized in that, The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the remote fault diagnosis method for a ballast water system as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the remote fault diagnosis method for a ballast water system as described in any one of claims 1 to 6.

Images