Ship course angle and yaw attitude separation method, device and system

Abstract

The application discloses a ship heading angle and yaw attitude separation method, device and system, and belongs to the technical field of ship attitude measurement and signal processing. The application is aimed at the yaw angle data collected by a combined inertial navigation system and other sensors, and the original signal is adaptively decomposed into a plurality of intrinsic mode function components and a residual component by performing empirical mode decomposition on the yaw angle signal. On the basis, the IMF components are combined, iterated and statistically evaluated from three dimensions of normal distribution compliance, probability density symmetry and frequency distribution rationality, the optimal mode combination scheme of the heading angle and the yaw attitude is automatically determined, and the accurate separation of the ship heading angle and the yaw attitude is realized. The application can accurately extract a smooth heading angle sequence and a yaw attitude sequence meeting zero mean and stationarity characteristics, and can be widely applied to the fields of ship motion control, sea wave parameter inversion, ship automatic driving auxiliary decision and the like.

Description

Technical Field

[0001] This invention relates to the field of ship attitude measurement and signal processing technology, specifically to a method, apparatus, and system for separating ship heading angle and bow attitude based on empirical mode decomposition. This invention is particularly applicable to ship yaw angle data acquired by attitude sensors such as integrated inertial navigation systems, and can separate heading changes caused by human control or navigation plans from the ship's bow motion attitude caused by sea surface winds and waves. It can be widely applied in ship motion control, marine element inversion in the ship's location, and other fields. Background Technology

[0002] In applications such as ship navigation, motion control, and marine environmental monitoring, accurately acquiring a ship's heading angle and yaw posture has significant engineering value. Shipborne sensors, such as integrated inertial navigation systems (INS), typically output a composite yaw angle signal that includes changes in heading angle and yaw motion. The heading angle reflects the large-scale directional change of the ship determined by human control or navigation plan, and its value range is typically [insert range here]. The yaw rate is characterized by gradual change and trend; the bow roll attitude is mainly caused by the interaction between random waves and the hull, exhibiting the statistical characteristics of a stationary random process, and approximately following a normal distribution with zero mean.

[0003] Traditional methods for distinguishing between heading angle and yaw attitude mainly rely on mechanical filtering, low-pass / high-pass digital filtering, or frequency domain filtering methods based on Fourier transform. However, these methods have the following shortcomings in practical applications: (1) The cutoff frequency is difficult to determine: The low-pass / high-pass filtering method relies on a preset cutoff frequency to separate the heading angle (low frequency) and the yaw attitude (high frequency). However, due to different sea states, different ship types and different navigation conditions, the frequency boundary between the heading angle and the yaw attitude is dynamic. A fixed cutoff frequency cannot be universally applied to all situations, resulting in an unsatisfactory separation effect.

[0004] (2) Non-stationarity of sensor output data: Due to human manipulation and navigation plans, the yaw angle data of ships is inherently non-stationary and nonlinear. Traditional frequency domain methods such as Fourier transform are based on the assumption of signal stationarity. Therefore, when processing ship yaw angle data, spurious frequency shifts are inevitable, and it is difficult to effectively separate the ship's steering operation and yaw motion components with clear physical meaning.

[0005] (3) Lack of adaptive mechanism: Existing methods usually require manual setting of filtering parameters based on experience, which lacks the ability to adapt to the characteristics of the signal itself. In complex systems with human-ship-environment coupling, the limitations of manually setting parameters are very prominent.

[0006] Mode decomposition of a signal is a method of decomposing signal components. Its core idea is to decompose a complex, non-stationary signal into a set of essential, simple, and physically meaningful sub-signals (called "eigenmode functions" or "modes"). Therefore, mode decomposition of signals has been widely used in fields such as mechanical fault diagnosis, seismic wave processing, and meteorological data processing.

[0007] Empirical Mode Decomposition (EMD) is an analytical method for nonstationary and nonlinear signals. It departs from the traditional approach of using harmonic basis functions based on Fourier decomposition and wavelet basis functions based on wavelet decomposition. Instead, it adaptively decomposes the signal based entirely on its waveform characteristics, without requiring pre-defined decomposition basis functions. After EMD decomposition, the signal yields several Intrinsic Mode Functions (IMFs) and a Residual Term (Res). Each IMF component exhibits single-component characteristics and is obtained in descending frequency order.

[0008] Since its introduction by Huang et al. in 1998, the EMD method has seen initial applications in marine engineering, such as wave spectrum analysis and vibration signal processing of marine structures. However, existing research has not yet explored the application of the EMD method to separate the heading angle and bow attitude from ship yaw angle signals. In particular, how to automatically and accurately identify which modal components belong to the heading angle and which belong to the bow attitude from the multiple IMF components and residual components generated by EMD decomposition, based on the physical characteristics of ship motion, remains a valuable technical problem. Summary of the Invention

[0009] This invention provides a method, apparatus, and system for separating ship heading angle and yaw attitude based on empirical mode decomposition (EMD). It addresses the problem in existing filtering methods that rely on preset cutoff frequencies or fixed basis functions, which struggle to accurately and universally separate physically meaningful heading angle and yaw attitude when processing non-stationary and nonlinear ship yaw angle data due to a lack of adaptive mechanisms. This method eliminates the need for a preset filter cutoff frequency, adaptively decomposes the yaw angle data based on its waveform characteristics, overcomes nonlinear and non-stationary factors introduced by human manipulation during navigation through short-time segmented processing, and automatically determines the optimal mode combination scheme by comprehensively evaluating multi-dimensional statistical features. This allows for accurate and robust separation of ship heading angle and yaw attitude from the composite yaw angle signal acquired by sensors.

[0010] In a first aspect, this invention proposes a method for separating a ship's heading angle and yaw attitude, comprising: performing short-time segmentation processing on ship yaw angle data collected by sensors to obtain multiple short-time data segments; performing empirical mode decomposition on each short-time data segment to obtain several intrinsic mode function components and a residual component; traversing all possible mode matching combinations, and matching the residual component with the lowest frequency... m The intrinsic mode function components are combined to form candidate heading angles, and the remaining high-frequency intrinsic mode function components are combined to form candidate yaw attitudes, where m is a positive integer; multi-dimensional statistical feature evaluation is performed on the candidate yaw attitudes under each mode matching combination; the optimal matching combination is determined based on the multi-dimensional statistical feature evaluation results; the final heading angle and yaw attitude are calculated based on the optimal matching combination; the extraction results of all short-time data segments are spliced ​​together in chronological order to output a complete heading angle sequence and yaw attitude sequence.

[0011] In some examples, in the short-time segmentation process, the length of the preset time window is greater than or equal to the wave spectrum accumulation time in the sea area.

[0012] In some examples, the actual length of each short data segment is 2 to 3 times the preset time window, and there is a time overlap area between adjacent short data segments.

[0013] In some examples, the multidimensional statistical feature assessment includes tests of normality, probability density distribution symmetry, and frequency distribution characteristics.

[0014] In some examples, the normality conformity test includes: plotting the normal quantile-quantile plot of the candidate yaw attitude to obtain a scatter distribution of the quantile points of the candidate yaw attitude relative to the theoretical quantiles of the normal distribution; performing linear regression fitting on the scatter distribution to obtain a regression line; calculating the correlation coefficient and determination coefficient between each data point in the scatter distribution and the regression line, as well as the Lilliefors test of the candidate yaw attitude. p Value; based on the correlation coefficient, the determination coefficient, and the p The value is used to determine whether the candidate bow attitude follows a normal distribution.

[0015] In some examples, the probability density distribution symmetry test includes: performing kernel density estimation on the candidate yaw attitude to obtain the probability density function of the candidate yaw attitude; calculating the integral ratio of the probability density function on the positive half-axis and the negative half-axis; and examining whether the median and quartile distributions of the candidate yaw attitude are symmetric about the zero point using a box plot.

[0016] In some examples, the frequency distribution characteristic test includes: performing a Fourier transform on the candidate yaw attitude sequence to obtain the amplitude spectrum of the candidate yaw attitude; setting an extremely low frequency threshold, and checking whether there are peak values ​​exceeding the preset amplitude threshold in the frequency range of the amplitude spectrum below the extremely low frequency threshold.

[0017] Secondly, this invention proposes a device for separating a ship's heading angle and yaw attitude, comprising: a data segmentation module configured to perform short-time segmentation processing on ship yaw angle data collected by sensors to obtain multiple short-time data segments; an empirical mode decomposition module configured to perform empirical mode decomposition on each short-time data segment to obtain several intrinsic mode function components and a residual component; and a mode matching module configured to traverse all possible mode matching combinations and match the residual component with the lowest frequency... m The intrinsic mode function components are combined to form a candidate heading angle, and the remaining high-frequency intrinsic mode function components are combined to form a candidate yaw attitude, wherein... m The input is a positive integer; the statistical evaluation module is configured to perform multi-dimensional statistical feature evaluation on the candidate yaw attitudes under each matching combination; the optimal matching determination module is configured to determine the optimal matching combination based on the multi-dimensional statistical feature evaluation results; the attitude calculation module is configured to calculate the final heading angle and yaw attitude based on the optimal matching combination; the result stitching module is configured to stitch the extracted results of all short-time data segments in chronological order and output a complete heading angle sequence and yaw attitude sequence.

[0018] Thirdly, the present invention proposes a ship heading angle and bow attitude separation system, including a processor and a memory, wherein the memory stores program instructions, and the processor executes the program instructions to realize the ship heading angle and bow attitude separation method.

[0019] Fourthly, the present invention proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method for separating the ship's heading angle and yaw attitude.

[0020] The present invention has the following beneficial effects: (1) The EMD method is based entirely on the waveform characteristics of the signal itself for adaptive decomposition. There is no need to preset the basis function or the filter cutoff frequency. It has good adaptability to different sea states, different ship types and different navigation conditions.

[0021] (2) The three dimensions of normality, probability density symmetry and frequency distribution rationality are comprehensively evaluated to ensure that the extracted bow attitude meets the physical and statistical laws of ship motion under random waves; at the same time, the large-scale and gradually changing trend of the heading angle is guaranteed.

[0022] (3) The introduction of short-time segmentation processing and overlapping area continuity verification mechanism effectively solves the problem of the influence of long-term trend changes in heading angle or the relatively drastic changes in heading caused by human manipulation on EMD decomposition, ensuring the continuity and smoothness of the extraction results on the time axis, and is suitable for long-term continuous observation scenarios in actual engineering.

[0023] (4) By traversing the modal combinations and automatically performing multi-dimensional statistical evaluation, the optimal separation scheme can be determined without manual intervention, making it suitable for embedding into shipborne or shore-based automated data processing processes.

[0024] (5) The extracted bow attitude sequence can be used not only for ship motion state monitoring and control, but also as input data for wave parameter inversion (such as significant wave height, wave period, wave direction spectrum, etc.), which has important secondary application value. Attached Figure Description

[0025] Figure 1 This is a flowchart of a method for separating ship heading angle and bow attitude based on empirical mode decomposition according to an embodiment of the present invention.

[0026] Figure 2 It refers to the ship yaw data collected by actual measurement according to an embodiment of the present invention, and the residual components and intrinsic mode components after EMD decomposition.

[0027] Figure 3 This refers to the ship's heading, bow attitude, and bow frequency components under the optimal matching combination according to an embodiment of the present invention.

[0028] Figure 4 This is the statistical characteristic distribution of bow roll according to an embodiment of the present invention.

[0029] Figure 5 This is the extraction of long-term heading angle data according to an embodiment of the present invention (the red and blue curves in the figure appear alternately, representing adjacent short-time segment intervals).

[0030] Figure 6 This is the extraction of long-term yaw attitude data according to an embodiment of the present invention (the red and blue curves in the figure appear alternately, representing adjacent short-time intervals).

[0031] Figure 7 This refers to the time period during which the heading angle increases and decreases dramatically, as described in an embodiment of the present invention (the black curve in the figure represents the yaw angle data collected by the sensor, while the red and blue curves represent the heading angle data extracted by the algorithm; the red and blue curves appear alternately, indicating adjacent short-segment time intervals).

[0032] Figure 8This is a region of significant amplitude variation in the locally magnified bow roll angle according to an embodiment of the present invention (the red and blue curves in the figure appear alternately, representing adjacent short-time intervals). Detailed Implementation

[0033] Example 1: A method for separating ship heading angle and yaw attitude based on empirical mode decomposition. For example... Figure 1 As shown, the method includes: performing short-time segmentation processing on the ship yaw angle data collected by the sensor to obtain multiple short-time data segments; performing empirical mode decomposition on each short-time data segment to obtain several intrinsic mode function components and a residual component; traversing all possible mode matching combinations, and matching the residual component with the lowest frequency... m The intrinsic mode function components are combined to form candidate heading angles, and the remaining high-frequency intrinsic mode function components are combined to form candidate yaw attitudes, where m is a positive integer. Multi-dimensional statistical feature evaluation is performed on the candidate yaw attitudes under each matching combination. Based on the multi-dimensional statistical feature evaluation results, the optimal matching combination is determined. The final heading angle and yaw attitude are calculated based on the optimal matching combination. The extraction results of all short-time data segments are concatenated in chronological order to output a complete heading angle sequence and yaw attitude sequence. The method is described in detail below.

[0034] Step 1: Short-time segmented processing of yaw angle data.

[0035] The long-term yaw angle sequence acquired by the sensor is divided into several short data segments according to a preset time window. The length of the preset time window can be set according to actual application requirements, preferably greater than or equal to the wave spectrum accumulation time in the sea area (e.g., 120 seconds). To ensure the continuity and correlation of the data after segmentation processing, and to fully evaluate the statistical characteristics of the yaw attitude, the actual length of each short data segment can be appropriately extended, for example, extended to 2 to 3 times the preset time window (e.g., 360 seconds) as a processing unit. There are time overlap areas between adjacent short data segments to ensure the continuity of the extraction results on the time axis.

[0036] Step 2: Perform EMD decomposition on each short-time data segment.

[0037] For the yaw angle signal in the current short-time data segment Perform empirical mode decomposition to obtain K Each intrinsic mode function component (IMF component) and a residual component ,satisfy: (1.1) Among them, each IMF component Arranged in descending order of frequency, i.e. Highest frequency Lowest frequency; residual component This represents a slowly changing trend term or DC component in the signal.

[0038] Step 3: Traverse the mode combinations and construct candidate heading angles and yaw attitudes.

[0039] For EMD decomposition results K The number of IMFs corresponding to a heading match combination (also known as a modal match combination) is defined as follows: (1 IMF component and 1 residual component). m ( ), indicating that the residual components With the lowest frequency m The combination of IMF components forms the heading angle. ,the remaining The combination of high-frequency IMF components represents the heading attitude. .Right now: (1.2) from arrive Iterate through all possible matching combinations to obtain The candidate heading angle sequence and yaw attitude sequence.

[0040] The above modal combination strategy is based on the following physical understanding: residual components This represents the slowly changing trend term remaining after all oscillation modes in the original signal have been removed. It exhibits the same slow-changing characteristics as the heading angle, which is influenced by human manipulation; therefore, the residual component must be the dominant component of the heading angle. Using the residual component as the dominant component, IMF components with frequencies increasing from low to high are successively incorporated into the heading angle, and the remaining high-frequency components are used to form the yaw attitude. This combination strategy ensures that candidate solutions cover the complete solution space from "lowest frequency heading angle (residual component only)" to "highest frequency heading angle (residual component + all low-frequency IMFs)".

[0041] Step 4: Evaluation of multi-dimensional statistical features and determination of the optimal matching combination.

[0042] Based on the statistical characteristics of the ship's bow attitude, Candidate samples of bow attitude obtained under each matching combination The number of heading IMFs corresponding to the optimal matching combination will be determined by statistical evaluation from the following three dimensions. .

[0043] Dimension 1: Normality test.

[0044] To examine the degree to which the sample data follows a normal distribution under the current combination, a normal quantile-quantile plot (normal QQ plot) is drawn for the candidate yaw attitudes, obtaining the scatter distribution of the quantile points of the candidate yaw attitudes relative to the theoretical quantiles of the normal distribution; a linear regression is performed on the scatter distribution to obtain the regression line; and the correlation coefficient between each data point in the scatter distribution and the regression line is calculated. c and coefficient of determination And the significance level of the Lilliefors test for candidate bow roll attitude sequences. p value.

[0045] Since the normal distribution is the most crucial statistical characteristic of bow attitude, and under the influence of random waves, the bow angle of a ship approximately follows a normal distribution with zero mean. Therefore, the closer the correlation coefficient of the normal QQ plot is to 1, the higher the coefficient of determination. The closer to 1, p The larger the value (greater than the significance level, which can generally be set to 0.05), the more the extracted bow attitude under the current matching combination conforms to the real physical laws.

[0046] By calculating the correlation coefficient of the regression line under all matching combinations c Coefficient of determination and significance level p The value determines the number of heading IMFs corresponding to the matching combination that best follows a normal distribution.

[0047] Dimension Two: Symmetry test of mathematical expectation and probability density distribution.

[0048] By calculating the probability density function (PDF) of the current combination of bow and sway samples, kernel density estimation (KDE) is performed on the samples; the mathematical expectation of the samples is then used. Assuming the expected value, calculate the integral ratio of the probability density of the positive and negative halves of the sample: (1.3) like The closer it is to the ideal value of 1, the more symmetrical the probability density distribution of the yaw attitude under the current matching combination is on both sides of the zero value, and the more it conforms to the physical characteristics of zero mean yaw motion.

[0049] In addition, box plots can be used to visually examine whether the median and quartile distributions of the sample data are symmetrical about the zero point, and whether the sample mean is close to zero, as an auxiliary basis for judgment.

[0050] Dimension 3: Frequency distribution characteristics test.

[0051] Perform a Fourier Transform (FT) on the bow attitude sequence under the current matching combination to examine whether its frequency component distribution is reasonable.

[0052] Under the influence of random ocean waves, the frequency components of a ship's bow roll motion are predominantly low-frequency, but should not include extremely low-frequency amplitude components (these extremely low-frequency components usually belong to heading changes). By examining the frequency-amplitude spectrum (amplitude spectrum), it is determined whether there are abnormal and amplitude low-frequency peaks. This involves setting an extremely low-frequency threshold and checking whether there are peaks exceeding the preset amplitude threshold within the frequency range below the extremely low-frequency threshold in the amplitude spectrum. For example, if a significant high-amplitude component appears in the extremely low-frequency band (e.g., below 0.01Hz), it indicates that the heading angle component may still not be completely separated in the current matching combination, and more low-frequency IMFs need to be included in the heading angle (i.e., increasing the heading angle). m value).

[0053] For candidate yaw attitudes, the evaluation results of the three dimensions mentioned above are considered together. Since the normality and zero-mean property of yaw attitude are its most important statistical characteristics, the evaluation results of dimensions one and two are given greater weight, while dimension three, as an auxiliary evaluation, is given less weight. Finally, the one with the best normality (correlation coefficient of the normal distribution regression line) is selected. c and coefficient of determination Closest to 1, significance level p The value is the largest, and the zero-mean property is the best. The closest matching combination to 1) is the number of IMFs in the corresponding heading.

[0054] Step 5: Final extraction of heading angle and yaw attitude.

[0055] Based on the number of heading IMFs corresponding to the optimal matching combination The final heading angle is calculated. and bow roll attitude : (1.4) Step Six: Segmented Results Assembly and Continuity Verification.

[0056] Steps two through five are performed on each short-time data segment to obtain the heading angle and yaw attitude sequences for each time period. For the extraction results of the time overlap region of adjacent data segments, the correlation coefficient and root mean square error (RMSE) are calculated to verify the continuity and reliability of the segmented processing. At the data junction (i.e., the boundary between the end of the previous segment and the beginning of the next segment), the mean and standard deviation of the junction error are calculated to evaluate the smoothness of the overall extraction results.

[0057] If the continuity of the overlapping area is poor (correlation coefficient less than or RMSE greater than the set percentage threshold), or the connection error is too large, consider sliding the preset time window, or appropriately increasing or decreasing the preset duration, and repeat steps two to five.

[0058] Finally, the extracted results of all segments are spliced ​​together in chronological order to obtain a complete long-term heading angle and yaw attitude sequence.

[0059] Example 2: A device for separating ship heading angle and bow attitude based on empirical mode decomposition.

[0060] This embodiment provides a device for separating a ship's heading angle and yaw attitude based on empirical mode decomposition (EMD). This device is used to implement the method described in Embodiment 1. The device includes a data segmentation module, an EMD module, a mode matching module, a statistical evaluation module, an optimal matching determination module, an attitude calculation module, and a result stitching module.

[0061] The data segmentation module is configured to perform short-time segmentation processing on the ship yaw angle data collected by the sensor, resulting in multiple short-time data segments. The function of this module corresponds to step one of Embodiment 1, and its segmentation strategy includes setting a preset time window, determining the data segment length, and the time overlap area between adjacent segments.

[0062] The Empirical Mode Decomposition (EMD) module is configured to perform EMD on each short-time data segment, obtaining several intrinsic mode function (IMF) components and one residual component. The functionality of this module corresponds to step two of Example 1.

[0063] The modal matching module is configured to iterate through all possible modal matching combinations and match the residual component with the lowest frequency. m The intrinsic mode function components are combined to form a candidate heading angle, and the remaining high-frequency intrinsic mode function components are combined to form a candidate yaw attitude, wherein... m It is a positive integer. The functionality of this module corresponds to the modal combination strategy described in formula (1.2) in step three of embodiment 1.

[0064] The statistical evaluation module is configured to perform multi-dimensional statistical feature evaluation on the candidate bow attitudes for each matching combination. The functionality of this module corresponds to step four of Embodiment 1. The multi-dimensional statistical feature evaluation includes three dimensions: normality distribution compliance test, probability density distribution symmetry test, and frequency distribution characteristic test.

[0065] The optimal matching determination module is configured to determine the optimal matching combination based on the evaluation results of the multi-dimensional statistical features and the performance of the three dimensions. The function of this module corresponds to the final stage of comprehensive evaluation and selection in step four of embodiment 1.

[0066] The attitude calculation module is configured to calculate the final heading angle and yaw attitude based on the optimal matching combination. The function of this module corresponds to step five and formula (1.4) in Embodiment 1.

[0067] The result stitching module is configured to stitch together the extracted results of all short-time data segments in chronological order, and to perform continuity checks and smoothness assessments on overlapping areas of adjacent segments, outputting a complete heading angle sequence and yaw attitude sequence. The functionality of this module corresponds to step six of Example 1.

[0068] The above modules can be stored in memory and executed by the processor to implement the method flow as described in Example 1.

[0069] Example 3: A system for separating a ship's heading angle from its bow roll attitude.

[0070] This embodiment provides a system for separating a ship's heading angle and yaw attitude. The system includes a processor and a memory. The memory stores program instructions. When the processor executes the program instructions, it can implement all or part of the steps of the ship's heading angle and yaw attitude separation method based on empirical mode decomposition as described in Embodiment 1.

[0071] The system also includes sensors connected to the processor for real-time acquisition of the vessel's yaw angle data. The sensors include at least one of an inertial measurement unit (IMU), a gyrocompass, and a global positioning system (GPS). The processor includes, but is not limited to, at least one of a central processing unit (CPU), a graphics processing unit (GPU), or a digital signal processor (DSP).

[0072] The system also includes a display connected to the processor for visually displaying the separated heading angle sequence and bow attitude sequence for real-time monitoring by the crew or automated systems.

[0073] The system can be deployed independently on the ship's bridge or embedded in an existing integrated bridge system or ship motion monitoring system to complete the entire processing chain from input of raw yaw angle data to output of heading angle and yaw attitude results.

[0074] Example 4: A computer-readable storage medium. For example, the computer-readable storage medium may be a read-only memory (ROM). Read-only memory (ROM), random access memory (RAM), and compact disc read-only memory (CD-ROM). Only memory, CD ROM, magnetic tape, floppy disk, and optical data storage devices, etc. This computer-readable storage medium is used to store computer-readable instructions, which, when executed by a computer, can realize the ship heading angle and bow attitude separation method based on empirical mode decomposition as described in Example 1.

[0075] The verification data comes from a sea trial conducted in the East China Sea in May 2025. The test vessel was equipped with a combined inertial navigation system to collect real-time yaw angle data. The unit of yaw angle is defined as degrees (°). ), with due north as The angle value increases in a clockwise direction (i.e., due north). due east due south due west The data sampling rate meets the requirements for subsequent EMD decomposition and spectral analysis.

[0076] In this embodiment, the specific parameters for short-time segmentation are set as follows: the preset time window is set to 120s; the actual length of the short-time data segment used is 3 times the preset time window to ensure the relevance and continuity of data processing, that is, the total length of each short-time data segment is 360s; the time overlap rate between adjacent data segments is approximately 66.7% (that is, the center time interval between adjacent segments is 120s, and the length of each segment is 360s, with an overlap of 240s).

[0077] The considerations for adopting the above segmentation strategy are as follows: On the one hand, the 120s window length is relatively short, and the main trend of the heading angle can be considered relatively stable within this period, which is conducive to EMD decomposition accurately capturing the gradual change trend of the heading; on the other hand, the actual segment length of 360s extends the data by 180s before and after the central window, providing sufficient contextual information and avoiding the impact of boundary effects on the processing results of the core 120s data segment. The large overlap area (240s) between adjacent segments ensures the continuity and smoothness of the spliced ​​data.

[0078] Taking the yaw angle data segment from 06:34:00 to 06:39:59 (UTC+8) on May 21, 2025 as an example, the yaw angle data within this short-term preset time window... Perform EMD decomposition. Raw yaw angle data. Residual components after EMD decomposition and intrinsic modal components like Figure 2 As shown.

[0079] The raw data shows that the ship's yaw angle is affected by changes in heading on a large scale, generally trending due east (approximately). The deflection trend shows a gradual change in the heading angle; however, on a small scale, it is affected by the bow roll motion caused by the interaction between random waves and the hull, showing a small-amplitude high-frequency jitter, reflecting the fluctuating characteristics of the bow roll attitude.

[0080] After EMD decomposition, the residual components It exhibits a monotonically increasing trend, reflecting the overall trend of course changes; five eigenmodes are obtained through decomposition. ~ It still contains slowly varying, low-frequency components that may reflect changes in heading. Therefore, not all intrinsic modes can be simply regarded as yaw attitude. Statistical evaluation is needed to determine the optimal combination.

[0081] Similarly, using the yaw angle data segment from 06:34:00 to 06:39:59 (UTC+8) on May 21, 2025, EMD decomposition of the yaw angle in this short-time data segment yielded 5 IMF components. ~ and 1 residual component traversal The main statistical characteristics of the bow attitude under each matching combination are shown in Table 1: Table 1. Main statistical characteristics of bow attitude under various matching combinations

[0082] As can be seen from Table 1: when When the expected value deviates significantly from 0 ( ), left and right integral ratio Significant deviation from 1, p The value is only 0.0010, therefore the normal distribution hypothesis is rejected. With a value of only 0.90640, all indicators show that the bow roll attitude extracted under this combination does not conform to the laws of physics.

[0083] when At that time, expectations Close to zero, median Approaching zero, the ratio of left and right integrals Closest to the ideal value of 1 The significance level is much higher than the commonly used significance level of 0.05, making it impossible to reject the normal distribution hypothesis. The correlation coefficient of the QQ plot is 0.99946, and the coefficient of determination is... It is the best among all combinations.

[0084] when At that time, although the expected value and the ratio of left to right integrals also performed well, This indicates that the data still significantly deviates from a normal distribution statistically.

[0085] Based on comprehensive comparison, the optimal matching combination for the current short-term data segment is determined to be: That is, the heading angle is determined by the residual component. and the three IMF components with the lowest frequency ( ~ The bow attitude is composed of the two highest frequency IMF components ( ) and Combination of elements.

[0086] Use the optimal matching combination The extracted heading angle and bow attitude are as follows: Figure 3 As shown, the statistical characteristic distribution of the bow roll attitude is as follows: Figure 4 As shown.

[0087] Depend on Figure 3 It can be seen that the extracted heading angle sequence closely matches the large-scale trend changes of the original yaw angle data, and the heading angle curve is smooth and reasonable; the extracted yaw attitude sequence is stable, with fluctuation amplitude within... Within the range, it conforms to the general magnitude characteristics of bow roll motion; the frequency distribution of the bow roll attitude is reasonable, with no abnormal low-frequency peaks.

[0088] Depend on Figure 4 It can be seen that the probability density distribution of the bow attitude exhibits a typical bell-shaped curve, and the kernel density estimation results are as follows: The density integral ratio is symmetrical about the center and the density integral ratio is symmetrical about the center. The median is close to 1; the box plot shows that both the median and quartiles are concentrated in the middle. Nearby; in the normal QQ plot, the sample points closely match the theoretical normal distribution regression line, and the correlation coefficient Coefficient of determination This indicates that the bow attitude extracted under this matching combination fits the normal distribution very well.

[0089] The above extraction method was applied to segments of continuous yaw angle data from 06:23 to 07:16 (UTC+8) on May 21, 2025, for approximately 53 minutes, to obtain a complete heading angle sequence. Figure 5 ) and bow attitude sequence ( Figure 6 ).

[0090] The overall results show that the extracted heading angle data is smooth and continuous, and fits well with the large-scale trend of the original yaw angle data; the yaw attitude meets the characteristics of zero mean and stability, and there are no obvious artificial boundary traces on the time axis.

[0091] Of particular note is that between approximately 06:45 and 06:55, as the ship's heading angle shifted from approximately 100° to around 140°, the bow roll amplitude showed a synchronous decreasing trend. Conversely, between approximately 06:55 and 07:05, when the heading angle shifted back from 140° to approximately 100°, the bow roll amplitude synchronously increased and returned to its pre-shift level. This phenomenon can be explained by the change in the encounter angle between the ship and the waves: when the heading changes, the relative direction between the ship and the waves changes, and the encounter frequency changes accordingly, leading to a change in the bow roll response amplitude. The accurate capture of this phenomenon indirectly confirms the consistency and accuracy of the physical mechanism between the heading angle and bow roll attitude extracted by this method.

[0092] Since this invention performs short-term segmentation processing on long-term yaw angle data, it is necessary to quantitatively evaluate the continuity and correlation between each segment.

[0093] (1) Correlation test of overlapping areas: The extraction results of overlapping areas of all adjacent time segments in the whole data were statistically analyzed. The correlation coefficient of the heading angle data within the overlapping areas was calculated to be 0.99972, and the root mean square error (RMSE) was [missing value]. The correlation coefficient of the bow roll attitude data is 0.96167, and the root mean square error is... The correlation coefficient is extremely close to 1 and the root mean square error is very small, proving that the segmented processing results have a high degree of consistency in the overlapping areas.

[0094] (2) Smoothness check at data connection points: At the connection boundary between adjacent segments, the mean difference (connection error) between the heading angle / bow attitude values ​​at the end of the preceding segment and the beginning of the following segment is: The standard deviation of the error is only The connection error is near zero and fluctuates very little, indicating that the segmented extraction results have good continuity and smoothness on the time axis, meeting the requirements of engineering applications.

[0095] (3) Magnified verification of local details: magnified local details of the periods when the heading angle increases and decreases sharply (e.g., Figure 7 As shown in the figure, it can be clearly seen that the extraction results of each short data segment have good overlap in the time overlap area, and the transition at the turning point of the heading angle is natural; local magnification shows that the period when the bow roll attitude amplitude changes significantly (such as Figure 8 As shown in the figure, the good connection between the bow roll attitude data of each segment can also be seen.

[0096] The method proposed in this invention can be applied to, but is not limited to, the following scenarios: (1) Ship motion status monitoring and control system: The extracted heading angle is input into the autopilot or track control system, and the bow attitude is input into the roll / bow reduction control system to achieve more precise ship motion control.

[0097] (2) Wave parameter inversion: Using the extracted bow attitude sequence as input data for shipborne wave measurement, and combining it with the ship transfer function, key marine environmental parameters such as significant wave height, spectral peak period, and wave direction spectrum are inverted to provide data support for marine meteorological forecasting and marine engineering operations.

[0098] (3) Navigation data recording and analysis: The course angle and yaw attitude after separation are recorded separately to provide more detailed motion data for shipping companies’ post-navigation analysis, accident retrospection, energy efficiency optimization, etc.

[0099] (4) Marine science research: to provide high-quality measured motion data for the verification and analysis of wave-hull coupling response mechanism in oceanographic research.

Claims

1. A method for separating a ship's heading angle and bow roll attitude, characterized in that, include: The ship yaw angle data collected by the sensor is processed into short-time segments to obtain multiple short-time data segments; Perform empirical mode decomposition on each short-time data segment to obtain several intrinsic mode function components and one residual component; Iterate through all possible modal matching combinations, and match the residual component with the lowest frequency. m The intrinsic mode function components are combined to form a candidate heading angle, and the remaining high-frequency intrinsic mode function components are combined to form a candidate yaw attitude, wherein... m It is a positive integer; Multi-dimensional statistical feature evaluation is performed on the candidate bow roll attitudes under each modality matching combination; Based on the results of the multi-dimensional statistical feature evaluation, the optimal matching combination is determined; The final heading angle and yaw attitude are calculated based on the optimal matching combination. The extracted results of all short-time data segments are concatenated in chronological order to output a complete heading angle sequence and a yaw attitude sequence.

2. The method for separating ship heading angle and bow attitude according to claim 1, characterized in that, In the short-time segmentation process, the length of the preset time window is greater than or equal to the accumulation time of the wave spectrum in the sea area.

3. The method for separating ship heading angle and bow attitude according to claim 2, characterized in that, The actual length of each short-time data segment is 2 to 3 times the preset time window, and there is a time overlap area between adjacent short-time data segments.

4. The method for separating ship heading angle and bow attitude according to claim 1, characterized in that, The multidimensional statistical feature evaluation includes tests for normality of distribution, symmetry of probability density distribution, and frequency distribution characteristics.

5. The method for separating ship heading angle and bow attitude according to claim 4, characterized in that, The normality conformity test includes: Plot the normal quantile-quantile plot of the candidate bow roll attitude to obtain the scatter distribution of the quantile points of the candidate bow roll attitude relative to the theoretical quantiles of the normal distribution; A linear regression was performed on the scatter distribution to obtain the regression line; Calculate the correlation coefficient and determination coefficient between each data point in the scatter distribution and the regression line, as well as the Lilliefors test for the candidate bow roll attitude. p value; Based on the correlation coefficient, the determination coefficient, and the... p The value is used to determine whether the candidate bow attitude follows a normal distribution.

6. The method for separating a ship's heading angle and bow attitude according to claim 4, characterized in that, The probability density distribution symmetry test includes: Kernel density estimation is performed on the candidate bow roll attitude to obtain the probability density function of the candidate bow roll attitude; Calculate the integral ratio of the probability density function along the positive and negative semi-axis; and The median and quartile distributions of the candidate bow roll attitudes were examined using box plots to determine whether they were symmetric about the zero point.

7. The method for separating ship heading angle and bow attitude according to claim 4, characterized in that, The frequency distribution characteristic test includes: Perform a Fourier transform on the candidate bow roll attitude sequence to obtain the amplitude spectrum of the candidate bow roll attitude; Set an extremely low frequency threshold and check whether there are peak values ​​exceeding the preset amplitude threshold in the frequency range of the amplitude spectrum below the extremely low frequency threshold.

8. A device for separating a ship's heading angle from its bow roll attitude, characterized in that, include: The data segmentation module is configured to perform short-time segmentation processing on the ship yaw angle data collected by the sensor to obtain multiple short-time data segments; The Empirical Mode Decomposition (EMD) module is configured to perform EMD on each short-time data segment to obtain several intrinsic mode function (IMF) components and one residual component. The modal matching module is configured to iterate through all possible modal matching combinations and match the residual component with the lowest frequency. m The intrinsic mode function components are combined to form a candidate heading angle, and the remaining high-frequency intrinsic mode function components are combined to form a candidate yaw attitude, wherein... m It is a positive integer; The statistical evaluation module is configured to perform multi-dimensional statistical feature evaluation on the candidate bow attitudes for each matching combination. The optimal matching determination module is configured to determine the optimal matching combination based on the evaluation results of the multi-dimensional statistical features. An attitude calculation module is configured to calculate the final heading angle and yaw attitude based on the optimal matching combination. The result stitching module is configured to stitch together the extracted results of all short-time data segments in chronological order, and output a complete heading angle sequence and yaw attitude sequence.

9. A system for separating a ship's heading angle from its bow attitude, characterized in that, It includes a processor and a memory, the memory storing program instructions, and the processor executing the program instructions to implement the method for separating the ship's heading angle and yaw attitude as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the method for separating the ship's heading angle and yaw attitude as described in any one of claims 1 to 7.

Images