A marine diesel engine state intelligent monitoring and early warning system

Abstract

This invention proposes an intelligent monitoring and early warning system for marine diesel engine conditions, comprising: a wear data analysis module, used to acquire in real time thermal images, vibration signal data, stress wave signal data, exhaust gas particulate matter concentration, cylinder liner thickness, and oil wear metal particle data collected during the piston ring's movement along the cylinder liner, and perform normalization processing to form and generate a wear feature matrix X based on several types of collected data; and a wear feature purification module, including a purification parameter dynamic adjustment unit and a wear feature purification unit. The purification parameter dynamic adjustment unit can obtain several intrinsic mode functions I of each type of collected data based on the wear feature matrix X through a VMD model and generate a minimized envelope entropy, thereby solving the problem that existing intelligent monitoring systems are unable to effectively extract weak wear features and have insufficient generalization ability for transient low-speed data, resulting in deterioration of sensor signal-to-noise ratio and poor intelligent monitoring effect.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring and early warning technology for marine diesel engine status, specifically to an intelligent monitoring and early warning system for marine diesel engine status. Background Technology

[0002] Marine diesel engines require intelligent status monitoring and early warning to identify potential faults in real time, optimize maintenance cycles, and avoid sudden downtime, thereby ensuring navigation safety, reducing operating costs, and extending equipment life.

[0003] However, under low-speed maneuvering conditions of marine diesel engines, such as when navigating in port or narrow waterways, the wear signal amplitude of key friction pairs such as piston rings and cylinder liners is significantly reduced due to low speed and large load fluctuations, overlapping with the frequency band of background noise such as wave impact. Furthermore, existing intelligent monitoring systems based on fixed threshold alarm algorithms struggle to effectively extract subtle wear characteristics, while AI models trained on steady-state conditions lack the ability to generalize to transient low-speed data, often resulting in missed or delayed alarms. This leads to a deterioration in the sensor signal-to-noise ratio and suboptimal intelligent monitoring performance.

[0004] Therefore, how to design an intelligent monitoring and early warning system for the status of marine diesel engines that can improve the effectiveness of intelligent monitoring and provide timely warnings has become an urgent problem for technicians in this field. Summary of the Invention

[0005] The purpose of this invention is to solve the problems of existing intelligent monitoring systems, which are unable to effectively extract weak wear features and have insufficient generalization ability for transient low-speed data, resulting in deterioration of sensor signal-to-noise ratio and poor intelligent monitoring effect.

[0006] To achieve the above objectives, the present invention provides an intelligent monitoring and early warning system for the condition of a marine diesel engine, comprising: a wear data analysis module, used to acquire in real time thermal images, vibration signal data, stress wave signal data, exhaust gas particulate matter concentration, cylinder liner thickness, and oil wear metal particle data collected during the piston ring's movement along the cylinder liner, and to perform normalization processing to form and generate a wear feature matrix X based on several types of collected data; The wear feature purification module includes a purification parameter dynamic adjustment unit and a wear feature purification unit. The purification parameter dynamic adjustment unit can obtain several intrinsic mode functions I for each type of acquired data and generate a minimized envelope entropy based on the wear feature matrix X and the VMD model. The wear feature purification unit is used to calculate the kurtosis of each intrinsic mode function I for each type of acquired data according to the wear feature matrix X and the minimized envelope entropy, and selects the kurtosis to be greater than a preset feature threshold. For each intrinsic mode function I, calculate and generate the purified wear feature vector for each type of acquired data. ; The wear warning module includes a warning threshold dynamic update unit and a warning trigger unit. The warning threshold dynamic update unit can update the purified wear feature vector received within a specific window length. Calculate and generate warning thresholds that can be dynamically updated. The early warning triggering unit can calculate the purified wear feature vector. With warning threshold percentage difference Based on the percentage difference The system generates Level 1, Level 2, and Level 3 early warning commands from low to high levels and outputs them outwards.

[0007] Optionally, the wear data analysis module includes a data processing unit and a matrix generation unit. The data processing unit can perform corresponding denoising and normalization processing on the data acquired in real time by the multi-channel data acquisition card to form several types of acquired data. The matrix generation unit is used to extract features from each type of acquired data to generate a wear feature matrix X.

[0008] Optionally, the data processing unit employs a median filtering algorithm to remove salt-and-pepper noise from the thermal image, and a wavelet threshold denoising model to decompose the vibration signal data and stress wave signal data into several frequency sub-bands and process them through a preset threshold to remove noise. Additionally, a moving average filtering algorithm is used to denoise the exhaust gas particulate matter concentration, cylinder liner thickness, and oil wear metal particle data.

[0009] Alternatively, the wear feature matrix X is obtained by setting the acquisition time sequence to m and extracting n features: .

[0010] Optionally, the intrinsic mode functions I of each type of collected data are obtained through a VMD model, and it is assumed that each type of collected data in the wear feature matrix X has K intrinsic mode functions I, to obtain: In the formula: Let X represent the type of data collected in the wear feature matrix. ; Represented as a frequency index; Represented as The Fourier transform, and Let n represent the time series of the m-th type of collected data in the wear feature matrix X, where n represents the time index. Represented as The Fourier transform, and This is represented as the m-th type of collected data in the first position. The time series of the k-th intrinsic mode function during the next iteration; Represented as The Fourier transform, and This is represented as the m-th type of collected data in the first position. The sequence of Lagrange multipliers introduced in the next iteration; Represented as a secondary penalty factor; Let the k-th eigenmode function of the m-th type of collected data be represented in the k-th case. The center frequency at the next iteration.

[0011] Alternatively, the purification parameter dynamic adjustment unit obtains the analytical signal through Hilbert transform. : In the formula: Represents the imaginary unit, and ; And capable of analyzing signals Modulus extraction is performed to obtain the envelope signal. : , And based on envelope signal The minimized envelope entropy is generated by adjusting the parameters of the VMD model.

[0012] Alternatively, let the minimized envelope entropy be... By analyzing the envelope signal The normalization process allows for the normalization of the envelope signal. : , And based on the normalized envelope signal By adjusting the VMD model parameters, the minimized envelope entropy is generated as follows: : .

[0013] Alternatively, the kurtosis can be set as... Then the kurtosis of each intrinsic mode function I for: In the formula: Let be the amplitude of the i-th eigenmode function at time n, where N is the signal length; It is represented as the mean of several intrinsic mode functions I.

[0014] Optionally, the warning threshold It is achieved by calculating all purified wear feature vectors received within a specific window length. mean in each dimension : In the formula: Represented as a specific window length; Represented as the i-th purified wear feature vector within a specific window The j-th component; And based on the mean Calculate all purified wear feature vectors within a specific window. Standard deviation in each dimension : , Based on standard deviation , obtain the warning threshold : In the formula: and All are represented as weighting coefficients; M represents the purified wear feature vector received within a specific window length. The quantity.

[0015] Alternatively, the warning triggering unit can calculate the difference percentage. : , When At that time, a Level 1 warning instruction is generated. when At that time, a level-two early warning instruction is generated. when At that time, a three-level early warning command is generated and output to the outside; in, and All of these represent instruction partitioning thresholds.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: The intelligent monitoring and early warning system for marine diesel engines proposed in this invention achieves multi-dimensional monitoring of the piston and cylinder liner operating status by real-time acquisition and normalization of thermal images, vibration signal data, stress wave signal data, exhaust gas particulate matter concentration, cylinder liner thickness, and oil wear metal particle data, generating a wear feature matrix X. This provides an accurate data foundation. Furthermore, the system utilizes dynamically adjustable modal numbers K and penalty factors for each type of acquired data. It generates a minimized envelope entropy and calculates and generates a purified wear feature vector for each type of acquired data. It can effectively eliminate background noise interference and accurately extract weak wear features, thus significantly improving the sensor's signal-to-noise ratio; it also calculates and dynamically updates the warning threshold. And by calculating the purified wear feature vector With warning threshold percentage difference Based on the percentage difference The system generates and outputs Level 1, Level 2, and Level 3 early warning commands from low to high levels to avoid the limitations of fixed threshold alarm algorithms. It also abandons the AI ​​model trained under steady-state conditions, improving its adaptability to transient low-speed data. This enables timely and accurate early warnings, effectively solving the problem of missed or delayed alarms in existing intelligent monitoring systems. It significantly improves the effectiveness and reliability of intelligent monitoring of marine diesel engines under low-speed maneuvering conditions.

[0017] As can be seen from the above, the present invention can effectively solve the problems of existing intelligent monitoring systems that are unable to effectively extract weak wear features and have insufficient generalization ability for transient low-speed data, resulting in deterioration of sensor signal-to-noise ratio and poor intelligent monitoring effect.

[0018] Other features and advantages of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0020] Figure 1 This is a schematic diagram of a marine diesel engine status intelligent monitoring and early warning system according to an embodiment of the present invention. Detailed Implementation

[0021] To enable those skilled in the art to more fully understand the technical solutions of the present invention, exemplary embodiments of the present invention will be described more comprehensively and in detail below with reference to the accompanying drawings. Obviously, the one or more embodiments of the present invention described below are merely one or more specific ways to implement the technical solutions of the present invention, and are not exhaustive. It should be understood that other ways belonging to a general inventive concept can be used to implement the technical solutions of the present invention, and should not be limited to the embodiments described exemplary. Based on one or more embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0022] Reference Figure 1 Embodiments of the present invention provide an intelligent monitoring and early warning system for the condition of a marine diesel engine, comprising: The wear data analysis module is used to acquire thermal images, vibration signal data, stress wave signal data, exhaust gas particulate matter concentration, cylinder liner thickness, and oil wear metal particle data in real time during the piston ring's movement along the cylinder liner, and perform normalization processing to form and generate a wear feature matrix X based on several types of acquired data. The wear feature purification module includes a purification parameter dynamic adjustment unit and a wear feature purification unit. The purification parameter dynamic adjustment unit can obtain several intrinsic mode functions I for each type of acquired data and generate a minimized envelope entropy based on the wear feature matrix X through a variational mode decomposition (VMD) model. The wear feature purification unit is used to calculate the kurtosis of each intrinsic mode function I for each type of acquired data according to the wear feature matrix X and the minimized envelope entropy, and selects the kurtosis to be greater than a preset feature threshold. For each intrinsic mode function I, calculate and generate the purified wear feature vector for each type of acquired data. ; The wear warning module includes a warning threshold dynamic update unit and a warning trigger unit. The warning threshold dynamic update unit can update the purified wear feature vector received within a specific window length. Calculate and generate warning thresholds that can be dynamically updated. The early warning triggering unit can calculate the purified wear feature vector. With warning threshold percentage difference Based on the percentage difference The system generates Level 1, Level 2, and Level 3 early warning commands from low to high levels and outputs them outwards.

[0023] Specifically, in the embodiments of the present invention, thermal image acquisition is achieved using a thermal imager installed outside the cylinder liner, which can comprehensively cover the cylinder liner surface to clearly capture the temperature distribution on the cylinder liner surface as the piston ring moves along the cylinder liner. Vibration signal data is acquired using vibration sensors installed on the outer wall of the cylinder liner or on the diesel engine block near the cylinder liner, which can effectively sense vibrations caused by friction between the piston ring and the cylinder liner. Stress wave signal data is acquired using stress wave sensors installed on the outer wall of the key stress area of ​​the cylinder liner, which can accurately detect stress wave changes caused by friction. Exhaust gas particulate matter concentration is acquired by an exhaust gas particulate matter concentration sensor installed on the diesel engine exhaust pipe, which can monitor the particulate matter concentration in the exhaust gas in real time. Cylinder liner thickness is achieved using a non-contact cylinder liner thickness measuring instrument. Oil wear metal particle data is acquired using an oil wear metal particle sensor installed in the lubricating oil circuit of the diesel engine, which can detect the content of wear metal particles in the lubricating oil, thereby reflecting the wear condition of key friction pairs such as piston rings and cylinder liners.

[0024] In one specific embodiment, the wear data analysis module acquires thermal images, vibration signal data, stress wave signal data, exhaust gas particulate matter concentration, cylinder liner thickness, and oil wear metal particle data in real time during the piston ring's movement along the cylinder liner based on a multi-channel data acquisition card.

[0025] Specifically, the multi-channel data acquisition card has multiple independent and synchronously operating acquisition channels, which can simultaneously connect to sensors corresponding to different types of data such as thermal images and vibration signals, and achieve real-time acquisition of these data by virtue of its high sampling rate and fast data processing and transmission capabilities.

[0026] In one embodiment, the wear data analysis module includes a data processing unit and a matrix generation unit. The data processing unit can perform corresponding denoising and normalization processing on the data acquired in real time by the multi-channel data acquisition card to form several types of acquired data. The matrix generation unit is used to extract features from each type of acquired data to generate a wear feature matrix X.

[0027] In one specific embodiment, the data processing unit uses a median filtering algorithm to remove salt-and-pepper noise from the thermal image, and a wavelet threshold denoising model to decompose the vibration signal data and stress wave signal data into several frequency sub-bands and process them through a preset threshold to remove noise. Additionally, a moving average filtering algorithm is used to denoise the exhaust gas particulate matter concentration, cylinder liner thickness, and oil wear metal particle data.

[0028] In one specific embodiment, the normalization processing of the data acquired in real time by the multi-channel data acquisition card by the data processing unit is implemented based on the deviation normalization model.

[0029] Specifically, the median filtering algorithm, wavelet threshold denoising model, moving average filtering algorithm, and deviation standardization model are all existing technologies. Therefore, their specific processing principles will not be elaborated upon in this invention.

[0030] In one specific embodiment, the feature extraction of the thermal image by the matrix generation unit is achieved through the gray-level co-occurrence matrix algorithm.

[0031] Specifically, the gray-level co-occurrence matrix algorithm generates a gray-level co-occurrence matrix by statistically analyzing the spatial distribution characteristics of gray levels in an image, and then extracts feature parameters that reflect the texture of the thermal image.

[0032] In one specific embodiment, the feature extraction of vibration signal data and stress wave signal data by the matrix generation unit is achieved by the Fast Fourier Transform (FFT) algorithm.

[0033] Specifically, the Fast Fourier Transform (FFT) algorithm can convert time-domain signals into frequency-domain signals, decompose complex time-domain waveforms into combinations of sine waves of different frequencies, thereby analyzing the frequency components of vibration signal data and stress wave signal data, and identifying characteristic frequencies related to wear.

[0034] Furthermore, the characteristics of exhaust gas particulate matter concentration, cylinder liner thickness, and oil wear metal particle data can all be directly extracted from the normalized data.

[0035] It is worth noting that the specific extraction principles of the degree co-occurrence matrix algorithm and the fast Fourier transform algorithm mentioned above are existing technologies, and will not be elaborated further in this invention.

[0036] In one specific embodiment, the wear feature matrix X is obtained by setting the acquisition time sequence to m and extracting n features: .

[0037] Specifically, the embodiments of the present invention comprehensively and accurately mine feature information closely related to piston ring and cylinder liner wear by constructing the wear feature matrix X, providing a rich, accurate and structured data foundation for subsequent modules.

[0038] In one embodiment, the intrinsic mode functions I of each type of acquired data are obtained through variational mode decomposition (VMD), and it is assumed that each type of acquired data in the wear feature matrix X has K intrinsic mode functions I, to obtain: In the formula: Let X represent the type of data collected in the wear feature matrix. ; Represented as a frequency index; Represented as The Fourier transform, and Let n represent the time series of the m-th type of collected data in the wear feature matrix X, where n represents the time index. Represented as The Fourier transform, and This is represented as the m-th type of collected data in the first position. The time series of the k-th intrinsic mode function during the next iteration; Represented as The Fourier transform, and This is represented as the m-th type of collected data in the first position. The sequence of Lagrange multipliers introduced in the next iteration; Represented as a secondary penalty factor; Let the k-th eigenmode function of the m-th type of collected data be represented in the k-th case. The center frequency at the next iteration.

[0039] In one specific embodiment, the purification parameter dynamic adjustment unit obtains the analytic signal through Hilbert transform. : In the formula: Represents the imaginary unit, and ; And capable of analyzing signals Modulus extraction is performed to obtain the envelope signal. : , And based on envelope signal The minimized envelope entropy is generated by adjusting the parameters of the Variational Mode Decomposition (VMD) model.

[0040] In one specific embodiment, let the minimized envelope entropy be... By analyzing the envelope signal The normalization process allows for the normalization of the envelope signal. : , And based on the normalized envelope signal By adjusting the parameters of the Variational Mode Decomposition (VMD) model, the minimized envelope entropy is generated. : .

[0041] Specifically, embodiments of the present invention analyze each type of acquired data in the wear feature matrix X using a variational mode decomposition (VMD) model. The signal is adaptively decomposed into several eigenmode functions I with specific sparsity characteristics. During the decomposition process, parameters such as the Lagrange multiplier, the second-order penalty factor, and the center frequency are continuously adjusted to optimize the decomposition effect. Then, an analytic signal is obtained by combining this with the Hilbert transform. And obtain the envelope signal by taking the modulus. Then, the envelope signal After normalization, calculate the minimized envelope entropy. To minimize the envelope entropy By selecting more representative intrinsic mode functions, subtle wear characteristics can be accurately extracted. This effectively overcomes the shortcomings of existing intelligent monitoring systems that rely on fixed threshold alarm algorithms to extract subtle wear characteristics, thus providing strong support for the reliable operation of marine diesel engines under complex low-speed conditions.

[0042] In one specific embodiment, the kurtosis is set to... Then the kurtosis of each intrinsic mode function I for: In the formula: Let be the amplitude of the i-th eigenmode function at time n, where N is the signal length; It is represented as the mean of several intrinsic mode functions I.

[0043] In one specific embodiment, the wear feature purification unit uses kurtosis... Greater than the preset feature threshold The intrinsic mode function I is used to generate a purified wear feature vector. : .

[0044] Specifically, embodiments of the present invention calculate the kurtosis of each intrinsic mode function I according to a given formula. Furthermore, by using the ratio of the fourth-order central moment to the square of the variance, the amount of pulse component in the signal is accurately measured, highlighting the wear and impact characteristics. Simultaneously, a preset feature threshold is used... Based on the baseline, select kurtosis Greater than the preset feature threshold The intrinsic mode function I obtained contains more significant wear feature information, so as to accurately capture wear features, remove noise interference, and generate a reliable purified wear feature vector. This provides strong support for subsequent accurate monitoring and diagnosis, effectively improves the results of intelligent monitoring, and ensures the reliable operation of marine diesel engines.

[0045] In addition, preset feature thresholds The settings are specifically designed to meet the actual application needs of users, and are used to distinguish between normal signals and abnormal signals containing wear characteristics. This invention does not limit these settings.

[0046] In one specific embodiment, the warning threshold It is achieved by calculating all purified wear feature vectors received within a specific window length. mean in each dimension : In the formula: Represented as a specific window length; Represented as the i-th purified wear feature vector within a specific window The j-th component; And based on the mean Calculate all purified wear feature vectors within a specific window. Standard deviation in each dimension : , Based on standard deviation , obtain the warning threshold : In the formula: and All are represented as weighting coefficients; M represents the purified wear feature vector received within a specific window length. The quantity.

[0047] In one specific embodiment, the warning triggering unit calculates the difference percentage. : , When At that time, a Level 1 warning instruction is generated. when At that time, a level-two early warning instruction is generated. when At that time, a three-level early warning command is generated and output to the outside; in, and All of these represent instruction partitioning thresholds.

[0048] Specifically, embodiments of the present invention calculate the mean of each dimension of all purified wear feature vectors within a specific window length. with standard deviation Then, the warning threshold is derived by combining the weighting coefficients. This allows for full consideration of the overall distribution and fluctuations of data within a specific window, thus enabling the warning threshold to be adjusted accordingly. The generated data more closely reflects the actual wear characteristics.

[0049] Meanwhile, the early warning triggering unit calculates the purified wear feature vector. With warning threshold percentage difference It generates different levels of warning instructions based on preset instruction thresholds.

[0050] By dynamically adapting to changes in wear characteristics, accurately capturing abnormal wear, and issuing early warnings of different levels in a timely manner, we can avoid missed or delayed alarms, effectively improve the intelligent monitoring effect, and ensure the safe operation of equipment.

[0051] The intelligent monitoring and early warning system for marine diesel engines proposed in this invention achieves multi-dimensional monitoring of the piston and cylinder liner operating status by real-time acquisition and normalization of thermal images, vibration signal data, stress wave signal data, exhaust gas particulate matter concentration, cylinder liner thickness, and oil wear metal particle data, generating a wear feature matrix X. This provides an accurate data foundation. Furthermore, the system utilizes dynamically adjustable modal numbers K and penalty factors for each type of acquired data. It generates a minimized envelope entropy and calculates and generates a purified wear feature vector for each type of acquired data. It can effectively eliminate background noise interference and accurately extract weak wear features, thus significantly improving the sensor's signal-to-noise ratio; it also calculates and dynamically updates the warning threshold. And by calculating the purified wear feature vector With warning threshold percentage difference Based on the percentage difference The system generates and outputs Level 1, Level 2, and Level 3 early warning commands from low to high levels to avoid the limitations of fixed threshold alarm algorithms. It also abandons the AI ​​model trained under steady-state conditions, improving its adaptability to transient low-speed data. This enables timely and accurate early warnings, effectively solving the problem of missed or delayed alarms in existing intelligent monitoring systems. It significantly improves the effectiveness and reliability of intelligent monitoring of marine diesel engines under low-speed maneuvering conditions.

[0052] While one or more embodiments of the present invention have been described above, those skilled in the art will recognize that the present invention can be implemented in any other form without departing from its spirit and scope. Therefore, the embodiments described above are illustrative and not restrictive, and many modifications and substitutions will be apparent to those skilled in the art without departing from the spirit and scope of the invention as defined in the appended claims.

Claims

1. A smart monitoring and early warning system for the condition of a marine diesel engine, characterized in that, include: The wear data analysis module is used to collect data and generate the wear feature matrix X; The wear feature purification module is used to calculate and generate purified wear feature vectors for each type of acquired data. ; The wear warning module includes a warning threshold dynamic update unit and a warning trigger unit. The warning threshold dynamic update unit can update the purified wear feature vector received within a specific window length. Calculate and generate warning thresholds that can be dynamically updated. The early warning triggering unit can calculate the purified wear feature vector. With warning threshold percentage difference Based on the percentage difference The system generates Level 1, Level 2, and Level 3 early warning commands from low to high levels and outputs them outwards.

2. The intelligent monitoring and early warning system for marine diesel engine status according to claim 1, characterized in that, The wear data analysis module acquires thermal images, vibration signal data, stress wave signal data, exhaust gas particulate matter concentration, cylinder liner thickness, and oil wear metal particle data in real time during the piston ring's movement along the cylinder liner, and performs normalization processing to form and generate a wear feature matrix X based on several types of acquired data. The wear feature purification module includes a purification parameter dynamic adjustment unit and a wear feature purification unit. The purification parameter dynamic adjustment unit can obtain several intrinsic mode functions I for each type of acquired data and generate a minimized envelope entropy based on the wear feature matrix X and the VMD model. The wear feature purification unit is used to calculate the kurtosis of each intrinsic mode function I for each type of acquired data according to the wear feature matrix X and the minimized envelope entropy, and selects the kurtosis to be greater than a preset feature threshold. Each intrinsic mode function I.

3. The intelligent monitoring and early warning system for marine diesel engine status according to claim 2, characterized in that, The wear data analysis module includes a data processing unit and a matrix generation unit. The data processing unit can perform corresponding noise reduction and normalization processing on the data acquired in real time by the multi-channel data acquisition card to form several types of acquired data. The matrix generation unit is used to extract features from each type of acquired data to generate a wear feature matrix X. The data processing unit employs a median filtering algorithm to remove salt-and-pepper noise from the thermal image, and a wavelet threshold denoising model to decompose the vibration signal data and stress wave signal data into several frequency sub-bands and process them through preset thresholds to remove noise. Additionally, a moving average filtering algorithm is used to denoise the exhaust gas particulate matter concentration, cylinder liner thickness, and oil wear metal particle data.

4. The intelligent monitoring and early warning system for marine diesel engine status according to claim 3, characterized in that, The wear feature matrix X is obtained by setting the acquisition time sequence to m and extracting n features: 。 5. The intelligent monitoring and early warning system for marine diesel engine status according to claim 4, characterized in that, The intrinsic mode functions I of each type of acquired data are obtained through the VMD model, and it is assumed that each type of acquired data in the wear feature matrix X has K intrinsic mode functions I, to obtain: In the formula: Let X represent the type of data collected in the wear feature matrix. ; Represented as a frequency index; Represented as The Fourier transform, and Let n represent the time series of the m-th type of collected data in the wear feature matrix X, where n represents the time index. Represented as The Fourier transform, and This is represented as the m-th type of collected data in the first position. The time series of the k-th intrinsic mode function during the next iteration; Represented as The Fourier transform, and This is represented as the m-th type of collected data in the first position. The sequence of Lagrange multipliers introduced in the next iteration; Represented as a secondary penalty factor; Let the k-th eigenmode function of the m-th type of collected data be represented in the k-th case. The center frequency at the next iteration.

6. The intelligent monitoring and early warning system for marine diesel engine status according to claim 5, characterized in that, The purification parameter dynamic adjustment unit obtains the analytical signal through Hilbert transform. : In the formula: Represents the imaginary unit, and And capable of analyzing signals Modulus extraction is performed to obtain the envelope signal. : ， And based on envelope signal The minimized envelope entropy is generated by adjusting the parameters of the VMD model.

7. The intelligent monitoring and early warning system for marine diesel engine status according to claim 6 is characterized in that the minimized envelope entropy is... By analyzing the envelope signal The normalization process allows for the normalization of the envelope signal. ： ， And based on the normalized envelope signal By adjusting the VMD model parameters, the minimized envelope entropy is generated as follows: : 。 8. The intelligent monitoring and early warning system for marine diesel engine status according to claim 7, characterized in that, Let the kurtosis be... Then the kurtosis of each intrinsic mode function I for: In the formula: Let be the amplitude of the i-th eigenmode function at time n, where N is the signal length; It is represented as the mean of several intrinsic mode functions I.

9. The intelligent monitoring and early warning system for marine diesel engine status according to claim 8, characterized in that, The warning threshold It is achieved by calculating all purified wear feature vectors received within a specific window length. mean in each dimension : In the formula: Indicated as a specific window length; Represented as the i-th purified wear feature vector within a specific window The j-th component; And based on the mean Calculate all purified wear feature vectors within a specific window. Standard deviation in each dimension : ， Based on standard deviation , obtain the warning threshold : In the formula: and All are represented as weighting coefficients; M represents the purified wear feature vector received within a specific window length. The quantity.

10. The intelligent monitoring and early warning system for the condition of marine diesel engines according to claim 9, characterized in that, The early warning triggering unit calculates the difference percentage. : ， When At that time, a Level 1 warning instruction is generated. when At that time, a level-two early warning instruction is generated. when At that time, a three-level early warning command is generated and output to the outside; in, and All of these represent instruction partitioning thresholds.

Images