Marine gearbox oil pipe system blockage identification method based on complex domain blind separation
By arranging vibration and acoustic sensors at both ends of the lubricating oil piping system of a marine gearbox and using a complex domain blind separation algorithm to identify blockages, the problems of high cost and poor applicability of existing detection methods are solved, and fast and low-cost blockage identification is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHONGQING GEARBOX
- Filing Date
- 2026-04-30
- Publication Date
- 2026-07-31
AI Technical Summary
Existing methods for inspecting marine gearbox lubricating oil lines are costly, have poor applicability, and low inspection efficiency, making it difficult to quickly identify blockage faults while the equipment is in operation.
A complex domain blind separation algorithm is adopted. Vibration and acoustic sensors are placed at both ends of the lubricating oil pipeline system to acquire vibration signals and perform FFT spectrum transformation and complex domain blind separation. The time delay value is calculated to identify the blockage point.
It enables rapid, low-cost, and accurate identification of lubricating oil system blockages during normal gearbox operation, reducing hardware costs and complexity while improving detection efficiency and applicability.
Smart Images
Figure CN122493884A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the technical field of marine gearboxes, specifically to a method for identifying blockages in the lubricating oil piping system of marine gearboxes based on complex domain blind separation. Background Technology
[0002] Marine gearboxes are one of the core components of a ship's propulsion system. In response to the current situation of ships operating at high speeds and for extended periods, higher requirements have been placed on the reliability of gearboxes and their maintenance.
[0003] As a crucial component of the gearbox, the marine gearbox lubrication system plays a vital role in its safe operation. Due to its complex layout, numerous hydraulic components, and harsh working environment, problems such as jamming between the engine-driven pump screws, stuck check valves, blockages in the engine-driven pump inlet filters and throttling elements, stuck relief valves, and abnormal cooling oil flow are common. The accurate identification and resolution of these pipeline faults is a significant factor affecting the safe navigation of vessels. Blockage in the lubrication system can lead to insufficient gearbox lubrication, wear on the transmission gears, and in severe cases, even affect the engagement and disengagement time of the marine gearbox and the safe and stable operation of the vessel. Therefore, identifying blockages in the gearbox lubrication system is of paramount importance.
[0004] There are many methods for determining whether a gearbox lubrication circuit is blocked. For example, the existing technology "A Rapid Detection Method and Device for a Wind Turbine Gearbox Lubrication Circuit System" (Publication No.: CN107543630A) involves connecting an external oil supply line to the gearbox's lubrication circuit system. Heated and cooled lubricating oil are supplied to the system sequentially from this external line. The oil temperature at various points along the circuit is measured during each of the two supply cycles. If a section of the circuit is blocked, the temperature change in that section will be different, thus indicating whether the corresponding circuit is unobstructed. Another method is to observe abnormal lubrication pressure in marine gearbox lubrication pipeline systems to determine if there is a blockage. This requires arranging multiple acceleration sensors on the lubrication pipeline and applying a fixed load excitation to acquire acceleration data at multiple points along the pipeline. If a blockage exists, the vibration mode of the pipeline will change. The blockage can be identified by analyzing the vibration mode indicators of the lubrication pipeline system. However, the above methods still have the following technical problems: Using an external oil supply pipeline to detect oil temperature changes by observing the flow of hot and cold lubricating oil through the pipeline requires additional heating and cooling equipment and multiple temperature monitoring elements. This results in a complex structure, cumbersome operation, and high testing costs, making it impossible to achieve rapid and convenient on-site judgment. On the other hand, the vibration mode and pressure monitoring method not only requires the deployment of multiple sensors and the application of fixed force loads, but also incurs high costs. Furthermore, it requires testing in a stopped state, has high requirements for the testing environment, and is difficult to identify pipeline blockages while the system is running. This method has poor applicability and low testing efficiency.
[0005] This invention provides a method for identifying gearbox lubricating oil pipeline blockage vibration based on complex domain blind separation. It collects vibration and sound signals of the gearbox lubricating oil pipeline system through at least two vibration and sound sensors, and uses a blind separation algorithm based on complex domain to optimize the iterative process. This method can identify blockage faults in the lubricating oil pipeline system during operation, ensuring the normal operation of the gearbox lubricating oil pipeline system. Summary of the Invention
[0006] This invention provides a method for identifying blockages in marine gearbox lubricating oil piping systems based on complex domain blind separation, which can solve the problems of high cost, poor applicability, and low detection efficiency of existing methods for detecting lubrication circuits.
[0007] This application provides the following technical solution: a method for identifying blockages in marine gearbox lubricating oil piping systems based on complex domain blind separation, comprising the following steps: Step 1: Place the vibration and acoustic sensors at both ends of the oil filling pipeline system to acquire vibration data of the oil filling pipeline system. and The lubricating oil piping system between the placement positions of the vibration sound sensor includes elbows, tees, and non-straight pipe sections; Step 2: For and Performing an FFT spectral transformation yields the frequency domain (complex domain) form of the vibration sound signal from the lubricating oil piping system. and ; Step 3: Use the complex field blind separation algorithm to... and Perform separation calculations to obtain the blind separation matrix. and ; Step 4: Based on the complex-domain blind separation matrix and Calculate the output signal components in the complex frequency domain. and ; Step 5: [Regarding...] and By performing an inverse IFFT spectrum transform, the time-domain form of the separated signal can be obtained; Step 6: Pipeline vibration acoustic signal components and The time delay value is obtained by performing cross-correlation estimation; Step 7: Based on the time delay value, combined with the known pipe inner and outer diameters and length information, complete the pipe system blockage identification.
[0008] Beneficial effects: This solution acquires vibration signals by placing vibration sensors at both ends of the pipeline and purifies the blockage signal using a complex frequency domain blind separation algorithm. It eliminates the need for additional heating and cooling equipment and multi-point oil temperature monitoring elements, significantly reducing hardware costs and system complexity. The entire detection process can be completed while the gearbox is running normally, without requiring shutdown for external load excitation, thus having lower requirements for on-site conditions and installation environment, and greater applicability. Furthermore, it eliminates the need for complex data processing and long waiting times; oil circuit blockages can be quickly identified through signal separation and time delay calculation, significantly improving detection efficiency and real-time performance. This solution effectively and reliably addresses the problems of high cost, poor applicability, and low efficiency associated with existing oil circuit detection methods.
[0009] Furthermore, the vibration sound sensor is a PZT vibration sound sensor with a frequency range of 3Hz-6kHz and a resonant frequency of 19kHz.
[0010] Beneficial effects: The PZT vibration and acoustic sensor can convert mechanical vibration signals into electrical signals. It has strong shock resistance and overload resistance. The abnormal vibrations caused by blockages usually have energy concentrated in the tens to thousands of hertz. This allows the sensor to sensitively capture the subtle vibration changes caused by blockages and improve the accuracy of signal acquisition.
[0011] Furthermore, the vibration data of the oil-filled piping system and To collect vibration and acoustic data on the surface of the piping system.
[0012] Beneficial effects: and The vibration signals are detected by vibration and acoustic sensors at both ends of the piping system. The surface vibration signal can be stably acquired during normal equipment operation. Unlike the vibration modal method in the prior art, which requires stopping the machine to apply external loads, or the oil supply process, which requires interrupting the oil temperature method, the surface vibration signal can be stably acquired during normal equipment operation. This achieves online detection without stopping the equipment, significantly improving detection efficiency. Moreover, only two sensors are needed to complete the detection. Compared with multi-point sensor solutions, the deployment is simpler and more adaptable to the confined space of engine rooms or ship cabins.
[0013] Furthermore, the FFT spectrum transformation includes the following steps: Step 1: Vibration sound signal and The propagation along the oil-filled piping system is a convolution process; therefore, the following vibration-acoustic temporal convolution model is established: (1) in, , s(n) It is a leaking mixed sound source signal. It is a signal that blocks the vibration source. It is a non-direct vibration source signal. and It is the system mixing matrix.
[0014] Step 2: Transform the above vibration sound time-domain convolution model to the frequency domain, and obtain: (2) If z is usually ignored, then we have (3) Beneficial effects: Transforming the time-domain convolution model to the frequency domain yields multiple signals of different frequencies, allowing for better analysis of the signal's frequency components and facilitating subsequent separation of the desired signal. Furthermore, the frequency domain transformation simplifies the process by focusing on the linear relationships between signals, enabling the neglect of variable Z and simplifying the formula, thus improving computational efficiency.
[0015] Furthermore, the blocking vibration source signal This is a vibration signal generated by the flow of lubricating oil impacting obstructions; it is a non-direct vibration source signal. It is a vibration signal generated during the normal operation of the gearbox.
[0016] Beneficial effects: because s(n) Leakage mixed sound source signal refers to the original mixed signal of all vibration sounds generated inside the lubricating oil piping system. It mainly consists of two parts: one is the blocking vibration source signal. One signal is a special vibration signal generated when the lubricating oil flows and impacts the blockage in the pipeline; the other is a non-direct vibration source signal. This signal is the vibration signal of the gearbox itself during normal operation. Filtering out this non-direct vibration source signal will yield the blocking vibration source signal, which will help in subsequent identification of the blocking point.
[0017] Furthermore, the blind separation algorithm includes the following steps: Step 1: Mean removal and pre-whitening of the signal: (4) in It is the sample covariance matrix. , is the sample mean of the sampled signal, and N is the signal length.
[0018] Step 2: Initialize the blind separation matrix : (5) (6) It is iterative, and its update algorithm is as follows: (7) For error, ,in This is an estimated value.
[0019] Step 3: Let the nonlinear function : (8) calculate The value of .
[0020] Step 4: Take the contents of step 3... By incorporating the classic blind separation algorithm, a preliminary estimate of the original signal is obtained. .
[0021] Step 5: Further refine the estimated values obtained in Step 4 The detailed method is as follows: From preliminary estimates Calculate the fourth moment : (9) in Indicates the fourth moment, express s is a unit vector, and s is a signal.
[0022] Again ,parameter Let the value be 3.348, and calculate... .
[0023] Next, we obtained (10) (11) in, for The first derivative, .
[0024] Next, calculate the intermediate correction factor: (12) Next, calculate the intermediate correction matrix: (13) in .
[0025] Next, we calculate the separation matrix: (14) Therefore, its refined result is
[0026] Beneficial effects: 1. By performing mean removal and pre-whitening processing on the signal, the correlation between signals can be removed, making different signals more independent, thus facilitating subsequent separation. Zc This is a clean signal after whitening.
[0027] 2. Initializing the separation matrix w is similar to establishing a function that can filter interference signals, so that the signal can be filtered later. This function can be iterated and gradually corrected according to the error value.
[0028] 3. The nonlinear function g(x) amplifies abrupt signals and suppresses ordinary noise, thus obtaining the most abrupt signal that suppresses background noise.
[0029] 4. Substitute the two signal expressions obtained in step 3 of this step into the classic blind separation algorithm, with the aim of initially separating the blocking signal.
[0030] 5. First calculate the fourth moment. The purpose is to use mathematical indicators to distinguish which signal is more like a blocking signal, providing a basis for subsequent corrections. Calculating intermediate correction coefficients provides correction amounts for subsequent correction matrices, eliminating residual noise and interference between signals. Then, the optimal solution is calculated using the separation matrix. After passing through the separation matrix, a pure blocking signal is obtained. Based on this signal, the time delay information of the blocking channel can be extracted, that is, the time it takes for the blocking signal to propagate from the blocking point to one of the vibration and acoustic sensor observation points. The product of this time and the propagation time of the leakage signal is the distance between the blocking point and the observation point, thereby realizing the location of the blocking point when the pipe length and lubricating oil components are unknown.
[0031] Furthermore, in step 7, the location of the blockage point is determined by dividing the length between the blockage point and the two vibration sensors by the signal propagation speed. The time from the blockage point to the two vibration sensors can be calculated. The time difference is equal to the time delay value. An equation can be constructed to calculate the distance between the blockage point and the vibration sensors.
[0032] Beneficial effects: By constructing a mathematical equation for the propagation time and time delay from the blockage point to the two vibration sensors, the location of the blockage point can be directly determined using the known total length of the pipeline and the propagation speed of the vibration signal, thus achieving precise positioning of the blockage in the lubricating oil pipeline system. Attached Figure Description
[0033] Figure 1 This is a flowchart of the blocking vibration identification method of the present invention.
[0034] Figure 2 This is a flowchart of the blind separation algorithm based on the complex field of this invention. Detailed Implementation
[0035] The following detailed description illustrates the specific implementation method: Example 1 like Figure 1 and Figure 2 As shown, the method for identifying blockages in marine gearbox lubricating oil piping systems based on complex domain blind separation includes the following steps: Step 1: Place vibration and acoustic sensors at both ends of the oil filling pipeline to acquire vibration data of the oil filling pipeline. and The lubricating oil piping system between the placement locations of the vibration and acoustic sensors may include non-straight pipe sections such as elbows and tees; the vibration and acoustic sensors are PZT vibration and acoustic sensors with a frequency range of 3Hz-6kHz and a resonant frequency of 19kHz. Furthermore, the vibration data of the oil-filled piping system... and To collect vibration and acoustic data on the surface of the piping system Step 2: For and Performing an FFT spectral transformation yields the frequency domain (complex domain) form of the vibration sound signal from the lubricating oil piping system. and The FFT spectrum transformation includes the following steps: 1. Vibration sound signal and The propagation along the oil-filled piping system is a convolution process; therefore, the following vibration-acoustic temporal convolution model is established: (1) in, , s(n) It is a leaking mixed sound source signal. It is a signal that blocks the vibration source. It is a non-direct vibration source signal. and It is the system mixing matrix.
[0036] 2. Transforming the above vibration sound time-domain convolution model to the frequency domain, we obtain: (2) If z is usually ignored, then we have (3) Step 3: Use the complex field blind separation algorithm to... and Perform separation calculations to obtain the blind separation matrix. and The blind separation algorithm includes the following steps: 1. Mean removal and pre-whitening of the signal: (4) in It is the sample covariance matrix. , is the sample mean of the sampled signal, and N is the signal length.
[0037] 2. Initialize the blind separation matrix : (5) (6) It is iterative, and its update algorithm is as follows: (7) For error, ,in This is an estimated value.
[0038] 3. Let the nonlinear function : (8) calculate The value of .
[0039] 4. (The text appears to be incomplete and contains several errors. A more accurate translation would require the full context.) By incorporating the classic blind separation algorithm, a preliminary estimate of the original signal is obtained. .
[0040] 5. Further refine the estimated values obtained in step 4. The detailed method is as follows: From preliminary estimates Calculate the fourth moment : (9) in Indicates the fourth moment, express s is a unit vector, and s is a signal.
[0041] Again ,parameter Let the value be 3.348, and calculate... .
[0042] Next, we obtained (10) (11) in, for The first derivative, .
[0043] Next, calculate the intermediate correction factor: (12) Next, calculate the intermediate correction matrix: (13) in .
[0044] Next, we calculate the separation matrix: (14) Step 4: Based on the complex-domain blind separation matrix and Calculate the output signal components in the complex frequency domain. and The refined results are as follows:
[0045] Step 5: [Regarding...] and By performing an inverse IFFT spectrum transform, the time-domain form of the separated signal can be obtained; Step 6: Pipeline vibration acoustic signal components and Cross-correlation estimation is used to obtain the time delay value. Cross-correlation estimation is a signal processing operation method. Its formula belongs to the existing technology and is used to calculate the relationship between the similarity between two signals and the time delay. The obtained time delay value is the time difference between the vibration signal being transmitted from the blockage point to the two vibration sensors.
[0046] Step 7: Based on the time delay value and combined with the known pipe inner and outer diameters and length information, complete the pipe blockage identification. Using the lengths L1 and L2 between the blockage point and the two vibration sensors, respectively, and the total length of L1 and L2, can be measured. The time taken to travel from the blockage point to each of the two vibration sensors can be calculated using the ratio of the lengths L1 and L2 to the signal propagation speed. The difference in these times equals the time delay value. Therefore, an equation can be constructed to easily calculate the distance between the pipe blockage point and the vibration sensors.
[0047] The principle behind this solution is as follows: First, according to steps 1 and 2 in this embodiment, vibration sensors are arranged at both ends of the lubricating oil pipeline to collect mixed vibration signals x1(n) and x2(n) containing blocking vibration and background interference. A time-domain convolution model is established to clarify that the mixed signal is formed by the superposition of blocking vibration source and interference source through different propagation paths. Then, by converting the time-domain convolution model into a frequency-domain product model, the signal calculation is simplified while highlighting the unique frequency characteristics of blocking vibration and avoiding the problem of high complexity in time-domain analysis.
[0048] Next, according to steps 1-3 of step 3 in this embodiment, the frequency domain mixed signal is preprocessed by whitening. By calculating the covariance matrix and introducing the whitening matrix, the correlation between signals is eliminated, making the blockage vibration signal and the background interference signal independent of each other, laying the foundation for subsequent blind separation. The complex value separation matrix is initialized and iteratively optimized. Combined with the nonlinear activation function g(x), the abrupt features of the blockage vibration signal are amplified, noise interference is suppressed, and the feature differences between the two types of signals are enhanced.
[0049] Then, according to steps 4 and 5 of step 3 in this embodiment, the enhanced feature signal is introduced into the classical blind separation algorithm to complete the initial separation of the mixed signal and obtain the initial estimate of the blocking vibration signal containing a small amount of residual interference signal. The sharpness of the signal is measured by calculating the fourth moment to determine the core features of the blocking vibration signal. Then, the initial separation matrix is finely adjusted and normalized by a series of intermediate correction coefficients and correction matrices to finally obtain the optimal blind separation matrix.
[0050] Subsequently, according to step 4 of this embodiment, the whitened mixed signal is filtered using the optimal blind separation matrix to output a pure blocking vibration signal.
[0051] Finally, according to step 6 of this embodiment, by performing cross-correlation calculation on the pure blockage vibration signal, the time delay of the signals collected by the two sensors is obtained. The time delay is the time it takes for the blockage signal to propagate from the blockage point to one of the sensor observation points. The product of this time and the propagation speed of the vibration signal is the distance between the blockage point and the observation point, thereby realizing the location of the blockage point when the pipe length and lubricating oil components are unknown. Combined with the known parameters of the pipe system (inner and outer diameters, length), the blockage location of the lubricating oil pipe system can be accurately identified.
[0052] The advantages of this solution are: This solution requires no additional pipeline modifications or downtime for testing. Through signal purification and analysis, it addresses the pain points of existing detection methods, such as high cost, poor applicability, and low detection efficiency, achieving efficient, accurate, and low-cost identification of lubricating oil pipeline blockage. When problems such as jamming between the pump screws, check valve jamming, blockage of the pump suction filter throttling element, overflow valve jamming, and cooling oil valve jamming occur in the lubricating oil piping system, vibration signals of the gearbox lubricating oil piping system are collected by at least two vibration sensors. A blind separation algorithm based on the complex domain is used to introduce the blind system identification principle into the detection and localization of blockages in the lubricating oil system. The channel identification problem of a blockage point on the pipeline between two collection points is studied. When the source blockage signal cannot be obtained, the data from two signal observation points on the pipeline on both sides of the blockage are used to identify the channel response of the blockage signal propagation. The time delay information of the channel relative to the source blockage channel is extracted from this channel, that is, the time for the blockage signal to propagate from the blockage point to one of the observation points is extracted. The product of this time and the propagation speed of the vibration signal is the distance between the blockage point and the observation point. This enables the localization of the blockage point when the pipeline length and lubricating oil components are unknown. It enables the identification and accurate location of blockage faults in the lubricating oil piping system during gearbox operation, allowing for timely repair and ensuring the normal operation of the ship.
[0053] The above are merely embodiments of the present invention, and the invention is not limited to the fields covered by this embodiment. Commonly known structures and characteristics in the solutions are not described in detail here. It should be noted that those skilled in the art can make various modifications and improvements without departing from the structure of the present invention, and these should also be considered within the scope of protection of the present invention. These modifications and improvements will not affect the effectiveness of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.
Claims
1. A method for identifying blockages in marine gearbox lubricating oil piping systems based on blind separation of complex domains, characterized in that, Includes the following steps: Step 1: Place the vibration and acoustic sensors at both ends of the oil filling pipeline system to acquire vibration data of the oil filling pipeline system. and The lubricating oil piping system between the placement positions of the vibration sound sensor includes elbows, tees, and non-straight pipe sections; Step 2: For and Performing an FFT spectral transformation yields the frequency domain (complex domain) form of the vibration sound signal from the lubricating oil piping system. and ; Step 3: Use the complex field blind separation algorithm to... and Perform separation calculations to obtain the blind separation matrix. and ; Step 4: Based on the complex-domain blind separation matrix and Calculate the output signal components in the complex frequency domain. and ; Step 5: [Regarding...] and By performing an inverse IFFT spectrum transform, the time-domain form of the separated signal can be obtained; Step 6: Pipeline vibration acoustic signal components and The time delay value is obtained by performing cross-correlation estimation; Step 7: Based on the time delay value, combined with the known pipe inner and outer diameters and length information, complete the pipe system blockage identification.
2. The method for identifying blockages in marine gearbox lubricating oil piping systems based on complex domain blind separation according to claim 1, characterized in that: The vibration and sound sensor is a PZT vibration and sound sensor with a frequency range of 3Hz-6kHz and a resonant frequency of 19kHz.
3. The method for identifying blockages in marine gearbox lubricating oil piping systems based on complex domain blind separation according to claim 2, characterized in that: Vibration data of the oil filling pipeline system and To collect vibration and acoustic data on the surface of the piping system.
4. The method for identifying blockages in marine gearbox lubricating oil piping systems based on complex domain blind separation according to claim 3, characterized in that: The aforementioned FFT spectral transformation includes the following steps: Step 1: Vibration sound signal and The propagation along the oil-filled piping system is a convolution process; therefore, the following vibration-acoustic temporal convolution model is established: in, , s(n) It is a leaking mixed sound source signal. It is a signal that blocks the vibration source. It is a non-direct vibration source signal. and It is the system mixing matrix; Step 2: Transform the above vibration sound time-domain convolution model to the frequency domain, and obtain: If z is usually ignored, then we have 。 5. The method for identifying blockages in marine gearbox lubricating oil piping systems based on complex domain blind separation according to claim 4, characterized in that: The blocking vibration source signal This is a vibration signal generated by the flow of lubricating oil impacting obstructions; it is a non-direct vibration source signal. It is a vibration signal generated during the normal operation of the gearbox.
6. The method for identifying blockages in marine gearbox lubricating oil piping systems based on complex domain blind separation according to claim 4, characterized in that: The blind separation algorithm includes the following steps: Step 1: Mean removal and pre-whitening of the signal: in It is the sample covariance matrix. , It is the sample mean of the sampled signal, and N is the signal length; Step 2: Initialize the blind separation matrix : It is iterative, and its update algorithm is as follows: For error, ,in This is an estimated value; Step 3: Let the nonlinear function : calculate The value; Step 4: Take the contents of step 3... By incorporating the classic blind separation algorithm, a preliminary estimate of the original signal is obtained. ; Step 5: Further refine the estimated values obtained in Step 4 The detailed method is as follows: From preliminary estimates Calculate the fourth moment : in Indicates the fourth moment, express s is a unit vector, and s is a signal; Again ,parameter Let the value be 3.348, and calculate... ; Next, we obtained in, for The first derivative, ; Next, calculate the intermediate correction factor: Next, calculate the intermediate correction matrix: in ; Next, we calculate the separation matrix: Therefore, its refined result is 。 7. The method for identifying blockages in marine gearbox lubricating oil piping systems based on complex domain blind separation according to claim 1, characterized in that: In step 7, the location of the blockage point is determined by dividing the length between the blockage point and the two vibration sensors by the signal propagation speed. The time from the blockage point to the two vibration sensors can be calculated. The time difference is equal to the time delay value. An equation can be constructed to calculate the distance between the blockage point and the vibration sensors.