Pulse frequency self-adaptive matching method and system for cleaning deposited carbon in engine combustion chamber

Abstract

The invention provides a pulse frequency adaptive matching method and system for engine combustion chamber carbon deposition cleaning, and the method comprises the steps: obtaining the carbon deposition degree distribution data, cylinder pressure data and engine vibration signals of a target engine combustion chamber uploaded by an edge calculation node; performing dynamic weight fusion on the carbon deposition degree distribution data and the cylinder pressure data to generate an engine state comprehensive score; the initial pulse frequency, matched with the engine state comprehensive score, of the cleaning equipment is searched for in a preset pulse frequency mapping table; generating a pulse frequency correction strategy according to the engine vibration signal in combination with a pre-trained reinforcement learning model; and dynamically adjusting the initial pulse frequency according to the correction strategy to generate a target pulse frequency, and sending the target pulse frequency to the edge computing node to control the cleaning equipment to execute pulse cleaning according to the target pulse frequency. According to the invention, the carbon deposition removal effect is improved, and the equipment vibration risk is reduced.

Description

Technical Field

[0001] The present application relates to the technical field of engine maintenance, and in particular to a pulse frequency adaptive matching method and system for cleaning carbon deposits in an engine combustion chamber. Background Art

[0002] In the field of engine maintenance, cleaning combustion chamber carbon deposits is a critical step in ensuring engine performance. Traditional cleaning methods make it difficult to dynamically adjust cleaning parameters based on the actual carbon deposit status and operating conditions of the engine. This results in unstable cleaning results and can easily cause problems such as equipment vibration and incomplete cleaning.

[0003] Currently, existing solutions utilize fixed-frequency pulse cleaning technology, controlling the operation of the cleaning equipment within a preset pulse frequency range. This solution monitors basic engine parameters (such as speed and temperature) and matches them to a pre-stored frequency parameter table, selecting the appropriate cleaning frequency for the cleaning operation. During the cleaning process, simple threshold judgments are used to adjust the frequency to ensure that the equipment operates within a safe range.

[0004] While this solution can achieve basic automated cleaning, it relies solely on limited parameters and a static frequency table, failing to accurately match the engine's actual carbon deposit distribution and dynamic operating conditions. Its frequency adjustment lacks in-depth analysis of vibration signals and carbon deposit status, limiting cleaning effectiveness. It also fails to effectively suppress abnormal vibration during the cleaning process, potentially impacting engine performance over time. Summary of the Invention

[0005] The present application provides a pulse frequency adaptive matching method and system for cleaning carbon deposits in an engine combustion chamber, which is used to solve the problems of low carbon deposit removal effect and high equipment vibration risk in the prior art.

[0006] In a first aspect, the present application provides a pulse frequency adaptive matching method for cleaning carbon deposits in an engine combustion chamber, comprising:

[0007] Obtain carbon deposit distribution data, cylinder pressure data, and engine vibration signals of the target engine combustion chamber uploaded by the edge computing node;

[0008] Dynamically weighting and fusing the carbon deposit distribution data and the cylinder pressure data to generate a comprehensive engine status score;

[0009] Searching a preset pulse frequency mapping table for an initial pulse frequency of the cleaning device that matches the comprehensive score of the engine status;

[0010] Generate a pulse frequency correction strategy based on the engine vibration signal and a pre-trained reinforcement learning model;

[0011] The initial pulse frequency is dynamically adjusted according to the correction strategy to generate a target pulse frequency, and the target pulse frequency is sent to the edge computing node to control the cleaning equipment to perform pulse cleaning according to the target pulse frequency.

[0012] Optionally, generating a pulse frequency correction strategy based on the engine vibration signal in combination with a pre-trained reinforcement learning model includes:

[0013] Separate the pulse vibration component of the cleaning equipment and the vibration component of the engine body from the engine vibration signal;

[0014] identifying a resonant frequency band based on the pulse vibration component;

[0015] The resonant frequency band is input into a pre-trained reinforcement learning model, and the historical cleaning data and real-time cylinder temperature feedback data are combined to generate a pulse frequency correction strategy.

[0016] Optionally, the resonant frequency band is input into a pre-trained reinforcement learning model, and historical cleaning data and real-time in-cylinder temperature feedback data are combined to generate a pulse frequency correction strategy, including:

[0017] The resonant frequency band is input into a pre-trained reinforcement learning model, and a matching module of the reinforcement learning model is used to filter out associated records from historical cleaning data whose overlap with the resonant frequency band exceeds a preset overlap threshold. The associated records include historical pulse frequency adjustment amounts, corresponding historical in-cylinder temperature change curves, and vibration attenuation rates when carbon deposit removal is completed.

[0018] In the case where there are multiple associated records, the calculation module of the reinforcement learning model calculates the current temperature change rate based on the real-time in-cylinder temperature feedback data, and compares the current temperature change rate with each of the historical in-cylinder temperature change curves;

[0019] According to the multiple comparison results, the corresponding historical pulse frequency adjustment amount is adjusted to obtain an adjusted historical pulse frequency adjustment amount, and the adjusted historical pulse frequency adjustment amount constitutes an adjusted historical pulse frequency adjustment amount set;

[0020] A pulse frequency correction strategy is generated according to the adjusted historical pulse frequency adjustment amount set.

[0021] Optionally, generating a pulse frequency correction strategy according to the adjusted historical pulse frequency adjustment amount set includes:

[0022] Dividing the resonant frequency band into a plurality of sub-frequency bands, and assigning a weight value to each sub-frequency band, wherein the weight value is the ratio of the energy of the corresponding sub-frequency band to the total energy of the resonant frequency band;

[0023] For each sub-frequency band, selecting a target historical pulse frequency adjustment value from the adjusted historical pulse frequency adjustment value set, wherein the target historical pulse frequency adjustment value is the adjusted historical pulse frequency adjustment value corresponding to the sub-frequency band;

[0024] Calculating the mean of the adjustment amounts of the corresponding sub-bands according to the target historical pulse frequency adjustment amount;

[0025] Multiplying the mean of the adjustment amount by the weight value of the corresponding sub-frequency band to obtain a weighted correction amount;

[0026] Combining the weighted correction values ​​of all sub-bands into an initial correction value set;

[0027] A pulse frequency correction strategy is generated based on the initial correction amount set, combined with the maximum allowable pulse frequency of the engine and the current carbon deposit level.

[0028] Optionally, generating a pulse frequency correction strategy based on the initial correction amount set, in combination with the maximum allowable pulse frequency of the engine and the current carbon deposition level, includes:

[0029] According to the current carbon deposition level, the corresponding allowable floating value is found from the preset floating range mapping table;

[0030] Calculating an upper limit value of each weighted correction amount in the initial correction amount set according to the maximum allowable pulse frequency of the engine;

[0031] Calculating a lower limit value of each weighted correction amount according to the initial pulse frequency;

[0032] Based on the allowable floating value, the upper limit value and the lower limit value, constraining the weighted correction amount;

[0033] Arrange all constrained weighted corrections in a preset frequency band order to generate a pulse frequency correction strategy.

[0034] Optionally, adjusting the corresponding historical pulse frequency adjustment amount according to the multiple comparison results to obtain the adjusted historical pulse frequency adjustment amount includes:

[0035] Extracting the historical temperature change rate within the same time period as the current cleaning time point from the historical in-cylinder temperature change curve based on the multiple comparison results;

[0036] Calculating the absolute value of the difference between the current temperature change rate and the historical temperature change rate;

[0037] According to the absolute value of the difference, searching for a corresponding correction coefficient from a preset coefficient mapping table;

[0038] The historical pulse frequency adjustment amount in each associated record is multiplied by the corresponding correction coefficient to obtain the adjusted historical pulse frequency adjustment amount.

[0039] Optionally, identifying a resonant frequency band based on the pulse vibration component includes:

[0040] Dividing the pulse vibration component into a plurality of analysis frequency bands according to preset frequency intervals;

[0041] For each analysis frequency band, calculating the average energy value of the analysis frequency band within a preset time window;

[0042] Compare the average energy values ​​of all analyzed frequency bands and select candidate frequency bands whose energy values ​​exceed the preset energy threshold;

[0043] Merge candidate frequency bands with adjacent frequencies and energy values ​​exceeding a preset energy threshold to form continuous energy peaks;

[0044] A frequency band of the continuous energy peaks whose duration exceeds a preset duration threshold is determined as a resonant frequency band.

[0045] In a second aspect, the present application provides a pulse frequency adaptive matching system for cleaning carbon deposits in an engine combustion chamber, comprising:

[0046] The acquisition module is used to obtain the carbon deposit distribution data, cylinder pressure data and engine vibration signal of the target engine combustion chamber uploaded by the edge computing node;

[0047] A first generating module is configured to dynamically weight and fuse the carbon deposit distribution data and the cylinder pressure data to generate a comprehensive engine status score;

[0048] A search module, configured to search a preset pulse frequency mapping table for an initial pulse frequency of a cleaning device that matches the comprehensive score of the engine state;

[0049] A second generation module is used to generate a pulse frequency correction strategy based on the engine vibration signal and a pre-trained reinforcement learning model;

[0050] The third generation module is used to dynamically adjust the initial pulse frequency according to the correction strategy, generate a target pulse frequency, and send the target pulse frequency to the edge computing node to control the cleaning equipment to perform pulse cleaning according to the target pulse frequency.

[0051] In a third aspect, the present application provides a computing device comprising a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute a pulse frequency adaptive matching method for cleaning carbon deposits in an engine combustion chamber as described in any one of the first aspects.

[0052] In a fourth aspect, the present application provides a computer storage medium having computer program instructions stored thereon, which, when executed by a processor, implements a pulse frequency adaptive matching method for cleaning carbon deposits in an engine combustion chamber as described in any one of the first aspects.

[0053] In the present application, a pulse frequency adaptive matching method for carbon deposit cleaning in an engine combustion chamber is provided, which includes: obtaining carbon deposit distribution data, cylinder pressure data and engine vibration signal of a target engine combustion chamber uploaded by an edge computing node; dynamically weighted fusion of the carbon deposit distribution data and the cylinder pressure data to generate a comprehensive engine status score; searching for an initial pulse frequency of a cleaning device that matches the comprehensive engine status score from a preset pulse frequency mapping table; generating a pulse frequency correction strategy based on the engine vibration signal and a pre-trained reinforcement learning model; dynamically adjusting the initial pulse frequency according to the correction strategy to generate a target pulse frequency, and sending the target pulse frequency to the edge computing node to control the cleaning device to perform pulse cleaning according to the target pulse frequency.

[0054] The technical solution provided by this application has the following beneficial effects:

[0055] This application uses edge computing nodes to collect the degree of carbon deposits in the engine combustion chamber, cylinder pressure, and vibration signals in real time, providing a comprehensive data foundation for intelligent cleaning and ensuring the accuracy of parameter decisions. The carbon deposit degree and cylinder pressure data are combined to generate a status score, quantify the actual working conditions of the engine, and overcome the limitations of single parameter judgment. Based on the status score, the initial pulse frequency is matched from the preset table to achieve rapid adaptation of cleaning parameters and avoid manual trial and error. The vibration signal is combined with the reinforcement learning model to dynamically optimize the frequency, and abnormal vibrations during the cleaning process are suppressed in real time to improve safety. The target pulse frequency is adjusted and executed in real time by the edge node, forming a "perception-decision-execution" closed loop to ensure stable and efficient cleaning results.

[0056] Furthermore, this application also separates the equipment pulse component and the engine body vibration from the engine vibration signal, accurately identifies the resonant frequency band, and inputs it into the reinforcement learning model, combining the historical cleaning effect and real-time in-cylinder temperature feedback to generate a dynamic frequency correction strategy.

[0057] In addition, through vibration signal separation and resonance frequency band identification, the vibration source that needs to be suppressed can be accurately located; reinforcement learning decision-making based on historical data and real-time temperature feedback can achieve intelligent dynamic adjustment of pulse frequency, effectively reducing resonance risk and optimizing cleaning energy efficiency, while avoiding cleaning damage caused by abnormal temperature.

[0058] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0060] Figure 1 A flow chart of a pulse frequency adaptive matching method for cleaning carbon deposits in an engine combustion chamber provided in an embodiment of the present application;

[0061] Figure 2 A schematic diagram of the structure of a pulse frequency adaptive matching system for cleaning carbon deposits in an engine combustion chamber provided in an embodiment of the present application;

[0062] Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0063] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0064] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0065] Researchers have found that existing engine carbon deposit cleaning technologies have problems such as parameter adjustment lag and inability to accurately match the real-time operating conditions of the engine, resulting in unstable cleaning effects and easy to cause equipment vibration. Based on this, an embodiment of the present application provides a pulse frequency adaptive matching method for engine combustion chamber carbon deposit cleaning. This method collects multi-dimensional data such as carbon deposit level, cylinder pressure and vibration signals in real time through edge computing nodes, generates a comprehensive engine status score through dynamic weighted fusion, determines the initial pulse frequency in combination with a preset mapping table, and uses a reinforcement learning model to dynamically correct the frequency parameters according to the vibration signal, ultimately achieving intelligent optimization and adjustment of the cleaning pulse frequency. The technical solution of the present application can be applied to various internal combustion engine maintenance scenarios, especially in the maintenance of high-end equipment such as aircraft engines and ship engines that have high requirements for cleaning accuracy and safety.

[0066] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0067] Figure 1 A flow chart of a pulse frequency adaptive matching method for cleaning carbon deposits in an engine combustion chamber provided in an embodiment of the present application is shown as follows: Figure 1 As shown, the method includes:

[0068] Step 101: Obtain carbon deposit distribution data, cylinder pressure data, and engine vibration signal of the target engine combustion chamber uploaded by the edge computing node.

[0069] In this step, the carbon deposit distribution data represents the quantitative data on the carbon deposit coverage and thickness in various areas of the combustion chamber, obtained through endoscopic imaging technology, expressed in percentage and millimeters. The cylinder pressure data represents the time-series variation of the in-cylinder pressure during the engine operating cycle, collected by the cylinder pressure sensor, expressed in MPa. The engine vibration signal represents the wideband vibration waveform, with a frequency range of 0-5 kHz, collected by a triaxial accelerometer mounted on the engine block.

[0070] In the embodiment of this application, an endoscopic imaging system is first used to capture images of the combustion chamber interior. A deep learning image segmentation algorithm is then used to identify carbon deposit areas and calculate the coverage area and thickness distribution, generating carbon deposit distribution data. Simultaneously, a high-precision cylinder pressure sensor is used to monitor cylinder pressure changes in real time, and cylinder pressure waveform data is obtained after processing by a signal conditioning circuit. A vibration sensor installed in the engine block captures the raw vibration signal, which is then subjected to bandpass filtering and noise reduction to extract effective vibration features. The edge computing node time-synchronizes and marks the three types of data, then uploads them to a cloud processing center via an encrypted communication protocol to ensure the real-time and integrity of the data collection.

[0071] For example, during routine maintenance on a certain model of marine diesel engine, the carbon deposit cleaning process was triggered. Technicians first inserted an endoscopic probe into the combustion chamber of the third cylinder, where a high-definition camera captured uneven carbon deposits on the piston top. The image processing system divided the combustion chamber into 36 grid cells and measured the carbon deposit thickness in each area, ranging from 0.08 to 0.25 mm. The area around the exhaust valve was the thickest, resulting in a weighted average of 0.18 mm for the overall carbon deposit thickness. Simultaneously, the cylinder pressure sensor recorded a peak compression stroke pressure of 7.2 MPa, 10% below the standard value of 8.0 MPa. The vibration sensor detected abnormal vibration energy peaks in the 900-1300 Hz frequency band.

[0072] Step 102: Dynamically weight the carbon deposit distribution data and the cylinder pressure data to generate a comprehensive engine status score.

[0073] In this step, dynamic weighted fusion refers to a feature fusion method that automatically adjusts the weights of various parameters based on the current engine operating conditions. The engine status comprehensive score is a quantitative indicator of the overall health of the engine, ranging from 0 to 100 points.

[0074] In this embodiment, the carbon deposit data is first spatially gridded, and a weighted carbon deposit thickness value is calculated for each grid cell. Simultaneously, the cylinder pressure data is waveform-featured and its similarity to a standard pressure curve is calculated. An adaptive weighting algorithm is employed to dynamically adjust the weighting coefficients of the carbon deposit and pressure data based on the current engine speed and load conditions, and the two features are normalized and fused. A temperature compensation factor is introduced during the fusion process to eliminate the influence of ambient temperature. Finally, a linear conversion is performed to generate a comprehensive engine status score ranging from 0 to 100.

[0075] For example, after receiving the above data, the system automatically sets the carbon deposit weighting factor to 0.65 and the pressure weighting factor to 0.35 based on the current engine speed (1200 rpm) and load (75%). Using the formula: Condition Score = 100 - (Carbon Deposit Thickness Index × 200 × 0.65 + Pressure Deviation Percentage × 0.35 × 80), the resulting overall engine condition score is 76 (with 15.6 points deducted for carbon deposits and 2.8 points deducted for pressure, for a total of 18.4 points).

[0076] Step 103: searching a preset pulse frequency mapping table for an initial pulse frequency of the cleaning device that matches the comprehensive score of the engine status.

[0077] In this step, the pulse frequency mapping table represents a three-dimensional lookup table (score×temperature×rotation speed dimension) storing the score-frequency correspondence relationship. The initial pulse frequency represents the basic operating frequency of the cleaning device, in Hz.

[0078] In this embodiment, the system maintains a multidimensional pulse-frequency mapping database, storing the optimal initial frequencies for different combinations of engine status scores, ambient temperatures, and speeds. During a query, the input parameters are first discretized, and a nearest neighbor search algorithm is used to locate the best matching record in the database. To improve query accuracy, bilinear interpolation is performed on adjacent records, and the results are validated. The final output initial pulse frequency must meet the device's safe operating range constraints.

[0079] For example, based on the score, the system queries a preset three-dimensional pulse frequency mapping table (dimensions include score, cooling water temperature, and rotational speed). For input parameters of 76 points, 85°C water temperature, and 1200 rpm, bilinear interpolation calculates the initial pulse frequency to be 23.6 Hz. The frequency verification module confirms that this value is within the device's permitted operating range of 18-30 Hz.

[0080] Step 104: Generate a pulse frequency correction strategy based on the engine vibration signal and a pre-trained reinforcement learning model.

[0081] In this step, the reinforcement learning model represents a frequency optimization decision model based on proximal policy optimization, which includes a 128-dimensional state space. The frequency correction strategy includes control instructions for frequency offset and adjustment rate.

[0082] In this embodiment, the vibration signal is first decomposed into frequency bands using a wavelet transform to extract the energy distribution of characteristic frequency bands relevant to pulse cleaning. A reinforcement learning model receives vibration characteristics, current pulse frequency, and engine state parameters as input, and outputs the frequency adjustment direction and amplitude through a policy network. Actual cleaning data is used for model training, with optimization objectives including vibration suppression and carbon deposit removal efficiency. During online inference, the model generates a frequency correction suggestion every second, which is then reviewed by the safety module to form the final correction strategy.

[0083] In addition, it should be noted that the above steps can be implemented when the engine is not disassembled.

[0084] For example, after cleaning began, the system detected a significant resonance peak at 1050 Hz in the engine vibration signal while operating at 23.6 Hz, with the amplitude exceeding the safety threshold by 30%. The reinforcement learning model analyzed the vibration spectrum characteristics, combined with a historical similar case (cleaning record number C-217) and the current cylinder head temperature (182°C), and output a frequency correction strategy: it recommended reducing the frequency to 21.8 Hz at a rate of 1.8 Hz per minute, which is expected to reduce the vibration amplitude to within a safe range.

[0085] Step 105: Dynamically adjust the initial pulse frequency according to the correction strategy to generate a target pulse frequency, and send the target pulse frequency to the edge computing node to control the cleaning device to perform pulse cleaning according to the target pulse frequency.

[0086] In this step, the target pulse frequency represents the final execution frequency after closed-loop optimization. Dynamic adjustment represents the proportional-integral-derivative control process with rate limitation.

[0087] In this embodiment, the system establishes a closed-loop control circuit to convert the correction strategy output by the reinforcement learning model into specific frequency adjustment instructions. A progressive adjustment algorithm is used to calculate the allowable frequency change within each control cycle, ensuring a smooth adjustment process. Vibration signal changes are monitored in real time, and adjustments are stopped when the desired effect is achieved. The final target pulse frequency is transmitted to the cleaning equipment controller via a digital communication interface, and the frequency mapping database is updated for subsequent use.

[0088] For example, the system executed an adjustment, smoothly adjusting the frequency from 23.6 Hz to 21.8 Hz in 6 minutes and 40 seconds using a closed-loop control algorithm. Real-time monitoring during the adjustment process showed that when the frequency dropped to 22.3 Hz, the 1050 Hz vibration amplitude had decreased by 45%. At the target frequency of 21.8 Hz, the vibration amplitude stabilized below the safety threshold. Simultaneously, real-time endoscope images showed that the carbon deposit removal effect met the expected standard, completing the adaptive cleaning operation.

[0089] This method achieves precise matching of operating frequencies during the engine carbon deposit cleaning process through intelligent multi-parameter fusion and adaptive control. The system automatically optimizes cleaning parameters based on the actual engine status, effectively suppressing abnormal vibration while ensuring cleaning effectiveness, extending engine life and reducing maintenance costs. It is suitable for preventive maintenance scenarios for various internal combustion engines.

[0090] To address the issues of equipment damage and low cleaning efficiency caused by vibration during the engine carbon deposit cleaning process, in some embodiments, step 104: generating a pulse frequency correction strategy based on the engine vibration signal in combination with a pre-trained reinforcement learning model, includes:

[0091] Step 201: Separate the pulse vibration component of the cleaning device and the engine body vibration component from the engine vibration signal.

[0092] In step 201, the pulse vibration component refers to the specific frequency vibration signal generated by the cleaning equipment during operation. It has periodic characteristics and the frequency range is related to the operating frequency of the equipment. The engine body vibration component refers to the broadband vibration signal generated during normal engine operation, which includes multiple vibration sources such as mechanical movement and combustion shock.

[0093] In the embodiment of the present application, an improved blind source separation technique is used to process the original vibration signal. First, a time-frequency analysis is performed on the vibration signal collected by the three-axis acceleration sensor, and the energy characteristics of each frequency band are extracted through wavelet transform. Then, a vibration feature matrix is ​​constructed, and the components with pulse periodic characteristics are separated using the independent component analysis algorithm. Finally, through correlation verification, the component in the separation result that is most relevant to the operating frequency of the cleaning equipment is identified as the pulse vibration component, and the rest is classified as the engine body vibration component. A sliding time window is used in the separation process to ensure real-time performance, and the window size is dynamically adjusted according to the engine speed.

[0094] Step 202: Identify a resonant frequency band based on the pulse vibration component.

[0095] In step 202, the resonant frequency band refers to a frequency range where the energy of the pulse vibration component is abnormally concentrated, which is caused by the coupling between the operating frequency of the equipment and the natural frequency of the engine structure.

[0096] In the embodiments of the present application, resonance detection is performed on the separated pulse vibration components. First, the signal's power spectral density is estimated, and an adaptive peak detection algorithm is used to identify frequency points with sudden energy increases. The energy integral of the frequency band surrounding each frequency point is then calculated and compared with a dynamic threshold. Finally, frequency regions that continuously exceed the threshold are merged into resonant frequency bands, and their center frequency, bandwidth, and peak energy are recorded. A speed tracking mechanism is introduced during the detection process to exclude normal vibration frequency bands associated with engine operating harmonics.

[0097] Step 203: Input the resonant frequency band into a pre-trained reinforcement learning model, combine historical cleaning data with real-time in-cylinder temperature feedback data, and generate a pulse frequency correction strategy.

[0098] In step 203, historical cleaning data includes vibration characteristics, frequency adjustment records, and cleaning effect evaluations of past successful cleaning cases. Real-time in-cylinder temperature feedback refers to real-time combustion chamber temperature data collected by embedded temperature sensors.

[0099] In this embodiment, a frequency decision-making system based on deep reinforcement learning is constructed. The model receives the resonance frequency band characteristics, the current pulse frequency, and the in-cylinder temperature as state inputs. Using a three-layer fully connected network, it calculates the value of each possible adjustment action. It selects the optimal action and outputs a correction strategy that includes the direction and magnitude of the frequency adjustment.

[0100] Here's a specific example:

[0101] During the carbon deposit cleaning process of a certain type of marine diesel engine, the system detected abnormal fluctuations in the original vibration signal in the 800-1500 Hz range. (Fast Fourier transform calculations, which determined the energy distribution of each frequency band, revealed that the energy value at 1120 Hz reached 0.15 g² / Hz, exceeding the preset safety threshold of 0.08 g² / Hz.) The vibration signal was first processed using an improved blind source separation technique: a 120-second sliding time window was constructed, and an independent component analysis algorithm was used to isolate the pulse vibration component. (The harmonic correlation of each component with the equipment's operating frequency of 25 Hz was calculated, and the component with the highest correlation coefficient of 0.92 was identified as the pulse vibration component.) Resonance detection was performed based on this component: within the 1120±35 Hz range, the signal energy integral reached 125 (calculated using the formula ∑(f=1085 to 1155)PSD(f)·Δf), exceeding that of adjacent frequency bands. After receiving this resonance feature, the reinforcement learning model combined it with three similar cases in the historical database (numbered C205 / C312 / C419, with an average cooling rate of 2.5°C / minute) and the current cylinder temperature of 185°C (measured by an embedded thermocouple). It calculated that the optimal adjustment strategy was to reduce the frequency from 25 Hz by 1.5 Hz (adjustment amount = base value 1.2 Hz × temperature compensation coefficient 1.25), with the adjustment rate controlled within 0.25 Hz / second.

[0102] In this application's embodiment, precise vibration signal separation and resonance identification, combined with an intelligent decision-making model, achieve real-time dynamic optimization of cleaning frequency. This method effectively suppresses harmful vibrations and improves cleaning efficiency, while avoiding the parameter adjustment lag associated with traditional methods that rely on manual experience. It is particularly suitable for preventive maintenance scenarios for high-value engines.

[0103] To further improve the accuracy and safety of pulse frequency adjustment during the engine carbon deposit cleaning process, in some embodiments, step 203: inputting the resonant frequency band into a pre-trained reinforcement learning model, combining historical cleaning data with real-time in-cylinder temperature feedback data, and generating a pulse frequency correction strategy includes:

[0104] Step 301: The resonant frequency band is input into a pre-trained reinforcement learning model. Through the matching module of the reinforcement learning model, associated records whose overlap with the resonant frequency band exceeds a preset overlap threshold are screened out from the historical cleaning data. The associated records include the historical pulse frequency adjustment amount, the corresponding historical cylinder temperature change curve, and the vibration attenuation rate when carbon deposit removal is completed.

[0105] In step 301, the matching module, representing the feature comparison component in the reinforcement learning model, uses the cosine similarity algorithm to calculate the frequency band feature matching degree. The overlap threshold is set to 70%, indicating that the historical record and the current resonant frequency band must have a frequency range of more than 70% overlap.

[0106] In this embodiment, the matching module first performs feature encoding on the resonant frequency band, extracting three key features: center frequency, bandwidth, and peak energy. It then performs a k-nearest neighbor search on the historical database, filtering out records with feature distances less than a threshold. Each associated record contains: a frequency adjustment (e.g., +1.2Hz), a temperature change curve (time-temperature data points), and a final vibration attenuation rate (e.g., 65%).

[0107] Step 302: When there are multiple associated records, the current temperature change rate is calculated based on the real-time in-cylinder temperature feedback data through the calculation module of the reinforcement learning model, and the current temperature change rate is compared with each of the historical in-cylinder temperature change curves.

[0108] In step 302, the calculation module represents the numerical processing unit in the reinforcement learning model, responsible for extracting and comparing temperature variation features. The temperature change rate represents the slope of the current in-cylinder temperature change over time, measured in °C / min. The specific process for calculating the current temperature change rate is as follows: the system uses a high-precision temperature sensor to collect in-cylinder temperature data at a frequency of once per second, continuously acquiring the temperature values ​​of the last 30 time points to form a time series. This series is then linearly fitted using the least squares method to obtain the linear equation for the temperature change over time: T = kt + b (where T is temperature, t is time, k is the slope, and b is the intercept). The slope k of the fitted line is used as the current temperature change rate, which physically represents the temperature change per minute (°C / min). During the calculation process, the system removes abnormal temperature sampling points (data with a temperature difference of more than 5°C from the previous point is considered abnormal). A sliding window mechanism is used to ensure real-time rate calculation, updating the results every 5 seconds. For example, if the fitted result is k = 2.3, it means that the current in-cylinder temperature is increasing at a rate of 2.3°C per minute.

[0109] In this embodiment, the system uses the least squares method to fit the current temperature change curve using the time series data from the temperature sensor to calculate the instantaneous rate of change. Simultaneously, the system analyzes the historical temperature curves in each associated record to extract the rate of change at the same time point. A matrix of the difference between the current and historical rates is then constructed for subsequent adjustment corrections.

[0110] Step 303: adjusting the corresponding historical pulse frequency adjustment amount according to the multiple comparison results to obtain adjusted historical pulse frequency adjustment amounts, wherein the adjusted historical pulse frequency adjustment amounts constitute an adjusted historical pulse frequency adjustment amount set.

[0111] In step 303, the adjusted historical pulse frequency adjustment value represents a correction factor calculated based on the temperature variation difference, ranging from 0.5 to 1.5. The adjusted historical pulse frequency adjustment value set includes multiple data sets of adjusted adjustments for final strategy generation.

[0112] In this embodiment, the temperature rate difference ratio is calculated for each associated record: Difference ratio = 1 + |current rate - historical rate| / historical rate. If the difference ratio is greater than 1.2, the adjustment factor is calculated as 1 / difference ratio; otherwise, the adjustment factor is calculated as 1. The historical adjustment amount is multiplied by the corresponding adjustment factor to obtain the corrected adjustment amount. All valid results are retained to form a set.

[0113] Step 304: Generate a pulse frequency correction strategy based on the adjusted historical pulse frequency adjustment amount set.

[0114] In this embodiment, a weighted average of the set of adjustment values ​​is first calculated, with the weights being the corresponding recorded vibration attenuation rates. The current frequency and device parameters are then combined to determine whether the value falls outside a safe range (±15%). Finally, a complete strategy is output, including the target frequency, adjustment rate, and expected vibration attenuation.

[0115] Here's a specific example:

[0116] During the carbon deposit cleaning process on a certain type of marine diesel engine, the system detected a resonant frequency band of 1100-1150 Hz (center frequency 1125 Hz, peak energy 0.16 g² / Hz). The reinforcement learning model's matching module then selected three related records from the historical database: Record A (frequency band 1080-1130 Hz, 83% overlap, 1.2 Hz historical down-regulation, a temperature increase from 180°C to 210°C in 15 minutes (a rate of change of 2°C / min), and a final vibration attenuation of 55%); Record B (1120-1170 Hz, 80% overlap, 0.8 Hz historical down-regulation, a temperature change rate of 3°C / min, and a 60% attenuation); and Record C (1090-1140 Hz, 88% overlap, 1.5 Hz historical down-regulation, a temperature change rate of 1.8°C / min, and a 50% attenuation). The system monitors the current cylinder temperature in real time, measuring a six-minute rise from 182°C to 185°C. The calculated current temperature rate of change is (185-182) / 6 = 0.5°C / min. This rate is compared with the historical records: Record A differs by 1.5°C / min, Record B by 2.5°C / min, and Record C by 1.3°C / min. The adjustment coefficient is calculated based on the temperature difference ratio (current rate / historical rate): Record A is 0.5 / 2 = 0.25, Record B is 0.5 / 3 ≈ 0.17, and Record C is 0.5 / 1.8 ≈ 0.28. Multiplying each historical adjustment by the corresponding coefficient yields the corrected adjustment: Record A is adjusted to 1.2 × 0.25 = 0.3 Hz, Record B is adjusted to 0.8 × 0.17 ≈ 0.14 Hz, and Record C is adjusted to 1.5 × 0.28 ≈ 0.42 Hz. The final weighted average (weighted by the degree of overlap) yields the target adjustment: (0.3 × 83 + 0.14 × 80 + 0.42 × 88) / (83 + 80 + 88) ≈ 0.29 Hz. This generates a frequency correction strategy of "downward adjustment by 0.3 Hz, with an adjustment rate not exceeding 0.1 Hz / s." Post-implementation monitoring shows a 40% reduction in 1125 Hz vibration energy, while maintaining cylinder temperature within a safe range.

[0117] In the examples of this application, precise optimization and adjustment of the pulse frequency are achieved through intelligent historical experience matching and dynamic parameter correction. This method not only retains the effectiveness of historical successful experience, but also adapts to current operating conditions, ensuring that the cleaning process always maintains the optimal working state, effectively improving cleaning efficiency and ensuring equipment safety.

[0118] To further improve the accuracy and adaptability of the pulse frequency correction strategy, in some embodiments, step 304: generating the pulse frequency correction strategy based on the adjusted historical pulse frequency adjustment amount set includes:

[0119] Step 401: Divide the resonant frequency band into multiple sub-frequency bands, and assign a weight value to each sub-frequency band. The weight value is the ratio of the energy of the corresponding sub-frequency band to the total energy of the resonant frequency band.

[0120] In step 401, the sub-band represents a number of small intervals divided by a fixed bandwidth of the resonance frequency band, and the bandwidth is usually 20-50 Hz. The weight value is used to reflect the coefficient of the importance of the vibration energy of each sub-band, ranging from 0 to 1. The specific process of dividing the resonance frequency band into multiple sub-bands is as follows: the system first obtains the start and end frequencies of the resonance frequency band (for example, the resonance frequency band is detected to be 1100-1200 Hz), and then uses a dynamic bandwidth division algorithm to divide it into sub-bands. The division basis includes three core elements: (1) the resonance energy distribution characteristics, by analyzing the second-order derivative of the power spectrum density curve to determine the energy mutation point as the potential segmentation boundary; (2) the natural frequency characteristics of the engine structure, referring to the modal analysis report of the engine model, to avoid dividing the natural frequency point at the edge of the sub-band; (3) the statistical law of historical data, based on the big data clustering results of the optimal segmentation scheme in past successful cases. In specific implementation, the system will first divide the resonant frequency band into equal parts with an initial bandwidth of 20 Hz, and then dynamically adjust the boundaries based on the above three factors: if an interval contains an energy peak (exceeding 30% of the adjacent interval), new segmentation points will be added at 5 Hz on both sides of the peak; if an interval contains a known natural frequency of the engine structure, the sub-band will be divided into ±5 Hz with this frequency as the center.

[0121] In this embodiment, the resonant frequency band is divided into several sub-bands using an equal-bandwidth partitioning method. The integral vibration energy (area under the power spectral density curve) is calculated for each sub-band. The weight of each sub-band is calculated as the energy value of the sub-band divided by the total energy value of the resonant frequency band. A 10% overlap between sub-bands is ensured to avoid edge effects.

[0122] Step 402: For each sub-frequency band, a target historical pulse frequency adjustment value is selected from the adjusted historical pulse frequency adjustment value set, where the target historical pulse frequency adjustment value is the adjusted historical pulse frequency adjustment value corresponding to the sub-frequency band.

[0123] In this embodiment, a mapping relationship between sub-frequency bands and historical records is established. For each sub-frequency band, the three historical records with the smallest difference from its center frequency are selected, and the middle value of their adjustment values ​​is taken as the target adjustment value. If the historical records are insufficient, the adjustment values ​​of adjacent sub-frequency bands are supplemented by linear interpolation.

[0124] Step 403: Calculate the average adjustment value of the corresponding sub-band according to the target historical pulse frequency adjustment value.

[0125] In step 403, the mean adjustment value represents the statistical average of multiple adjustments for the same sub-band. The process for calculating the sub-band adjustment value mean is as follows: the system first obtains all target historical pulse frequency adjustments corresponding to the sub-band. Each adjustment value is then assigned a weight based on the vibration attenuation rate in its original historical records (the higher the vibration attenuation rate, the greater the weight, calculated as: weight = attenuation rate / sum of all recorded attenuation rates). The mean is then calculated using a weighted average algorithm: Each adjustment value is multiplied by its corresponding weight, the sum is then divided by the total weight (calculated as: mean adjustment value = ∑(adjustment value × weight) / ∑weight). During the calculation process, the system first uses boxplot analysis to identify and remove outliers that deviate from the mean by more than 1.5 times the interquartile range to ensure the robustness of the mean. For any remaining valid adjustments, if there are fewer than three, the two closest adjustments are borrowed from adjacent sub-bands to supplement the remaining values. Linear interpolation is used to correct for any deviations in the borrowed values, and the final, verified mean adjustment value is output. For example, a sub-band obtains three effective adjustment values: -1.2 Hz (weight 0.4), -1.5 Hz (weight 0.3), and -1.0 Hz (weight 0.3). The calculated average is (-1.2×0.4-1.5×0.3-1.0×0.3) / (0.4+0.3+0.3)=-1.23 Hz.

[0126] In this embodiment, a weighted average is calculated for multiple target adjustments corresponding to each sub-band, with the weight being the vibration attenuation rate of the original historical record. A graph-based window filtering algorithm is used to remove outliers during the mean calculation to ensure robustness of the result.

[0127] Step 404: multiply the mean of the adjustment amount by the weight value of the corresponding sub-frequency band to obtain a weighted correction amount.

[0128] In step 404 , the weighted correction value represents the frequency correction value after energy weight adjustment.

[0129] In this embodiment, the final contribution value of each sub-band is obtained by multiplying the mean adjustment value of each sub-band by its weight value. At the same time, a temperature compensation factor (current temperature / historical average temperature) is introduced for fine-tuning to make the correction amount more suitable for the current working conditions.

[0130] Step 405: Combine the weighted correction values ​​of all sub-bands into an initial correction value set.

[0131] In step 405 , the initial correction value set includes a complete data set of weighted correction values ​​for all sub-bands.

[0132] In this embodiment, the weighted correction values ​​of each sub-band are arranged from low to high frequency to construct a frequency-correction value correspondence table. The correction values ​​between adjacent sub-bands are smoothed (cubic spline interpolation) to ensure the continuity of frequency adjustment.

[0133] Step 406: Generate a pulse frequency correction strategy based on the initial correction value set, combined with the maximum allowable pulse frequency of the engine and the current carbon deposit level.

[0134] In step 406 , the maximum allowed pulse frequency represents the pulse frequency operating range allowed by the engine hardware, and the current carbon deposit level represents the adjustment amplitude scaling factor set according to the severity of carbon deposits.

[0135] In this embodiment, the engine manual is first consulted to obtain the maximum allowable frequency deviation (e.g., ±15%). A scaling factor (0.8, 1.0, or 1.2) is then selected based on the current carbon deposit level (mild, moderate, or severe). Finally, all weighted corrections are subjected to a double constraint: multiplied by the carbon deposit factor and then truncated to within the allowable frequency range to generate the final correction strategy table.

[0136] Here's a specific example:

[0137] During the carbon deposit cleaning process of a certain type of ship diesel engine, the system detects that the current resonance frequency band is 1080-1180Hz (total energy integral value is 150 units). First, the frequency band is divided into 5 sub-bands: 1080-1100Hz (energy value 18), 1100-1120Hz (energy value 35), 1120-1140Hz (energy value 42), 1140-1160Hz (energy value 32), 1160-1180Hz (energy value 23), and the weights of each sub-band are 0.12 (18 / 150), 0.23 (35 / 150), 0.28 (42 / 150), 0.21 (32 / 150), and 0.15 (23 / 150), respectively. The adjustment values ​​for each sub-band are matched from the adjusted historical pulse frequency adjustment values. For the 1080-1100 Hz band, the three historical adjustment values ​​are -0.8 Hz, -1.0 Hz, and -0.6 Hz, respectively, with an average of -0.8 Hz. The average adjustment values ​​for the 1100-1120 Hz band are -1.2 Hz, -1.5 Hz for the 1120-1140 Hz band, -1.0 Hz for the 1140-1160 Hz band, and -0.7 Hz for the 1160-1180 Hz band. The weighted correction values ​​for each sub-band band are calculated by multiplying the average adjustment value by its weight: -0.8 × 0.12 = -0.096 Hz, -1.2 × 0.23 = -0.276 Hz, -1.5 × 0.28 = -0.42 Hz, -1.0 × 0.21 = -0.21 Hz, and -0.7 × 0.15 = -0.105 Hz. The initial correction values ​​are [-0.096, -0.276, -0.42, -0.21, -0.105] Hz. Considering the current moderate carbon deposit level (adjustment factor 1.0) and the engine's maximum allowable adjustment range of ±1.5 Hz, the resulting pulse frequency correction strategy is a 0.42 Hz reduction in the primary resonance range of 1120-1140 Hz, and proportional reductions of 0.1-0.28 Hz in other regions, keeping the total adjustment range within a safe range.

[0138] In this application's embodiment, a method of frequency band segmentation and energy weighting is used to precisely suppress complex resonant modes. This approach considers the differences in vibration energy across frequency bands while also taking into account engine safety constraints, making frequency adjustment more refined and personalized, effectively improving the safety and effectiveness of the cleaning process.

[0139] To further improve the safety and reliability of the pulse frequency correction strategy, in some embodiments, step 405: generating a pulse frequency correction strategy based on the initial correction amount set, combined with the maximum allowable pulse frequency of the engine and the current carbon deposit level, includes:

[0140] Step 501: According to the current carbon deposition level, a corresponding allowable floating value is searched from a preset floating range mapping table.

[0141] In step 501, the floating range mapping table represents a lookup table storing the corresponding allowable frequency adjustment ranges for different carbon deposition levels. The allowable floating value represents the maximum frequency adjustment range allowed under the current carbon deposition state.

[0142] In this embodiment, the system maintains a three-dimensional lookup table (carbon deposit level x engine model x operating condition type) and matches the current carbon deposit score (e.g., 0-100) to the corresponding allowable floating value. Carbon deposit levels are divided into three levels: mild (±10%), moderate (±15%), and severe (±20%). The specific values ​​are obtained from the technical manual based on the engine model.

[0143] Step 502: Calculate the upper limit value of each weighted correction value in the initial correction value set according to the maximum allowable pulse frequency of the engine.

[0144] In step 502 , the upper limit value represents the maximum allowable frequency deviation determined by engine hardware parameters.

[0145] In this embodiment, the maximum allowable pulse frequency (e.g., ±20% of the rated frequency) is read from the engine control unit, and the upper limit of each weighted correction is calculated as: the maximum allowable engine pulse frequency × a first preset proportional coefficient. The calculation takes frequency-temperature coupling into account, and the upper limit is automatically reduced by 10% when the cylinder temperature exceeds a threshold.

[0146] Step 503: Calculate the lower limit value of each weighted correction amount according to the initial pulse frequency.

[0147] In step 503, the lower limit value represents the minimum frequency adjustment amount to ensure the cleaning effect.

[0148] In the embodiment of the present application, the lower limit is set to initial pulse frequency × second preset proportional coefficient. Dynamic adjustment is also performed based on the current carbon deposit thickness: when the local carbon deposit thickness is greater than 0.2 mm, the lower limit is relaxed to initial frequency × 0.85 to ensure sufficient cleaning intensity.

[0149] Step 504: Based on the allowable floating value, the upper limit value and the lower limit value, constrain the weighted correction amount.

[0150] In step 504 , the constraint processing represents a process of placing triple restrictions on the correction amount.

[0151] In this embodiment, each weighted correction is subjected to constraint verification: first, it is checked to see if it exceeds the allowed floating value, and then it is verified to be within the upper and lower limits. For corrections that exceed the constraints, the nearest neighbor method is used to adjust them to the nearest valid value, and the correction flag is recorded for subsequent analysis.

[0152] Step 505: Arrange all constrained weighted correction amounts in a preset frequency band order to generate a pulse frequency correction strategy.

[0153] In step 505, the frequency band sequence represents a sequence of correction amounts arranged from low to high frequency.

[0154] In this embodiment, the constrained corrections are sorted by sub-band center frequency to generate a frequency-adjustment comparison table. Adjustments between adjacent bands are smoothed (using cubic spline interpolation) to ensure continuity of frequency changes. The final strategy includes the target frequency, adjustment rate, and expected effect evaluation for each band.

[0155] Here's a specific example:

[0156] During a carbon deposit cleaning operation on a certain marine diesel engine, the system generated a final strategy based on an initial set of correction values: [-0.096, -0.276, -0.42, -0.21, -0.105] Hz. The system first consulted a floating range mapping table based on the current carbon deposit score of 75 (moderate), obtaining an allowable floating range of ±1.5 Hz (calculated as: base value ±1.2 Hz × carbon deposit factor 1.25). Engine parameters indicated a maximum allowable pulse frequency of 26.5 Hz (initial 25 Hz + 1.5 Hz), with a lower limit of 22.5 Hz (25 Hz × 0.9). Each weighted correction value was triple-constrained: -0.276 Hz for the 1100-1120 Hz range was verified to meet all conditions; -0.42 Hz for the 1120-1140 Hz range converted to an absolute frequency of 24.58 Hz (25 - 0.42), which was within the allowable range; and correction values ​​for all other frequency bands passed verification. The resulting strategy was generated sequentially by frequency band: 1080-1100Hz was adjusted down by 0.1Hz (rounded down from -0.096Hz), 1100-1120Hz by 0.28Hz, 1120-1140Hz by 0.42Hz, 1140-1160Hz by 0.21Hz, and 1160-1180Hz by 0.11Hz. The adjustment rate was uniformly limited to 0.15Hz / s (calculated based on a maximum adjustment of 0.42Hz and a safety period of 3 seconds). Post-execution monitoring showed a 52% reduction in vibration energy in the primary resonance zone of 1120-1140Hz (0.42Hz adjustment × 123% / Hz attenuation factor). The adjusted operating frequencies of all frequency bands remained within the safe range of 22.58-24.9Hz.

[0157] In this application's embodiment, multi-dimensional safety constraints and a dynamic adjustment mechanism ensure that the frequency correction strategy not only meets engine safety requirements but also effectively improves cleaning performance. This method is particularly suitable for large diesel engines with complex operating conditions, maximizing cleaning efficiency while ensuring equipment safety.

[0158] In order to further improve the matching accuracy between the historical pulse frequency adjustment amount and the current working condition, in some embodiments, step 303: adjusting the corresponding historical pulse frequency adjustment amount according to the multiple comparison results to obtain the adjusted historical pulse frequency adjustment amount includes:

[0159] Step 601: extracting the historical temperature change rate within the same time period as the current cleaning time point from the historical in-cylinder temperature change curve based on multiple comparison results.

[0160] In step 601, the historical temperature change rate represents the temperature change slope at the same time point in the historical cleaning record. The same time period refers to the same time interval from the start of cleaning to the current moment.

[0161] In an embodiment of the present application, the system establishes a time alignment mechanism to extract the historical temperature data segment consistent with the current cleaning time for each associated record (e.g., if the current cleaning is 15 minutes, the data of the first 15 minutes of the historical record will be extracted).

[0162] Step 602: Calculate the absolute value of the difference between the current temperature change rate and the historical temperature change rate.

[0163] In step 602, the absolute value of the difference represents the absolute difference between the current rate and the historical rate.

[0164] In this embodiment, for each associated record, the following calculation is performed: absolute value of difference = |current rate - historical rate|. A dynamic difference ratio is also calculated: absolute value of difference / historical rate, which is used for subsequent coefficient mapping. This calculation uses a sliding window check to eliminate transient fluctuations.

[0165] Step 603: According to the absolute value of the difference, a corresponding correction coefficient is searched from a preset coefficient mapping table.

[0166] In step 603, the coefficient mapping table represents a table storing correspondence between difference ratios and correction coefficients.

[0167] In this embodiment of the present application, a piecewise linear mapping rule is preset: when the difference ratio is ≤ 0.2, the coefficient is 1.0; when 0.2 < difference ratio ≤ 0.5, the coefficient is 0.8; when 0.5 < difference ratio ≤ 1.0, the coefficient is 0.6; when the difference ratio is > 1.0, the coefficient is 0.4. The system queries this table in real time to obtain the precise coefficients and supports dynamic updating of mapping rules.

[0168] Step 604: multiply the historical pulse frequency adjustment amount in each associated record by the corresponding correction coefficient to obtain the adjusted historical pulse frequency adjustment amount.

[0169] In this embodiment, for each record, the following equation is applied: Corrected Adjustment = Original Adjustment × Correction Factor. A minimum valid adjustment (e.g., ±0.3 Hz) is also set. If the result is less than this value, a guaranteed minimum adjustment is used. All correction results are stored in a temporary collection for subsequent strategy generation.

[0170] Here's a specific example:

[0171] During the carbon deposit cleaning process of a certain type of marine diesel engine, the system detected a temperature change rate of 0.5°C / minute (calculated from the cylinder temperature rising from 182°C to 185°C in 6 minutes: (185-182) / 6 = 0.5°C / minute). For three records matched to the historical database: Record A (historical rate of 2°C / minute, original adjustment of -1.2Hz), Record B (historical rate of 3°C / minute, original adjustment of -0.8Hz), and Record C (historical rate of 1.8°C / minute, original adjustment of -1.5Hz), the system first calculated the absolute temperature difference between each record (Record A: |0.5-2| = 1.5°C / minute; Record B: |0.5-3| = 2.5°C / minute; Record C: |0.5-1.8| = 1.3°C / minute). According to the preset coefficient mapping table (temperature difference ≤ 1°C, coefficient 1.0; 1-2°C, coefficient 0.6; 2-3°C, coefficient 0.3; >3°C, coefficient 0.1), the correction coefficients are determined: Record A is 0.3 (1.5 in the 1-2°C range), Record B is 0.3 (2.5 in the 2-3°C range), and Record C is 0.6 (1.3 in the 1-2°C range). Multiplying the original adjustment by the corresponding coefficient yields the corrected adjustment: Record A is adjusted to -1.2 × 0.3 = -0.36 Hz, Record B to -0.8 × 0.3 = -0.24 Hz, and Record C to -1.5 × 0.6 = -0.9 Hz.

[0172] In this application's embodiment, the accuracy of reusing historical empirical data is improved through refined matching and dynamic correction of temperature variation characteristics. This method not only retains the effective adjustment amount of successful cases, but also adapts to the current operating conditions. The resulting frequency correction strategy is both reliable and adaptable, effectively resolving the strategy inaccuracy problem caused by temperature variation in traditional methods.

[0173] To further improve the accuracy and reliability of identifying the resonant frequency band, in some embodiments, step 202 of identifying the resonant frequency band based on the pulse vibration component includes:

[0174] Step 701: Divide the pulse vibration component into multiple analysis frequency bands according to preset frequency intervals.

[0175] In step 701, the analysis frequency band represents the small intervals of equal width into which the spectrum of the vibration signal is divided. The frequency interval represents a fixed bandwidth value set according to the engine characteristics.

[0176] In this embodiment, a dynamic segmentation technique is used to divide the frequency bands: first, the engine operating frequency range is determined (e.g., 500-3000 Hz), then this range is divided equally into 25 Hz fixed bandwidths (configurable), generating 100 analysis frequency bands (e.g., 500-525 Hz, 525-550 Hz, etc.). The segmentation takes into account the modal characteristics of the engine structure to ensure that natural frequencies do not fall on the frequency band boundaries.

[0177] Step 702: For each analysis frequency band, calculate the average energy value of the analysis frequency band within a preset time window.

[0178] In step 702, the time window is used as a sliding time period for energy statistics. The average energy value represents the integrated mean of the power spectrum density in the frequency band.

[0179] In this embodiment, a 10-second sliding time window (with a 5-second overlap) is set. For each analysis frequency band, the following steps are performed: calculate the power spectral density value (obtained by fast Fourier transform) for all frequency points within the band, calculate the area integral, and divide it by the band width to obtain the average energy value per Hz (unit: g² / Hz). The calculation process uses the trapezoidal integration method to improve accuracy.

[0180] Step 703: Compare the average energy values ​​of all analyzed frequency bands and select candidate frequency bands whose energy values ​​exceed a preset energy threshold.

[0181] In step 703, the energy threshold represents a critical energy level for determining the presence of resonance, and the candidate frequency band represents a possible resonance frequency band that has been preliminarily screened.

[0182] In this embodiment of the present application, the energy threshold = overall average energy × 3 + background noise energy (estimated by wavelet denoising). All analyzed frequency bands are sorted, and frequency bands with energy values ​​exceeding the threshold and ranking in the top 20% are selected as candidates. The energy of the candidate frequency band is also required to be at least 1.5 times that of the adjacent frequency bands to avoid false detection.

[0183] Step 704: Merge candidate frequency bands with adjacent frequencies and energy values ​​exceeding a preset energy threshold to form continuous energy peaks.

[0184] In step 704 , the continuous energy peaks represent continuous high energy regions on the frequency spectrum.

[0185] In this embodiment, a connected domain analysis is performed on the candidate frequency bands: if the center frequency difference between the two bands is ≤ 1.5 times the bandwidth (37.5 Hz) and the energy ratio is between 0.7 and 1.3, they are merged into a single energy peak. After merging, the peak energy (maximum value) and bandwidth (start and end frequency difference) are recalculated.

[0186] Step 705: Determine a frequency band in the continuous energy peaks whose duration exceeds a preset duration threshold as a resonant frequency band.

[0187] In step 705 , the duration threshold represents a time standard for determining resonance stability.

[0188] In the present embodiment, continuous energy peaks must persist within at least three consecutive time windows (30 seconds) with an energy fluctuation of ≤20%. For energy peaks that meet these requirements, their center frequency, bandwidth, peak energy, and duration are recorded as the final resonant frequency band output.

[0189] Here's a specific example:

[0190] During the carbon deposit cleaning process of a certain type of marine diesel engine, the system identified resonant frequency bands from the collected pulse vibration components. The system first divided the 800-2000 Hz frequency range into 60 analysis bands (e.g., 800-820 Hz, 820-840 Hz, etc.) at 20 Hz intervals. For each band, the average energy value within the most recent 30-second window was calculated. The energy value for the 1100-1120 Hz band was 0.25 units (calculated by summing the power spectral density values ​​of all frequency points within the band and dividing by the band width of 20 Hz), 0.32 units for the 1120-1140 Hz band, and 0.28 units for the 1140-1160 Hz band. The system preset an energy threshold of 0.2 units (determined based on historical data statistics, taking it as three times the average energy under normal operating conditions) to select the three adjacent bands mentioned above as candidates. These bands were then merged to form a continuous energy peak between 1100 and 1160 Hz (with a maximum peak energy of 0.32 units). The stability of the energy peak was continuously monitored within three consecutive detection cycles (45 seconds in total). The energy values ​​of each cycle were 0.30, 0.31, and 0.29 units, respectively (fluctuation less than 10%), meeting the preset duration threshold (30 seconds) and stability requirements (fluctuation less than 15%). It was finally determined to be a resonant frequency band, and the system recorded its characteristic parameters such as center frequency 1130 Hz, bandwidth 60 Hz, and peak energy 0.32 units for subsequent frequency adjustment strategy generation.

[0191] In the present embodiment, a resonance identification method using multi-stage screening and stability verification effectively distinguishes true resonance from transient interference. This method takes into account both frequency-domain energy characteristics and time-domain stability, improving the accuracy of resonance detection and providing a reliable basis for subsequent frequency optimization. It is particularly suitable for scenarios with complex vibration noise, such as ship engines.

[0192] Figure 2 A schematic diagram of the structure of a pulse frequency adaptive matching system for cleaning carbon deposits in an engine combustion chamber provided in an embodiment of the present application is shown as follows: Figure 2 As shown, the system includes:

[0193] The acquisition module 21 is used to obtain the carbon deposit distribution data, cylinder pressure data and engine vibration signal of the target engine combustion chamber uploaded by the edge computing node.

[0194] The first generating module 22 is configured to dynamically weight and fuse the carbon deposit distribution data and the cylinder pressure data to generate a comprehensive engine status score.

[0195] The search module 23 is configured to search a preset pulse frequency mapping table for an initial pulse frequency of the cleaning device that matches the comprehensive score of the engine status.

[0196] The second generating module 24 is configured to generate a pulse frequency correction strategy based on the engine vibration signal in combination with a pre-trained reinforcement learning model.

[0197] The third generation module 25 is used to dynamically adjust the initial pulse frequency according to the correction strategy, generate a target pulse frequency, and send the target pulse frequency to the edge computing node to control the cleaning equipment to perform pulse cleaning according to the target pulse frequency.

[0198] Figure 2 The pulse frequency adaptive matching system for cleaning carbon deposits in the combustion chamber of an engine can be executed Figure 1 The implementation principles and technical effects of the pulse frequency adaptive matching method for cleaning carbon deposits in an engine combustion chamber, as described in the illustrated embodiment, will not be elaborated upon. The specific manner in which the various modules and units perform their operations in the pulse frequency adaptive matching system for cleaning carbon deposits in an engine combustion chamber, as described in the aforementioned embodiment, has been described in detail in the relevant embodiments of the method and will not be further elaborated upon here.

[0199] In one possible design, Figure 2 The pulse frequency adaptive matching system for cleaning carbon deposits in the combustion chamber of an engine according to the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0200] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0201] The processing component 32 is as follows Figure 1 The embodiment provides a pulse frequency adaptive matching method for cleaning carbon deposits in an engine combustion chamber.

[0202] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0203] The storage component 31 is configured to store various types of data to support operations in the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random-access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0204] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0205] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0206] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0207] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0208] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The illustrated embodiment is a pulse frequency adaptive matching method for cleaning carbon deposits in an engine combustion chamber.

[0209] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0210] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0211] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, or of course, by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiments.

[0212] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A pulse frequency adaptive matching method for cleaning carbon deposits in an engine combustion chamber, characterized in that: Applied to the cloud, including: Obtain carbon deposit distribution data, cylinder pressure data, and engine vibration signals of the target engine combustion chamber uploaded by the edge computing node; Dynamically weighting and fusing the carbon deposit distribution data and the cylinder pressure data to generate a comprehensive engine status score; Searching a preset pulse frequency mapping table for an initial pulse frequency of the cleaning device that matches the comprehensive score of the engine status; Generate a pulse frequency correction strategy based on the engine vibration signal and a pre-trained reinforcement learning model; The initial pulse frequency is dynamically adjusted according to the correction strategy to generate a target pulse frequency, and the target pulse frequency is sent to the edge computing node to control the cleaning equipment to perform pulse cleaning according to the target pulse frequency.

2. The method according to claim 1, characterized in that The pulse frequency correction strategy is generated based on the engine vibration signal and in combination with a pre-trained reinforcement learning model, including: Separate the pulse vibration component of the cleaning equipment and the vibration component of the engine body from the engine vibration signal; identifying a resonant frequency band based on the pulse vibration component; The resonant frequency band is input into a pre-trained reinforcement learning model, and the historical cleaning data and real-time in-cylinder temperature feedback data are combined to generate a pulse frequency correction strategy.

3. The method according to claim 2, characterized in that The resonant frequency band is input into a pre-trained reinforcement learning model, and the historical cleaning data and real-time in-cylinder temperature feedback data are combined to generate a pulse frequency correction strategy, including: The resonant frequency band is input into a pre-trained reinforcement learning model, and a matching module of the reinforcement learning model is used to filter out associated records from historical cleaning data whose overlap with the resonant frequency band exceeds a preset overlap threshold. The associated records include historical pulse frequency adjustment amounts, corresponding historical in-cylinder temperature change curves, and vibration attenuation rates when carbon deposit removal is completed. In the case where there are multiple associated records, the calculation module of the reinforcement learning model calculates the current temperature change rate based on the real-time in-cylinder temperature feedback data, and compares the current temperature change rate with each of the historical in-cylinder temperature change curves; According to the multiple comparison results, the corresponding historical pulse frequency adjustment amount is adjusted to obtain an adjusted historical pulse frequency adjustment amount, and the adjusted historical pulse frequency adjustment amount constitutes an adjusted historical pulse frequency adjustment amount set; A pulse frequency correction strategy is generated according to the adjusted historical pulse frequency adjustment amount set.

4. The method according to claim 3, characterized in that Generating a pulse frequency correction strategy according to the adjusted historical pulse frequency adjustment amount set includes: Dividing the resonant frequency band into a plurality of sub-frequency bands, and assigning a weight value to each sub-frequency band, wherein the weight value is the ratio of the energy of the corresponding sub-frequency band to the total energy of the resonant frequency band; For each sub-frequency band, selecting a target historical pulse frequency adjustment value from the adjusted historical pulse frequency adjustment value set, wherein the target historical pulse frequency adjustment value is the adjusted historical pulse frequency adjustment value corresponding to the sub-frequency band; Calculating the mean of the adjustment amounts of the corresponding sub-bands according to the target historical pulse frequency adjustment amount; Multiplying the mean of the adjustment amount by the weight value of the corresponding sub-frequency band to obtain a weighted correction amount; Combining the weighted correction values ​​of all sub-bands into an initial correction value set; A pulse frequency correction strategy is generated based on the initial correction amount set, combined with the maximum allowable pulse frequency of the engine and the current carbon deposit level.

5. The method according to claim 4, characterized in that The pulse frequency correction strategy is generated based on the initial correction amount set, combined with the maximum allowable pulse frequency of the engine and the current carbon deposition level, including: According to the current carbon deposition level, the corresponding allowable floating value is found from the preset floating range mapping table; Calculating an upper limit value of each weighted correction amount in the initial correction amount set according to the maximum allowable pulse frequency of the engine; Calculating a lower limit value of each weighted correction amount according to the initial pulse frequency; Based on the allowable floating value, the upper limit value and the lower limit value, constraining the weighted correction amount; Arrange all constrained weighted corrections in a preset frequency band order to generate a pulse frequency correction strategy.

6. The method according to claim 3, characterized in that The adjusting the corresponding historical pulse frequency adjustment amount according to the plurality of comparison results to obtain the adjusted historical pulse frequency adjustment amount includes: According to the multiple comparison results, extracting the historical temperature change rate within the same time period as the current cleaning time point from the historical cylinder temperature change curve; Calculating the absolute value of the difference between the current temperature change rate and the historical temperature change rate; According to the absolute value of the difference, searching for a corresponding correction coefficient from a preset coefficient mapping table; The historical pulse frequency adjustment amount in each associated record is multiplied by the corresponding correction coefficient to obtain the adjusted historical pulse frequency adjustment amount.

7. The method according to claim 2, characterized in that The step of identifying a resonant frequency band based on the pulse vibration component includes: Dividing the pulse vibration component into a plurality of analysis frequency bands according to preset frequency intervals; For each analysis frequency band, calculating the average energy value of the analysis frequency band within a preset time window; Compare the average energy values ​​of all analyzed frequency bands and select candidate frequency bands whose energy values ​​exceed the preset energy threshold; Merge candidate frequency bands with adjacent frequencies and energy values ​​exceeding a preset energy threshold to form continuous energy peaks; A frequency band of the continuous energy peaks whose duration exceeds a preset duration threshold is determined as a resonant frequency band.

8. A pulse frequency adaptive matching system for cleaning carbon deposits in the engine combustion chamber, characterized in that: include: The acquisition module is used to obtain the carbon deposit distribution data, cylinder pressure data and engine vibration signal of the target engine combustion chamber uploaded by the edge computing node; A first generating module is configured to dynamically weight and fuse the carbon deposit distribution data and the cylinder pressure data to generate a comprehensive engine status score; A search module, configured to search a preset pulse frequency mapping table for an initial pulse frequency of a cleaning device that matches the comprehensive score of the engine state; A second generation module is used to generate a pulse frequency correction strategy based on the engine vibration signal and a pre-trained reinforcement learning model; The third generation module is used to dynamically adjust the initial pulse frequency according to the correction strategy, generate a target pulse frequency, and send the target pulse frequency to the edge computing node to control the cleaning equipment to perform pulse cleaning according to the target pulse frequency.

9. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a pulse frequency adaptive matching method for cleaning carbon deposits in an engine combustion chamber as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, the pulse frequency adaptive matching method for cleaning carbon deposits in an engine combustion chamber as claimed in any one of claims 1 to 7 is implemented.

Images