A computing method for engine noise analysis

Abstract

The application discloses a calculation method for engine noise analysis, which comprises the following steps: collecting vibration signals and rotating speed pulse signals at key positions of an engine, carrying out direct current removal and anti-aliasing filter processing on the vibration signals, carrying out smoothing processing on the rotating speed signals and carrying out instantaneous compensation on non-stationary sections to form a stabilized signal for angle resampling; determining disturbance intensity based on the instantaneous slope of the stabilized rotating speed curve, adaptively adjusting the distribution density of angle interpolation points to generate an angle domain vibration signal with encoded disturbance information; carrying out multi-scale fast Fourier transform on the angle domain signal by using short-time and long-time analysis windows, adaptively determining order resolution, identifying the amplitude and duration of transient abnormal orders from the short window order spectrum, and extracting abnormal order frequency drift trends from the long window order spectrum; and fusing the short window and long window order spectra to generate a two-dimensional order spectrum, thereby realizing multi-scale order feature analysis and visualization of the engine noise signal.

Description

Technical Field

[0001] This invention relates to the field of engine fault monitoring technology, specifically a calculation method for engine noise analysis. Background Technology

[0002] Engines operating at sea are often in a continuous state of operation. Due to the confined space and the difficulty of maintenance during shutdown, any sudden mechanical failure can lead to operational interruption or even safety accidents. To ensure the long-term reliable operation of engines, health monitoring is usually carried out through vibration and noise analysis to identify early failure characteristics such as shaft eccentricity, gear meshing wear, and abnormal combustion.

[0003] In existing technologies, order analysis methods are often used to convert vibration signals from the frequency domain to an order domain related to rotational speed, in order to eliminate the influence of rotational speed fluctuations on the spectrum. The key step in implementing order analysis is angle domain resampling. Currently, the equal-angle interpolation resampling method is widely used. This method is based on the assumption that the rotational speed changes linearly within the interval between adjacent rotational speed pulses, and performs uniform interpolation at a constant angle interval to reconstruct the angle domain waveform.

[0004] However, under offshore operating conditions, engine speed often exhibits non-stationary changes due to factors such as wave loads and sudden load changes, resulting in significant nonlinear micro-perturbations. Traditional methods cannot keep up with these instantaneous speed fluctuations. In the segment where speed changes rapidly, the interpolation points are too sparse, leading to angle distortion. In the high-speed region, sparse interpolation makes it difficult to meet the sampling conditions for higher orders, causing order drift and aliasing. Weak anomalous features caused by speed perturbations, such as the sidebands of early gear wear or the intermittent order shifts of combustion anomalies, are easily overwhelmed by the above interpolation errors and are difficult to accurately identify in the order domain.

[0005] It is evident that traditional equal-angle interpolation resampling methods cannot simultaneously meet the requirements of transient capture and high-order resolution, thus limiting the effective extraction and reliable early warning of weak early engine fault features. Summary of the Invention

[0006] In view of the above-mentioned shortcomings mentioned in the background art, the purpose of this invention is to provide a calculation method, computer equipment, storage medium and computer program product for engine noise analysis.

[0007] A first aspect of the present invention provides a calculation method for engine noise analysis, the method comprising the following steps: S1 collects vibration signals and speed pulse signals from multiple key parts of the engine, and preprocesses them to obtain preprocessed vibration signals and stabilized speed curves; S2, the disturbance intensity is determined based on the instantaneous slope of the stabilized rotational speed curve, and the distribution density of the equal angle interpolation points is adaptively adjusted according to the disturbance intensity to generate an angular domain vibration signal encoded with disturbance information; S3. Based on the rotational speed change amplitude and signal energy, the order resolution of the short-time analysis window and the long-time analysis window are adaptively determined. The short-time analysis window and the long-time analysis window are applied to the angular domain vibration signal to perform fast Fourier transform. The amplitude and duration of the transient abnormal order are identified from the obtained short-window order spectrum, and the frequency drift trend of the abnormal order is extracted from the obtained long-window order spectrum. S4 merges the short-window order spectrum and the long-window order spectrum to generate a two-dimensional order plot with time and order as coordinate axes. The transient anomaly amplitude, duration and frequency drift trend are marked at the corresponding order positions in the plot.

[0008] A second aspect of the present invention provides a computer device including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the method as described in any of the preceding claims.

[0009] A third aspect of the present invention provides a storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the method described in any of the preceding claims.

[0010] A fourth aspect of the present invention provides a program product including a computer program that, when executed by a processor, implements the steps of the method described in any of the preceding claims.

[0011] Compared with the prior art, the present invention has at least the following beneficial technical effects: The method of this invention enables adaptive order analysis of vibration signals under non-stationary engine speed conditions, achieving dynamic adjustment of angle resampling and multi-scale order feature extraction. This effectively identifies transient abnormal orders and frequency drift trends, improves the resolution of high-order components and the identifiability of weak abnormal features. Simultaneously, it fuses short-window and long-window order spectra to generate a visualized two-dimensional order diagram, providing clear abnormal order annotations. This facilitates continuous monitoring and early fault analysis under multiple operating conditions, thereby enhancing the accuracy and reliability of engine operating status assessment. Attached Figure Description

[0012] Figure 1 This is a flowchart illustrating a calculation method for engine noise analysis disclosed in an embodiment of the present invention. Figure 2 This is a schematic diagram of a two-dimensional order diagram generated according to an embodiment of the present invention. Detailed Implementation

[0013] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0014] This invention provides a calculation method for engine noise analysis. In offshore operation scenarios, it can be used to accurately identify early fault characteristics such as abnormal engine combustion, shaft misalignment, and gear meshing wear, thereby realizing the transformation from passive repair to proactive prevention.

[0015] Reference Figure 1 As shown, the method of the present invention includes the following steps: S1 collects vibration signals and speed pulse signals from multiple key parts of the engine, and preprocesses them to obtain preprocessed vibration signals and stabilized speed curves; Vibration and speed signals are the most direct information reflecting the mechanical health of an engine. Vibration signals include characteristic frequency components caused by faults such as shaft misalignment, gear meshing wear, and combustion abnormalities, while speed signals serve as the fundamental reference for order analysis. Therefore, sensors are placed in key parts of the engine to collect signals. These key parts include, but are not limited to, the shaft system, gearbox, and combustion chamber. Shaft vibration can reflect rotor imbalance or misalignment faults, gearbox vibration can reflect tooth wear or broken teeth, and vibration near the combustion chamber can reflect combustion abnormalities.

[0016] Vibration signals are acquired through an accelerometer, while rotational speed pulse signals are obtained through a photoelectric sensor or magnetoelectric sensor installed near the shaft. The output pulse sequence can be converted into an instantaneous rotational speed curve.

[0017] It should be understood that during use, accelerometers may introduce DC bias due to power supply ripple or conditioning circuitry, causing the overall signal to deviate from zero and affecting the accuracy of subsequent angle domain waveforms. Furthermore, when the data acquisition system operates at a fixed sampling rate, the signal may contain frequency components higher than the effective analysis frequency band. If resampling is performed directly, these high-frequency components will alias into the low-frequency band due to the Nyquist condition not being met, resulting in spurious spectral lines. Simultaneously, the speed pulse signal may be affected by encoder jitter, unstable pulse triggering, or electromagnetic interference during measurement, causing local glitches or abnormal jumps. Direct use of these signals can lead to errors in interpolation timing calculations. Especially when the engine experiences rapid acceleration or deceleration due to wave loads or sudden load changes, the speed change within adjacent pulse intervals is not linear, and simple smoothing may lag behind the actual instantaneous speed change. Therefore, the acquired raw signals cannot be directly used for subsequent order analysis and require necessary preprocessing, as follows: As an example, preprocessing is performed to obtain the preprocessed vibration signal and the stabilized rotational speed curve, including: The vibration signal is subjected to DC removal and anti-aliasing filtering to obtain a preprocessed vibration signal; and the rotational speed pulse signal is smoothed and instantaneous rotational speed compensation is performed on non-stationary segments to obtain a stable rotational speed curve.

[0018] For vibration signals, DC removal is performed by calculating and subtracting the signal mean to eliminate constant offset components introduced by sensor bias voltage or acquisition circuitry, bringing the signal mean to zero and ensuring the symmetry of the subsequent angular domain waveform. Anti-aliasing filtering is performed before or after analog-to-digital conversion, using a low-pass filter to remove frequency components above the upper limit of the analysis bandwidth. The cutoff frequency of this low-pass filter is set based on the highest order to be analyzed and the upper frequency limit corresponding to the engine's highest speed, thus preventing frequency aliasing due to insufficient equivalent sampling rate during subsequent angular domain resampling. After these processes, a preprocessed vibration signal is obtained.

[0019] For speed pulse signals, smoothing can be achieved using methods such as moving average filtering or median filtering to suppress local glitches and spikes caused by pulse trigger jitter or encoder manufacturing tolerances, thus obtaining a continuous speed curve. However, conventional smoothing filtering may produce a response lag when the speed changes drastically, making the smoothed speed curve unable to accurately track the actual instantaneous speed change. Therefore, instantaneous speed compensation is needed for non-stationary segments. That is, in segments where the rate of change of speed exceeds a preset threshold, the instantaneous speed value is directly calculated using the original pulse interval within a shorter time window, replacing the lag result of the smoothed value, so that the speed curve in this segment more closely follows the instantaneous fluctuations of the actual speed. Through the combined processing of smoothing and instantaneous compensation, a stable speed curve is obtained.

[0020] S2, the disturbance intensity is determined based on the instantaneous slope of the stabilized rotational speed curve, and the distribution density of the equal angle interpolation points is adaptively adjusted according to the disturbance intensity to generate an angular domain vibration signal encoded with disturbance information; The traditional equal-angle interpolation resampling method is the most widely used angle domain transformation technique in order analysis. Its basic process is as follows: fit the rotational speed curve using the rotational speed pulse signal, determine the sampling time corresponding to the equal-angle increment, and then reconstruct the angle domain waveform through an interpolation algorithm. This method relies on the assumption that the rotational speed change within adjacent pulse intervals is linear or approximately linear, and the interpolation points can be uniformly distributed along this linear segment.

[0021] However, under offshore operating conditions, engine speed often exhibits non-stationary changes due to factors such as quasi-periodic fluctuations in wave loads, sudden changes in operating loads, and frequent adjustments to control commands. The instantaneous speed curve contains a rich and non-stationary micro-disturbance structure. These micro-disturbances are not random measurement noise, but rather a physical reflection of the engine's actual operating state. If these micro-disturbances are ignored and uniform interpolation is performed across the entire speed range at constant angular intervals, the interpolation points become too sparse in the rapid speed change range, leading to angular distortion. In the high-speed range, sparse interpolation fails to meet the sampling conditions for higher orders, causing order drift and aliasing. More importantly, weak anomalous features caused by speed disturbances (such as low-amplitude sidebands generated by early gear meshing wear and intermittent order shifts caused by combustion anomalies) are easily obscured or misjudged by the aforementioned interpolation errors, making accurate identification in the order domain difficult.

[0022] Therefore, this step first calculates the instantaneous slope of the stabilized speed curve. It should be understood that this instantaneous slope is the derivative of speed with respect to time, directly reflecting the rate and direction of speed change at each moment, and is a fundamental quantity for capturing the characteristics of micro-perturbations. Based on this, indicators that can quantify the intensity of micro-perturbations are extracted from the time-varying characteristics of the instantaneous slope, serving as the basis for subsequent adaptive adjustment of the interpolation density.

[0023] As an example, determining the disturbance intensity based on the instantaneous slope of the stabilized rotational speed curve includes: The instantaneous speed slope of each sampling point is normalized to obtain a normalized slope feature value; the deviation of the normalized slope feature value from the slope feature benchmark value of the same engine under the same historical operating conditions is calculated; the deviation is processed by a weighting function or smoothing filter to generate the disturbance intensity at the corresponding time point.

[0024] Understandably, simply using the absolute value of the instantaneous slope to measure the intensity of a disturbance is insufficient, because the normal speed fluctuation amplitude varies under different operating conditions. For example, under low speed and high load conditions, the slope fluctuation amplitude caused by normal combustion cycle changes may be relatively large. If the absolute value of the slope is used directly as the disturbance intensity, normal fluctuations may be misjudged as abnormal disturbances.

[0025] Therefore, this step first normalizes the instantaneous slope to eliminate the influence of differences in operating conditions, obtaining a normalized slope characteristic value. The historical slope characteristic benchmark value under the same operating conditions can be extracted from the engine's historical operating database; this benchmark value represents the statistical expectation of normal speed fluctuations under that operating condition. The current normalized slope characteristic value is compared with this benchmark value to calculate the deviation. The magnitude of the deviation reflects the degree of abnormality of the current speed perturbation relative to the normal state.

[0026] The deviation is processed along the time direction by weighting function or smoothing filter. Weighting function can give higher weight to recent deviations to enhance response speed, while smoothing filter can filter out isolated noise and enhance the response of continuous disturbance segments, and finally generate the disturbance intensity value corresponding to each time point.

[0027] As an example, adaptively adjusting the distribution density of equal-angle interpolation points according to the disturbance intensity to generate an angular domain vibration signal encoded with disturbance information includes: S21, determine the angle interpolation interval corresponding to each time period on the stabilized speed curve according to the disturbance intensity, and perform non-uniform angle segmentation on the stabilized speed curve accordingly to obtain a dense angle interpolation node sequence and a sparse angle interpolation node sequence. Let the stabilized speed curve be Its instantaneous slope is The disturbance intensity values ​​at each time point have been obtained through step S2. This step will determine the disturbance intensity. Transformed into the density distribution of interpolation points in the angle domain: Specifically, let the standard equal-angle interpolation interval be... Define the angle interpolation interval adjustment coefficient. Let the actual angle interpolation interval be: ; in, It is a monotonically increasing function that satisfies , and when hour ,when When it increases It increases accordingly. As an example, It can be designed as: ; in, This is the scaling factor. This represents the power exponent. Through the above mapping relationship, the intensity of the disturbance in each time period can be determined. Determine the corresponding angle interpolation interval .

[0028] Next, Divided into continuous small time periods on the time axis The disturbance intensity value corresponding to each time period The angle interpolation interval is: ; The angle increment corresponding to this period is: ; The number of interpolation nodes required during this period is: ; Let the first The angle value corresponding to each angle interpolation node is... The time period is Then the angular interval between adjacent nodes satisfies: ; After traversing all time periods, the sequence of angle interpolation nodes generated in each time period is concatenated in chronological order. Let the first... The node sequence corresponding to each time period is , will By concatenating the node sequences of each time period, a complete angle interpolation node sequence can be obtained. ,in This represents the total number of interpolation nodes.

[0029] Based on this, the angle interpolation interval corresponding to the time period to which each node belongs is used. The node sequence is divided into two subsequences. Specifically, an angle interpolation interval threshold is set. When node Time period The fact that the node was generated during a period of high disturbance intensity and dense angle interpolation indicates that it was classified into a dense angle interpolation node sequence. ;when The fact that the node was generated during a period of low disturbance intensity indicates that the angle interpolation is sparse, thus classifying it into the sparse angle interpolation node sequence. .

[0030] S22, the preprocessed vibration signal is segmented and dynamically interpolated using the dense angle interpolation node sequence and the sparse angle interpolation node sequence to generate the angular domain vibration signal.

[0031] Let the preprocessed vibration signal be Dense angle interpolation node sequence and sparse angle interpolation node sequence The nodes are merged according to their corresponding sampling times to restore a complete sequence of angle interpolation nodes. The complete sequence has dense nodes in segments with high perturbation intensity and sparse nodes in segments with low perturbation intensity.

[0032] Each node Corresponding theoretical timestamp This is obtained by solving the following equation: ; In actual calculations, For discrete sampling sequences, the above integration is performed numerically using trapezoidal integrals. Starting from the initial time... Begin, along The time series is accumulated point by point with the angle increment, and when the accumulated angle first reaches or exceeds At that time, accurate values ​​are determined through local interpolation. .

[0033] Obtain the timestamp sequence corresponding to all nodes. Then, the preprocessed vibration signal Interpolation is performed to obtain the angular domain vibration signal. Values ​​at each node: ; Interpolation methods can include linear interpolation, cubic spline interpolation, or band-limited interpolation. Due to the complete angle interpolation node sequence... The nodes are dense in sections with high disturbance intensity and sparse in sections with low disturbance intensity, thus generating the final angular domain vibration signal. exist Smaller segments have higher local sampling rates, corresponding to periods of high disturbance intensity; in Larger sections have lower local sampling rates, corresponding to periods of low disturbance intensity. Therefore, disturbance intensity information is encoded in the angular domain vibration signal in the form of an interpolation point density distribution.

[0034] S3. Based on the rotational speed change amplitude and signal energy, the order resolution of the short-time analysis window and the long-time analysis window are adaptively determined. The short-time analysis window and the long-time analysis window are applied to the angular domain vibration signal to perform fast Fourier transform. The amplitude and duration of the transient abnormal order are identified from the obtained short-window order spectrum, and the frequency drift trend of the abnormal order is extracted from the obtained long-window order spectrum. Single-time-window or fixed-resolution FFT analysis cannot simultaneously capture short-term transient features and long-term trend changes. Specifically, short windows, due to their small time span and high temporal resolution, can closely follow instantaneous changes in the signal, which is beneficial for identifying sudden transient anomalies; however, their frequency resolution is low, making it difficult to separate dense order components. Long windows, due to accumulating more rotation periods, have high order resolution and can accurately characterize subtle order shifts; however, their time averaging effect can obscure short-term transient anomalies, causing features to be submerged.

[0035] Furthermore, the difference between the speed change range and the signal energy will significantly affect the rationality of the analysis window selection. Specifically, in the range of rapid speed change, the risk of dynamic aliasing between false sidebands and real fault characteristics is high; when the signal energy is weak, the fault characteristics may be masked by the noise floor.

[0036] To address the aforementioned technical problems, this invention employs an adaptive dual-window strategy, dynamically determining the order resolution of the two analysis windows based on real-time operating conditions, which are used for the two differentiated objectives of transient identification and long-term trend extraction, respectively.

[0037] As an example, the order resolution of the short-time analysis window and the long-time analysis window is adaptively determined based on the speed change amplitude and signal energy, including: S31, calculate the rotational speed change amplitude of the stabilized rotational speed curve during the current analysis period, and the signal energy of the angular domain vibration signal during the corresponding period; Specifically, let the current analysis period be... The stabilized speed curve during this period is Rotational speed variation It can be measured by the difference between the maximum and minimum rotational speeds during that period: ; Alternatively, the standard deviation of rotational speed can be used to characterize the magnitude of change, providing a more robust reflection of the overall level of fluctuation. ; in, This represents the average rotational speed during that period.

[0038] Simultaneously, calculate the angular domain vibration signal. Signal energy within the corresponding time period Let the angle interval corresponding to this time period be . The signal energy can be calculated using the following formula: ; In actual calculations, since the angular domain signal is a discrete sequence, the above integral can be approximated as the sum of squares of discrete sampling points.

[0039] S32, when the rotational speed change exceeds a preset threshold or the signal energy is lower than a preset threshold, shorten the length of the short-time analysis window and the long-time analysis window; otherwise, increase the length of the short-time analysis window and the long-time analysis window.

[0040] It is understandable that the speed variation range It can directly reflect the degree of non-stability of the current operating condition. When Exceeding the preset speed change threshold When the engine is in a rapidly changing operating condition, the risk of dynamic aliasing between the spurious sideband components generated by the non-uniform phase error caused by minor speed disturbances and the actual fault characteristics increases significantly. In this case, shortening the short window length can improve the time-domain resolution, allowing the analysis window to closely follow instantaneous changes and lock the instantaneous position of the spurious sidebands in the time domain, preventing them from being obscured by the time averaging effect of the long window. At the same time, shortening the long window length is to avoid including data from non-stationary periods in the long-term trend calculation and to prevent disturbance periods from contaminating the trend extraction.

[0041] When signal energy Below the preset energy threshold This means that the overall energy of the vibration signal is relatively weak during this period, and subtle fault characteristics may have been masked by the noise floor. Shortening the short window length can reduce the amount of data within the window, reduce the cumulative effect of noise, and improve the local signal-to-noise ratio; similarly, shortening the long window length can prevent data from low signal-to-noise ratio periods from being mixed into long-term trend analysis.

[0042] Conversely, when and This indicates that the current rotational speed is stable and the signal energy is sufficient. In this case, increasing the length of both windows allows the short window to be appropriately widened to balance computational efficiency, while the long window can accumulate more rotational cycle data, achieving higher order resolution. This is beneficial for multi-cycle coherent accumulation of real fault characteristics and for extracting accurate long-term frequency drift trends.

[0043] The above window length adjustment strategy can be expressed as: Let the length of the short window be... The length of the long window is Then we have: ; ; in, , These are the reference lengths for short windows and long windows, respectively. , To adjust the step size. It should be noted that regardless of whether the step size is shortened or increased, the short window length... Always less than the length of the long window This ensures that the two functions have different time resolution and order resolution.

[0044] After adaptively determining the lengths of the two analysis windows, short-time and long-time analysis windows are applied to the angular domain vibration signal and a fast Fourier transform is performed.

[0045] As an example, applying short-time and long-time analysis windows to the angular domain vibration signal to perform a fast Fourier transform includes: S33, based on the interpolation point density distribution of the angular domain vibration signal, the short-time analysis window is slid with a first sliding step size during periods of dense interpolation points and with a second sliding step size during periods of sparse interpolation points. A fast Fourier transform is performed on the data segment within each window to obtain the short-window order spectrum; wherein, the first sliding step size is smaller than the second sliding step size. Specifically, this step fully utilizes the disturbance information encoded in the angular domain vibration signal. Let the angular domain vibration signal... The sequence of interpolation points on the angle axis is as follows The angular interval between adjacent interpolation points is This interval reflects the perturbation intensity of the code; that is, the smaller the interval, the greater the perturbation intensity of the corresponding segment.

[0046] The sliding step size of the short window adaptively changes according to the interpolation point density. Let the window length of the short window be... (Measured in angles), the sliding step size of the short window on the angle axis is... During periods of high interpolation density (i.e., (smaller segments), short windows with the first sliding step size Sliding; during the sparse period of the interpolation point (i.e. (larger segments), short windows with the second sliding step Slide, and satisfy: ; The smaller initial sliding step size indicates a higher overlap between adjacent analysis windows and finer temporal coverage. It should be understood that areas with dense interpolation points correspond to periods of high disturbance intensity, where spurious sideband components caused by non-uniform phase errors may move rapidly. A smaller sliding step size allows for more precise tracking of the instantaneous positional evolution of these spurious sidebands in the time domain, thus enabling their locking and identification. Conversely, areas with sparse interpolation points correspond to stable periods of low disturbance intensity, where spurious sidebands naturally fade, and the true fault characteristics are relatively stable. A larger sliding step size is used to reduce redundant computation.

[0047] For each sliding position, extract the data segment within the window. ( Perform a Fast Fourier Transform on this data segment to obtain the short-window order spectrum at that moment. Let the number of data points within the window be . After FFT, the orders are obtained. Corresponding amplitude ,in This is the time center of the window. After performing the above operation on all sliding positions, the short window order spectrum as a function of time and order is obtained. It should be understood that this spectrum has high temporal resolution and is suitable for identifying sudden increases in amplitude and duration of transient anomalies. That is, transient anomalies are usually characterized by a sudden increase and short-term maintenance of a certain order amplitude within a short period of time, which is related to combustion anomalies or instantaneous impacts.

[0048] S34, apply the long-time analysis window to all the angular domain vibration signals in the current analysis period and perform a fast Fourier transform to obtain the long-window order spectrum.

[0049] Unlike short windows that slide in segments, long windows typically cover the entire current analysis period. The corresponding total or most of the corner domain data. Let the length of the long window be... (Measured in angles), and Perform an FFT on all data points within the window to obtain the order spectrum of the long window. .

[0050] It should be understood that the order resolution of the long-window order spectrum is lower due to the accumulation of more rotation periods. Inversely proportional to the length of the window, that is Much smaller than the order resolution of a short window Therefore, the long-window order spectrum can accurately characterize the frequency drift trend of anomalous orders over long time scales. During stable periods with low disturbance intensity, spurious sidebands naturally fade away, and the features in the long-window order spectrum are more reliable, accurately reflecting the slow evolution of real faults, such as the slight shift in the center frequency of the meshing order caused by gear wear over time.

[0051] Based on this, from the short window order spectrum The system identifies the amplitude and duration of transient anomalies, specifically by setting an amplitude threshold. When a certain order At any moment amplitude When a transient anomaly occurs at that time, its peak amplitude is recorded. The duration is obtained by determining the start and end times when the threshold is exceeded. And, from the order spectrum of long windows Extracting the frequency drift trend of anomalous orders, specifically: tracking specific orders by comparing the order spectrum of long windows in different analysis periods. The trajectory of the center frequency over time The drift direction and rate are obtained as the frequency drift trend of the abnormal order, providing a basis for judging the fault evolution trend.

[0052] S4 merges the short-window order spectrum and the long-window order spectrum to generate a two-dimensional order plot with time and order as coordinate axes. The transient anomaly amplitude, duration and frequency drift trend are marked at the corresponding order positions in the plot.

[0053] Short Window Order Spectrum And the order spectrum of long windows The engine's order characteristics were characterized from both transient and long-term dimensions. Short-window order spectrum. It is time and order The two-dimensional function has high time resolution and can reflect the instantaneous energy changes of each order component within a short time window, making it suitable for capturing sudden amplitude increases caused by combustion anomalies or instantaneous impacts; long-window order spectrum It is a class Although the univariate function has a lower time resolution, it accumulates rotational cycle data over a longer period, resulting in a finer order resolution and accurately reflecting the slow evolution of fault characteristics over time. The two types of spectra contain complementary information; observing either one alone cannot fully present the complete state of the engine on both transient and long-term time scales.

[0054] Therefore, this step merges the short-window order spectrum and the long-window order spectrum into a unified two-dimensional order graph for visualization.

[0055] Specifically, the two-dimensional order plot uses time as the horizontal axis and order as the vertical axis. During its construction, the entire analysis time interval is divided into continuous short segments. ( The duration of each segment can be set to be similar to or slightly longer than the short window sliding step. For each time segment, the short window order spectrum and the long window order spectrum corresponding to that segment are fused to generate the fused order amplitude for that time segment.

[0056] Various strategies can be employed for fusion. As an example, during periods where transient characteristics need to be highlighted, the amplitude of the short-window order spectrum can be prioritized, ensuring that sudden energy jumps in order are clearly presented in the graph. During periods of stable rotational speed and low disturbance intensity, the amplitude of the long-window order spectrum can be prioritized, leveraging its high signal-to-noise ratio and accurate order localization capability to guarantee the accuracy and continuity of the order trajectories in the graph. The fused [element / method / process]... The first time segment, the first Amplitude of the order It can be represented as: ; in, To incorporate the weighting coefficients, the perturbation intensity corresponding to the segment can be used. Or the range of speed change Adaptive determination. When the disturbance intensity is large and the speed changes drastically. A larger value is chosen to make the short-window results dominant, thus highlighting transient changes; when the disturbance intensity is small and the rotation speed is stable, Choose a smaller value to allow the long window result to dominate, so as to give full play to its stable order resolution advantage.

[0057] The fusion order magnitude of all time segments Arranged by time and order, forming a The matrix, where This is the highest order of analysis. For example... Figure 2 As shown, the matrix is ​​mapped to a color map, where the row index of each element in the matrix corresponds to its order. Column index corresponding to time Element value The color gradient is converted into color depth or hue through a color mapping function. A higher amplitude value results in a brighter color or a warmer hue; a lower amplitude value results in a darker color or a cooler hue. Therefore, users can intuitively identify the continuous growth or sudden leaps in abnormal orders, as well as the overall change pattern of energy at each order over time, through the spatial distribution of the color gradient.

[0058] Based on the generated two-dimensional order map, the three types of features identified and extracted are further labeled at corresponding positions in the map.

[0059] For transient anomalous orders identified from the short-window order spectrum, the time-order position corresponding to that order in the two-dimensional order plot is marked with a symbol (such as a circle, asterisk, or highlighted border), and its peak amplitude is noted in text next to the mark or in the legend. and duration This annotation allows users to quickly pinpoint the timing, order of impact, and severity of transient anomalies.

[0060] The frequency drift trend extracted from the long-window order spectrum is marked in the two-dimensional order plot using arrows, trend lines, or superimposed gradient curves. Specifically, the center frequency position of the same order in the long-window order spectrum for different analysis periods can be displayed. Connecting these points in a smooth curve on the graph, with the arrow pointing in the direction of drift, helps users intuitively determine whether the center frequency of this abnormal order has systematically shifted over time. This shift is often associated with a gradual increase in gear meshing clearance or a slow degradation of shaft alignment, and is an important basis for early fault warning and trend prediction.

[0061] Understandably, through the aforementioned fusion and annotation processes, the two-dimensional order diagram unifies the transient and long-term trend information, originally scattered across two spectra, into a single view, making the time-domain manifestation and frequency-domain evolution of fault characteristics readily apparent. Furthermore, this diagram can be queried and retrieved based on time, operating condition, and measurement point location, and historical order diagrams under the same operating condition can be overlaid for comparative analysis. Through intuitive comparison of color differences and order energy change trends, managers can effectively track the evolution trends of engine combustion uniformity, shaft alignment, and gear meshing wear, issuing early warnings and scheduling planned maintenance before faults worsen, thereby achieving continuous tracking of engine health and proactive prevention of early faults.

[0062] This invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in any of the preceding claims.

[0063] Specifically, the computer device can be an industrial control computer, embedded processing unit, or remote server deployed at the engine monitoring site. The memory stores the acquired vibration signal data, speed pulse signal data, preprocessed vibration signals and stabilized speed curves, angle interpolation node sequences, angular domain vibration signals, short-window order spectra, long-window order spectra, and generated two-dimensional order maps, as well as intermediate data and processing results. The processor executes the computer program to sequentially complete all methodological steps, including signal preprocessing, disturbance intensity determination based on instantaneous slope, adaptive angular domain resampling, multi-scale window FFT analysis, order feature identification and extraction, and two-dimensional order map fusion and annotation, thereby achieving multi-scale order feature analysis and visualization of the engine noise signal.

[0064] This invention also provides a storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the method described in any of the preceding claims.

[0065] The storage medium can be a non-volatile storage medium, such as a read-only memory (ROM), flash memory, a hard disk drive (HDD), or a solid-state drive (SSD). The computer program stored in the storage medium contains program instructions for performing the aforementioned steps. By loading the computer program in the storage medium onto a processor for execution, the method of the present invention can be implemented on any compatible computing device.

[0066] This invention also provides a program product, including a computer program that, when executed by a processor, implements the steps of the method described in any of the preceding claims.

[0067] This program product can be provided in the form of a software installer, executable file, or code repository, and supports deployment and operation in various operating system environments. Through this program product, users can easily deploy the method of this invention on different hardware platforms to achieve multi-scale order analysis of engine operating status and early fault warning.

[0068] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A calculation method for engine noise analysis, characterized in that, Includes the following steps: S1 collects vibration signals and speed pulse signals from multiple key parts of the engine, and preprocesses them to obtain preprocessed vibration signals and stabilized speed curves; S2, determine the disturbance intensity based on the instantaneous slope of the stabilized rotational speed curve, and adaptively adjust the distribution density of equal-angle interpolation points according to the disturbance intensity to generate an angular domain vibration signal encoded with disturbance information; S3. Based on the rotational speed change amplitude and signal energy, adaptively determine the order resolution of the short-time analysis window and the long-time analysis window. Apply the short-time analysis window and the long-time analysis window to the angular domain vibration signal to perform fast Fourier transform. Identify the amplitude and duration of the transient abnormal order from the obtained short-window order spectrum. Extract the frequency drift trend of the abnormal order from the obtained long-window order spectrum. S4, merge the short-window order spectrum and the long-window order spectrum to generate a two-dimensional order plot with time and order as coordinate axes, and mark the transient anomaly amplitude, duration and frequency drift trend at the corresponding order position in the plot; Determining the disturbance intensity based on the instantaneous slope of the stabilized speed curve includes: normalizing the instantaneous speed slope of each sampling point to obtain a normalized slope feature value; calculating the deviation of the normalized slope feature value from the slope feature benchmark value of the same engine under the same historical operating conditions; and processing the deviation through a weighting function or smoothing filter to generate the disturbance intensity at the corresponding time point. The method adaptively adjusts the distribution density of equal-angle interpolation points according to the disturbance intensity to generate an angular domain vibration signal encoded with disturbance information, including: S21, determining the angle interpolation interval corresponding to each time period on the stabilized rotational speed curve according to the disturbance intensity, and performing non-uniform angle segmentation on the stabilized rotational speed curve accordingly to obtain a dense angle interpolation node sequence and a sparse angle interpolation node sequence; S22, using the dense angle interpolation node sequence and the sparse angle interpolation node sequence to perform segmented dynamic interpolation on the preprocessed vibration signal to generate the angular domain vibration signal.

2. The calculation method for engine noise analysis according to claim 1, characterized in that: The preprocessing process to obtain a preprocessed vibration signal and a stabilized speed curve includes: performing DC removal and anti-aliasing filtering on the vibration signal to obtain a preprocessed vibration signal; and performing smoothing processing on the speed pulse signal and instantaneous speed compensation on non-stationary segments to obtain a stabilized speed curve.

3. The calculation method for engine noise analysis according to claim 1, characterized in that: The order resolution of the short-time analysis window and the long-time analysis window is adaptively determined based on the rotational speed change amplitude and signal energy, including: S31, calculating the rotational speed change amplitude of the stabilized rotational speed curve in the current analysis period, and the signal energy of the angular domain vibration signal in the corresponding period; S32, when the rotational speed change amplitude exceeds a preset threshold or the signal energy is lower than a preset threshold, shortening the length of the short-time analysis window and the long-time analysis window; otherwise, increasing the length of the short-time analysis window and the long-time analysis window.

4. The calculation method for engine noise analysis according to claim 3, characterized in that: Applying short-time analysis windows and long-time analysis windows to the angular domain vibration signal for Fast Fourier Transform (FFT) includes: S33, based on the interpolation point density distribution of the angular domain vibration signal, sliding the short-time analysis window with a first sliding step size during periods of dense interpolation points and with a second sliding step size during periods of sparse interpolation points, performing FFT on the data segment within each window to obtain the short-window order spectrum; wherein the first sliding step size is smaller than the second sliding step size; S34, applying the long-time analysis window to all the angular domain vibration signals of the current analysis period and performing FFT to obtain the long-window order spectrum.

5. The calculation method for engine noise analysis according to claim 1, characterized in that: The key components include, but are not limited to, the shaft system, gearbox, and combustion chamber.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 4.

7. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

8. A program product comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

Images