Hydrogen internal combustion engine knock early warning and positioning method and system, storage medium and equipment

Abstract

The application discloses a hydrogen internal combustion engine knock early warning and positioning method and system, a storage medium and equipment. The method comprises the following steps: arranging at least three acoustic emission sensors on a cylinder head and a cylinder body to form an array, and collecting ultrasonic stress wave signals in a combustion process; performing high-speed synchronous sampling on signals of each channel, and intercepting data in a corresponding crank angle window according to an engine working cycle; performing band-pass filtering on the intercepted signals, extracting high-frequency characteristics, calculating a knock index, and determining a current cylinder knock risk level; acquiring a cylinder body temperature in real time, dynamically correcting a sound velocity parameter based on a sound velocity-temperature function, and inversely calculating spatial coordinates of a knock source through a time difference algorithm; when the knock index exceeds a threshold value, outputting early warning and position information to an engine ECU, triggering ignition timing and hydrogen injection strategy adjustment, and realizing accurate knock positioning and fine prevention and control.

Description

Technical Field

[0001] This invention relates to the field of internal combustion engine knock monitoring, specifically to a method, system, storage medium, and device for early warning and location of knock in hydrogen internal combustion engines. Background Technology

[0002] With the rapid development of hydrogen fuel technology, zero-carbon emission hydrogen internal combustion engines have become an important direction for the transformation and upgrading of the internal combustion engine industry. However, hydrogen itself has the inherent characteristics of fast combustion speed and poor anti-knock properties. Under high pressure ratio and high load conditions, the air-fuel mixture at the end of the cylinder is very prone to local spontaneous combustion before the main flame front arrives, which can lead to knocking.

[0003] Currently, several mature knock detection technologies have been developed in the industry. Among them, the in-cylinder pressure detection method collects combustion pressure signals by installing pressure sensors in the combustion chamber, extracts high and low frequency components through empirical mode decomposition, and calculates the knock factor. This method has the highest detection accuracy and is mostly used in engine bench calibration scenarios. The cylinder block vibration detection method uses piezoelectric accelerometers to collect mechanical vibration signals on the cylinder block surface. By constructing a vibration benchmark under normal operating conditions, it identifies abnormal amplitudes and is the mainstream application solution for current mass-produced vehicles. The ion current detection method relies on a dedicated detection device to collect in-cylinder ion current signals and has high sensitivity to the combustion state near the ignition point. In addition, there is an abnormal noise identification scheme that combines microphone audio acquisition and multiple vibration sensors. By matching the crankshaft angle range where knock occurs frequently with the pressure change rate characteristics, a comprehensive judgment of knock events can be made.

[0004] However, non-invasive detection solutions are susceptible to interference from engine background noise, can only capture macroscopic structural vibrations, have a detection time lag and insufficient early warning lead, making it difficult to achieve proactive vibration suppression; and known solutions only support cylinder-level knock detection, cannot locate specific knock areas, and cannot support hotspot-oriented optimization and precise control, thus failing to meet the high knock control requirements of hydrogen internal combustion engines. Summary of the Invention

[0005] In view of the shortcomings of the prior art, the purpose of this invention is to provide a method, system, storage medium and device for early warning and location of knocking in hydrogen internal combustion engines, in order to solve the above-mentioned problems described in the prior art.

[0006] A first aspect of the present invention is to provide a method for early warning and location of knocking in a hydrogen internal combustion engine, the method comprising: At least three acoustic emission sensors are arranged on the cylinder head and cylinder block of the hydrogen internal combustion engine to form an acoustic emission sensor array to collect ultrasonic stress wave signals generated during hydrogen combustion. The output signals of each acoustic emission sensor are sampled at high speed and synchronously, and the sampled data within the corresponding crankshaft angle window is extracted according to the engine working cycle; The intercepted signals from each channel are subjected to bandpass filtering to extract high-frequency ultrasonic features and calculate the knock index. The knock risk level of the current cylinder is determined based on the knock index. Real-time engine block temperature parameters are acquired, sound velocity parameters are dynamically corrected based on the material sound velocity-temperature function, and the spatial coordinates of the knock source in the cylinder are obtained by using the time difference of arrival algorithm. When the knock index exceeds a preset threshold, a knock warning signal and the location information of the knock source are output to the engine ECU, triggering the engine ECU to adjust the ignition timing and hydrogen injection strategy.

[0007] According to one aspect of the above technical solution, the acoustic emission sensor array is configured with four non-coplanar acoustic emission sensors to form a non-coplanar tetrahedral array. The specific steps for arranging the four acoustic emission sensors include: A spark plug-side sensor is arranged on the top of the cylinder head, close to the spark plug base, as a time reference channel for the acoustic emission sensor array. An intake valve side sensor is arranged on the outside of the intake valve passage in the cylinder head, and the sound distance between the intake valve side sensor and a potential ignition source is controlled to be no more than 30mm. An exhaust valve side sensor is placed in the cylinder head bridge area directly above the exhaust valve, corresponding to the high-incidence area of ​​spontaneous combustion of the end mixture; A cylinder wall side sensor is arranged in the middle section of the cylinder block bushing, and the distance between the cylinder wall side sensor and the other three acoustic emission sensors in the Z-axis direction is controlled to be no less than 40mm. The geometric precision factor (GDOP) of the non-coplanar tetrahedral array is controlled to be less than 3.

[0008] According to one aspect of the above technical solution, the steps of performing high-speed synchronous sampling, cyclic truncation, bandpass filtering, and calculating the detonation index on the output signals of each of the acoustic emission sensors specifically include: The same sampling clock is used to synchronously sample the signals of each channel, the sampling frequency is not less than 1MHz, and the phase error between multiple channels is not greater than 0.5μs; A second-order Butterworth bandpass filter is used to filter the sampled signal. The filter passband range is 20kHz-500kHz, and the corresponding time domain signal is output. The detonation index is constructed based on the energy gradient of the filtered high-frequency ultrasonic signal.

[0009] According to one aspect of the above technical solution, in the step of inverting the spatial coordinates of the detonation source using the time difference of arrival algorithm, the calculation of the time difference of arrival specifically includes: Using the spark plug-side sensor as a reference channel, the filtered discrete signals of each channel are captured within a preset crankshaft angle window; Fast Fourier transform is performed on the discrete signals of the reference channel and the other channels respectively to obtain the corresponding frequency domain signals; Construct the phase-weighted cross-power spectrum between each channel and the reference channel, and perform an inverse Fourier transform on the phase-weighted cross-power spectrum to obtain the generalized cross-correlation function; The time delay corresponding to the maximum peak value of the generalized cross-correlation function is taken as the arrival time difference of the corresponding channel relative to the reference channel; Calculate the peak confidence level of each channel. When the peak confidence level is lower than a preset threshold or the arrival time difference exceeds the maximum propagation delay range limited by the engine geometry and material sound speed, the corresponding channel is marked as a low confidence channel, its weight is reduced, or it is removed.

[0010] According to one aspect of the above technical solution, the step of using the arrival time difference algorithm to invert and obtain the spatial coordinates of the knock source within the cylinder specifically includes: Based on the elastic modulus and density of the cylinder material, the corrected sound velocity at the current temperature is calculated according to the sound velocity-temperature function of the material; Using the spatial coordinates of the detonation source as unknowns and the known installation coordinates of each acoustic emission sensor as geometric constraints, multiple sets of hyperboloid constraint equations are established based on the corrected sound velocity and the arrival time difference, forming an overdetermined set of equations. The target function for positioning is established by using the peak confidence scores of each channel as weights, and the weighted least squares iterative solution is used to obtain the estimated location of the detonation source in the current working cycle. Calculate the positioning residual for the current loop. When the positioning residual exceeds a preset threshold, mark the position estimation result of the loop as a low-confidence result.

[0011] According to one aspect of the above technical solution, after obtaining the estimated location of the detonation source in the current working cycle, multi-period sliding window clustering processing is also performed, specifically including: The detonation source locations estimated from the most recent M effective working cycles are used to form a sliding window, where M ranges from 5 to 10. Based on the cylinder geometry and positioning error, the neighborhood radius and minimum number of neighborhood points are set. Within the sliding window, a density clustering method based on Euclidean distance threshold is used to remove isolated noise points and select the main clusters. The weighted centroid of the main cluster is calculated using the peak confidence level or the detonation index corresponding to each point within the main cluster as weights. The weighted centroid is output as the final knock position and mapped to the corresponding geometric region of the cylinder; If there is no main cluster within the sliding window that meets the point requirement, then a location confidence deficiency flag is output.

[0012] According to one aspect of the above technical solution, the step of triggering the engine ECU to adjust the ignition timing and hydrogen injection strategy specifically includes: Based on the cylinder region where the knock source is located, an ignition angle delay command for the corresponding cylinder is sent to the engine ECU. Alternatively, it can send direct injection pressure and hydrogen injection timing adjustment commands to the engine ECU to change the air-fuel mixture equivalence ratio in the knocking region.

[0013] A second aspect of the present invention is to provide a hydrogen internal combustion engine knock warning and positioning system, applied to the method described in the above-mentioned technical solution, the system comprising: The signal acquisition module is used to arrange at least three acoustic emission sensors on the cylinder head and cylinder block of the hydrogen internal combustion engine to form an acoustic emission sensor array, and to acquire the ultrasonic stress wave signal generated during hydrogen combustion. The synchronous sampling module is used to perform high-speed synchronous sampling of the output signals of each of the acoustic emission sensors, and to extract the sampling data within the corresponding crankshaft angle window according to the engine working cycle; The signal processing module is used to perform bandpass filtering on the intercepted signals from each channel, extract high-frequency ultrasonic features and calculate the knock index, and determine the knock risk level of the current cylinder based on the knock index. The positioning and calculation module is used to acquire the engine cylinder block temperature parameters in real time, dynamically correct the sound velocity parameters based on the material sound velocity-temperature function, and use the arrival time difference algorithm to invert the spatial coordinates of the knock source in the cylinder. The closed-loop control module is used to output a knock warning signal and the location information of the knock source to the engine ECU when the knock index exceeds a preset threshold, thereby triggering the engine ECU to adjust the ignition timing and hydrogen injection strategy.

[0014] A third aspect of the present invention is to provide a readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the steps of the method described in the above-described technical solution.

[0015] A fourth aspect of the present invention is to provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method described in the above technical solutions.

[0016] Compared with existing technologies, the advantages of using the hydrogen internal combustion engine knock warning and location method, system, storage medium and equipment shown in this invention are as follows: This invention employs an acoustic emission sensor array to acquire high-frequency ultrasonic stress wave signals propagating from the cylinder block's metal substrate. Combined with bandpass filtering to extract the knock characteristic frequency band, it can capture knock precursor signals in the early stages of knock development, before macroscopic structural vibrations have formed. This effectively avoids interference from low-frequency mechanical noises such as piston knocking and valve vibration. Compared to traditional vibration and acoustic non-invasive detection methods, it significantly extends the warning lead time, improves detection accuracy, and is well-suited to the characteristics of rapid hydrogen combustion and knock development, providing a time basis for proactive knock suppression. Simultaneously, based on the time-of-arrival algorithm, the spatial coordinates of the knock source are inverted and dynamically corrected, offsetting sound velocity drift caused by temperature changes under operating conditions. This ensures positioning accuracy across all operating conditions and precisely locates the specific area where knock occurs. It provides data support for cylinder cooling channel optimization, hydrogen injection timing adjustment, and combustion chamber structure improvement, meeting the needs of directional knock control in hydrogen internal combustion engines. Attached Figure Description

[0017] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a flowchart illustrating the method for early warning and location of knocking in a hydrogen internal combustion engine provided in an embodiment of the present invention. Figure 2 This is a structural block diagram of a hydrogen internal combustion engine knock warning and positioning system provided in an embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Several embodiments of the present invention are shown in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that the disclosure of the present invention will be more thorough and complete.

[0019] It should be noted that when a component is said to be "fixed to" another component, it can be directly on the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0021] Example 1 Please see Figure 1 The first embodiment of the present invention provides a method for early warning and location of knocking in a hydrogen internal combustion engine, the method comprising steps S10-S50: Step S10: Arrange at least three acoustic emission sensors on the cylinder head and cylinder block of the hydrogen internal combustion engine to form an acoustic emission sensor array to collect ultrasonic stress wave signals generated during hydrogen combustion.

[0022] Specifically, firstly, based on the three-dimensional structure of the engine cylinder head and cylinder block, a machined surface with good flatness is selected as the mounting point. For example, the roughness of the mounting surface is required to be controlled within Ra1.6. The paint, oxide layer and oil stains on the mounting surface are removed, and a high-temperature resistant acoustic coupling agent, such as high-temperature silicone grease or ceramic-based coupling adhesive, is applied. The thickness of the coupling layer is controlled to be less than 0.1mm. Then, the sensor is fixed by threaded mounting brackets or high-temperature structural adhesive to ensure that the coupling surface is completely fitted without air gaps.

[0023] When there are three sensors, they can be arranged on the spark plug side, exhaust valve side and intake valve side of the top surface of the cylinder head to form a planar triangular array, which can meet the basic knock warning and two-dimensional planar positioning requirements and is suitable for cost-sensitive mass production vehicle scenarios; when there are four sensors, high-precision positioning in three-dimensional space can be achieved, which is mostly used for bench calibration and high-performance models.

[0024] After the sensor is installed, a standard lead-breaking test is used to verify the coupling consistency. A 0.5mm diameter lead core is broken vertically at a fixed position from the sensor. Under the same standard sound source, the peak amplitude deviation of the output signal of each channel should be controlled within 20% to ensure the consistency of subsequent time delay calculations.

[0025] Step S20: High-speed synchronous sampling of the output signals of each acoustic emission sensor is performed, and sampling data within the corresponding crankshaft angle window is extracted according to the engine working cycle.

[0026] Specifically, high-speed synchronous sampling and cyclic truncation are used to preprocess the signal. At the hardware level, a multi-channel synchronous acquisition circuit is used, with the clock output from the same high-precision temperature-compensated crystal oscillator serving as the sampling clock for all channels to ensure that the clocks are strictly from the same source. The acquisition system connects to the engine crankshaft position sensor's speed signal and top dead center signal to obtain the real-time crankshaft phase. It prioritizes the use of an equal-angle sampling mode, that is, a point is collected every fixed angle of rotation, rather than simply equal-time sampling, to ensure that the angle position of the analysis window is always aligned at different speeds.

[0027] In practice, the top dead center of ignition is taken as the zero point of time, and the effective interception window is set from 0°CA to 40°CA after top dead center. This covers the combustion stage where knocking is common. Since the main combustion period of hydrogen internal combustion engines is usually concentrated within 30°CA after top dead center, knocking events mostly occur in the 10-30°CA range. This window not only fully covers the knocking period, but also avoids introducing too much interference from exhaust and intake strokes.

[0028] By converting the crankshaft angle window to a time window using the relationship between engine speed and crankshaft angle, the conversion formula is: Window Time = (Angle Range / 360) / (Speed ​​ / 60). For example, at 3000 rpm, 40° CA corresponds to a time length of approximately 2.22 ms. Data capture is triggered once per engine working cycle, outputting a fixed length of single-cycle sampled data.

[0029] Step S30: Bandpass filtering is performed on the intercepted channel signals to extract high-frequency ultrasonic features and calculate the knock index. The knock risk level of the current cylinder is determined based on the knock index.

[0030] Specifically, in practice, bandpass filtering involves implementing a digital bandpass filtering algorithm in an embedded processor to remove low-frequency mechanical noise and retain the ultrasonic signal in the detonation characteristic frequency band. After filtering, the energy characteristics of the high-frequency signal are extracted, and a detonation index is constructed to quantify detonation risk: the filtered signal is divided into frames according to a fixed time window, typically 10 μs in length, with an inter-frame overlap rate of 50% to ensure a smooth energy curve. The signal energy is calculated frame by frame, i.e., the sum of squared amplitudes of all sampling points within the frame, yielding a sequence of energy changes over time. The first-order difference of this energy sequence is then calculated to obtain the energy gradient. The maximum gradient value within the window, after normalization, is taken as the detonation index for the current cycle.

[0031] The reason for choosing energy gradient rather than absolute energy as the criterion is that absolute energy varies greatly with engine load and speed, and the basic level differs significantly under different operating conditions, making it difficult to unify the threshold; while energy gradient reflects the energy mutation caused by sudden spontaneous combustion, and is more robust to changes in operating conditions.

[0032] It should also be noted that the knock detection threshold is determined through bench calibration. Specifically, multi-cycle data is collected under normal operating conditions without knock, and 3-5 times the maximum energy gradient value under each condition is taken as the corresponding threshold. This is used to create a two-dimensional speed-load MAP table and stored in the controller. During real-time operation, the threshold is obtained by looking up the table based on the current operating condition. The real-time knock index is then compared with the threshold to output the knock risk level.

[0033] Step S40: Real-time acquisition of engine cylinder block temperature parameters, dynamic correction of sound velocity parameters based on material sound velocity-temperature function, and inversion of the spatial coordinates of the knock source in the cylinder using the time difference of arrival algorithm.

[0034] Specifically, during correction and inversion, the cylinder block temperature parameters are first obtained. This can be achieved by reading the engine coolant temperature signal or by using a patch temperature sensor mounted on the cylinder wall to obtain the local substrate temperature. In mass production scenarios, coolant temperature is typically combined with load and engine speed for correction estimation, while in bench tests, patch temperature sensors can be used directly for higher accuracy. By substituting the temperature value into a pre-calibrated material sound velocity-temperature function, the corrected sound velocity under the current operating conditions is calculated, replacing the fixed sound velocity parameter. Subsequently, the time difference of arrival (TDOA) localization algorithm is invoked, taking into account the TDOA of each channel, sensor calibration coordinates, and the corrected sound velocity, and iteratively optimizing the solution to obtain the spatial coordinates of the detonation source.

[0035] The positioning algorithm is embedded in the controller as a function module. It is called once in each loop, and the calculation time of a single loop is usually controlled within 1ms, which meets the requirements of real-time computing on the vehicle. It outputs the estimated value of the knock source position in the current loop and completes the mapping from one-dimensional time observation to three-dimensional spatial coordinates.

[0036] Step S50: When the knock index exceeds a preset threshold, a knock warning signal and the location information of the knock source are output to the engine ECU, triggering the engine ECU to adjust the ignition timing and hydrogen injection strategy.

[0037] Specifically, the system communicates with the engine ECU via the vehicle bus, also known as the CAN bus, which typically uses the CAN2.0B protocol. The message priority is set to high priority, and the cycle is no more than 10ms. The communication protocol for knock warning messages and location information messages is predefined.

[0038] When the knock index exceeds the calibrated threshold, an early warning message is immediately sent to the ECU. The message contains information such as the knock risk level and the knock source area code. The risk level can be divided into three levels: mild, moderate and severe, corresponding to different control intensities.

[0039] Furthermore, the ECU is pre-configured with lookup logic for the corresponding control strategy. Upon receiving a warning signal, it calls the corresponding control parameters based on the knock location and level, and performs ignition timing or hydrogen injection strategy adjustments to suppress knock development before macroscopic structural vibrations appear, thus forming a complete detection-location-control closed loop.

[0040] Compared with existing technologies, the advantages of using the hydrogen internal combustion engine knock warning and location method shown in this embodiment are as follows: This embodiment employs an acoustic emission sensor array to acquire high-frequency ultrasonic stress wave signals propagating from the cylinder block's metal substrate. Combined with bandpass filtering to extract the knock characteristic frequency band, it can capture knock precursor signals in the early stages of knock development, before macroscopic structural vibrations have formed. This effectively avoids interference from low-frequency mechanical noises such as piston knocking and valve vibration. Compared to traditional vibration and acoustic non-invasive detection schemes, it significantly extends the warning lead time, improves detection accuracy, and is well-suited to the characteristics of rapid hydrogen combustion and knock development, providing a time basis for proactive knock suppression. Simultaneously, by using a time-of-arrival algorithm to invert the spatial coordinates of the knock source and dynamically correct them, it can offset the sound velocity drift caused by temperature changes under operating conditions, ensuring positioning accuracy across all operating conditions. It can precisely locate the specific area where knock occurs, providing data support for cylinder cooling channel optimization, hydrogen injection timing adjustment, and combustion chamber structure improvement, meeting the needs of directional knock control in hydrogen internal combustion engines.

[0041] Example 2 The second embodiment of the present invention also provides a method for knock warning and location of hydrogen internal combustion engines. The method shown in this embodiment is basically the same as the method shown in the first embodiment, except that: In this embodiment, the acoustic emission sensor array is configured with four non-coplanar acoustic emission sensors, forming a non-coplanar tetrahedral array; The specific steps for arranging the four acoustic emission sensors include: A spark plug-side sensor is arranged on the top of the cylinder head, close to the spark plug base, as a time reference channel for the acoustic emission sensor array. An intake valve side sensor is arranged on the outside of the intake valve passage in the cylinder head, and the sound distance between the intake valve side sensor and a potential ignition source is controlled to be no more than 30mm. An exhaust valve side sensor is placed in the cylinder head bridge area directly above the exhaust valve, corresponding to the high-incidence area of ​​spontaneous combustion of the end mixture; A cylinder wall side sensor is arranged in the middle section of the cylinder block bushing, and the distance between the cylinder wall side sensor and the other three acoustic emission sensors in the Z-axis direction is controlled to be no less than 40mm. The geometric precision factor (GDOP) of the non-coplanar tetrahedral array is controlled to be less than 3.

[0042] Before formal installation, the spatial coordinates of four candidate installation points are marked in the engine's 3D CAD model. These coordinates are then substituted into the geometric accuracy factor formula for simulation calculation. The expression is as follows:

[0043] In the formula, This is the TDOA observation matrix, where each row of the matrix consists of the unit direction vector components of the corresponding sensor relative to the reference sensor; The trace operation represents the sum of the values ​​of all elements along the main diagonal of the matrix; GDOP is the amplification factor of the time measurement error into the positioning error. The smaller the value, the better the array geometry and the higher the positioning accuracy. All typical knock locations within the cylinder are traversed, and the coordinates of each mounting point are adjusted to ensure that the maximum GDOP value across the entire area is less than 3. During the adjustment process, it is crucial to ensure that the Z-axis distance between the cylinder wall side sensor and the other three cylinder head sensors is not less than 40mm to prevent the array from becoming nearly coplanar, which could lead to a sharp deterioration in vertical positioning accuracy.

[0044] Furthermore, due to the significant spatial concentration of knocking in hydrogen internal combustion engines, it mainly focuses on four typical hot spots: the spark plug flame core area, the intake valve side, the exhaust valve side, and the cylinder wall boundary layer. Acoustic emission signals propagate in metallic media following an exponential decay law.

[0045] In the formula, The amplitude of the acoustic emission signal when it reaches the sensor; is the acoustic attenuation coefficient of metallic materials, and its value is related to the signal frequency and the internal grain characteristics of the material. Let be the linear distance of acoustic propagation from the detonation source to the sensor. This formula shows that the signal amplitude decays exponentially with the propagation distance; for every doubling of the propagation distance, the signal amplitude decreases by approximately 6 dB, resulting in a significant drop in the signal-to-noise ratio. This scheme controls the distance difference between each typical detonation source and the array centroid to not exceed half the distance difference between the nearest and farthest sensors, ensuring that the acoustic path difference from all hotspot areas to the sensors is within a reasonable range. The signal-to-noise ratio at the farthest point is not less than 50% of that at the nearest point, preventing the signal attenuation of distant hotspots below the noise level and ensuring no detection blind spots across the entire area.

[0046] Furthermore, finite element modal analysis can be used to obtain the vibration modes of the cylinder block within the commonly used speed range, identifying the displacement node positions of the low- and mid-frequency modes. During operation, the cylinder block excites multiple structural vibration modes. Low- and mid-frequency vibrations exhibit antinodes and nodes at different locations. The displacement amplitude is largest at antinodes, resulting in the strongest interference to the ultrasonic sensor; at nodes, the displacement is close to zero, resulting in the weakest interference. Sensors are preferentially placed near the high-rigidity ribs of the cylinder head and cylinder block. These locations typically correspond to the nodes of the low- and mid-frequency vibration modes. Mechanical vibration crosstalk can be suppressed through the structural layout itself, improving the signal-to-noise ratio in the ultrasonic band without the need for additional complex algorithms—a hardware-level anti-interference design.

[0047] The installation of the four sensors must strictly adhere to positional constraints. The spark plug-side sensor is mounted on the top of the cylinder head, on a plane adjacent to the spark plug base, no more than 10mm from the edge of the spark plug base. A small, specially machined boss can be used to ensure the flatness of the mounting surface. Besides serving as a time reference channel, this sensor also assists in monitoring ignition energy and the initial development state of the flame nucleus. After installation, the synchronization of its signal rise edge with the ignition timing must be verified to ensure that each ignition cycle can serve as a time reference. The intake valve-side sensor is located on the cylinder head sidewall outside the intake manifold, avoiding the injector mounting location to reduce injection impact noise. A coordinate measuring machine should be used to confirm that the linear sound path between the center of the sensor's sensitive surface and the edge of the intake valve seat is no greater than 30mm to ensure rapid capture of pre-combustion and backfire precursors on the intake side. The exhaust valve side sensor is installed directly above the exhaust valve on the cylinder head nose area, aligned with the high-knock zone 15mm outside the exhaust valve. The mounting surface must avoid the thin-walled area of ​​the coolant jacket to reduce acoustic interference from water flow noise. The sensor's temperature resistance rating must be above 250℃ to adapt to high-temperature exhaust conditions. The cylinder wall side sensor is installed in the middle section of the cylinder block side bushing, corresponding to the piston crown height at top dead center. The mounting base can be modified using the cylinder block water plug hole to ensure reliable acoustic coupling with the cylinder wall metal, while ensuring proper sealing to prevent coolant leakage.

[0048] After the sensor array is installed, overall performance verification and calibration will be performed. First, a unified coordinate system is established, typically with the cylinder center axis as the Z-axis, the cylinder top surface as the XY plane, and the spark plug center as the origin. The precise coordinates of each sensor in this coordinate system are calibrated using a coordinate measuring machine. A standard lead-breaking test is used to verify positioning accuracy: at known coordinate locations within the cylinder, such as the exhaust valve edge, intake valve edge, and a designated location on the cylinder wall, a standard lead-breaking operation is performed to simulate a point sound source. Response signals from each channel are collected, and the positioning algorithm is run to obtain calculated coordinates. The deviation between the calculated values ​​and the actual coordinates is compared. If the deviation exceeds the allowable range, the sensor coupling state, coordinate calibration values, and synchronization accuracy must be rechecked until the positioning error meets the design requirements. Simultaneously, the signal sensitivity of each channel is verified, and the coupling pressure and installation torque are adjusted to ensure that the peak amplitude deviation of each channel under the same sound source is controlled within 20%, guaranteeing the accuracy and stability of subsequent time delay estimation.

[0049] In this embodiment, the steps of performing high-speed synchronous sampling, cyclic truncation, bandpass filtering, and calculating the detonation index on the output signals of each acoustic emission sensor specifically include: The same sampling clock is used to synchronously sample the signals of each channel, the sampling frequency is not less than 1MHz, and the phase error between multiple channels is not greater than 0.5μs; A second-order Butterworth bandpass filter is used to filter the sampled signal. The filter passband range is 20kHz-500kHz, and the corresponding time domain signal is output. The detonation index is constructed based on the energy gradient of the filtered high-frequency ultrasonic signal.

[0050] Specifically, the sampling hardware uses an analog-to-digital converter chip that supports multi-channel synchronous conversion. The analog front end is equipped with an anti-aliasing low-pass filter to filter out signal components above 500kHz in the analog domain to prevent sampling aliasing. The conversion start signal of all channels is driven by the same clock source to ensure that the sampling time of each channel is strictly aligned.

[0051] Then, a dual second-order section structure is used to implement the IIR digital filter, with a passband range of 20kHz-500kHz. During the design, the five coefficients b0, b1, b2, a1, and a2 of the transfer function are calculated first based on the sampling frequency and cutoff frequency using the Butterworth filter design formula. Here, b0, b1, and b2 are the numerator polynomial coefficients, and a1 and a2 are the denominator polynomial coefficients. The filtering operation uses a direct type II topology, performing point-by-point recursive operations on the input sampled sequence to output the filtered time-domain signal.

[0052] After the filter design is completed, the amplitude-frequency characteristics need to be verified. The attenuation at the cutoff frequencies of 20kHz and 500kHz should not exceed 3dB, and the attenuation in the frequency bands below 10kHz and above 600kHz should be greater than 20dB to ensure effective filtering of low-frequency mechanical vibrations and high-frequency electromagnetic interference. Furthermore, to avoid the filter's phase delay affecting time delay estimation, the group time delay parameters of the filter can be pre-determined and fixedly compensated in subsequent time delay calculations, or a zero-phase filtering algorithm can be used to eliminate phase shift.

[0053] Next, the detonation index is calculated. First, the filtered signal is divided into frames with a time window of 10μs length, and the number of sampling points in each frame is equal to the sampling frequency multiplied by 0.00001. Then, the signal energy is calculated frame by frame, that is, the sum of the squares of the amplitudes of all sampling points in the frame, to obtain the energy time series. Then, the first-order difference operation is performed on the energy series to obtain the energy gradient series. Finally, the maximum value of the gradient within the window is taken, and after normalization by the operating conditions, it is used as the detonation index of the current cycle.

[0054] Because abnormal jumps in a single frame are filtered during the calculation process, if the energy of a single frame suddenly increases but subsequent frames do not show a continuous upward trend, it is judged as instantaneous pulse interference and is not included in the gradient statistics.

[0055] Furthermore, the threshold calibration is performed on an engine test bench, covering typical operating conditions such as idling, partial load, and full load. By collecting multi-cycle gradient data under non-knock conditions, 3-5 times the maximum gradient value for each operating condition is taken as the knock detection threshold, and a corresponding two-dimensional speed-load MAP table is created. During real-time operation, the above threshold is obtained by looking up the table according to the current operating condition, thereby completing the knock risk level determination.

[0056] In this embodiment, the calculation of the time difference of arrival specifically includes the following steps in the step of inverting the spatial coordinates of the detonation source using the time difference of arrival algorithm: Using the spark plug-side sensor as a reference channel, the filtered discrete signals of each channel are captured within a preset crankshaft angle window; Fast Fourier transform is performed on the discrete signals of the reference channel and the other channels respectively to obtain the corresponding frequency domain signals; Construct the phase-weighted cross-power spectrum between each channel and the reference channel, and perform an inverse Fourier transform on the phase-weighted cross-power spectrum to obtain the generalized cross-correlation function; The time delay corresponding to the maximum peak value of the generalized cross-correlation function is taken as the arrival time difference of the corresponding channel relative to the reference channel; Calculate the peak confidence level of each channel. When the peak confidence level is lower than a preset threshold or the arrival time difference exceeds the maximum propagation delay range limited by the engine geometry and material sound speed, the corresponding channel is marked as a low confidence channel, its weight is reduced, or it is removed.

[0057] Specifically, in the selection of the reference channel and the time-domain signal truncation, after each cycle of data truncation, the spark plug-side sensor channel is designated as the reference channel, and the remaining sensors are used as measurement channels. The truncation window is set from the ignition moment, typically from 10°CA before top dead center (TDC) to 40°CA after TDC. The purpose of truncation at 10°CA is to retain sufficient baseline signal for calculating background noise level and reference energy, ensuring that the knock signal falls completely within the analysis window. The discrete signal of the reference channel is denoted as... , No. The discrete signals of the measurement channels are denoted as follows: ,in The index represents the discrete sampling point. The truncated discrete signal is first de-DC processed by subtracting the signal mean to eliminate the influence of DC bias on cross-correlation calculations. At the same time, a Hanning window is superimposed to suppress spectral leakage. If the sampling rate is too high and the data volume is too large, appropriate downsampling can be performed without violating the Nyquist theorem to compress the computation and improve the real-time performance of the operation.

[0058] Specifically, the discrete signals of the reference channel and each measurement channel are subjected to Fast Fourier Transform to obtain the corresponding frequency domain signals:

[0059] In the formula, This represents the Fast Fourier Transform operation; The frequency domain complex signal corresponding to the reference channel; For the first The frequency domain complex signal corresponding to each measurement channel; For frequency variables. The number of FFT points is taken as the smallest power of 2 greater than the signal length, such as 256 points or 512 points. If the number is insufficient, zeros are padded to improve the frequency domain resolution.

[0060] The phase-weighted cross-power spectrum is then constructed, and its expression is:

[0061] In the formula, Reference channel frequency domain signal The complex conjugate; Modular operation for complex numbers; To prevent extremely small positive numbers with a denominator of zero, the typical value is [value missing]. ; For the first Phase-weighted cross-power spectrum between the channel and the reference channel.

[0062] The above formula, by normalizing the cross-power spectrum by amplitude and retaining only phase information, can effectively suppress the broadening effect of amplitude attenuation and multipath interference on the cross-correlation peak, making the time delay peak sharper.

[0063] The generalized cross-correlation function is solved, and the time delay is extracted to output the final time delay result. An inverse fast Fourier transform is performed on the phase-weighted cross-power spectrum to obtain the generalized cross-correlation function in the time domain, expressed as:

[0064] In the formula, This indicates the inverse fast Fourier transform operation; For time delay variables; For the first The generalized cross-correlation function of a channel relative to a reference channel characterizes the similarity between the two signals at different time delays. The sequence zeros correspond to zero time delay, while the left and right sides correspond to positive and negative time delays, respectively.

[0065] The maximum peak value is searched within the effective physical delay range of the cross-correlation function. The time index corresponding to the peak value is multiplied by the sampling interval, which gives the arrival time difference between the measurement channel and the reference channel. The expression is:

[0066] In the formula, This represents finding the independent variable that maximizes the function. The possible values ​​of ; That is, the first The time difference of arrival of the channel signal relative to the reference channel signal.

[0067] To overcome the time resolution limitation of the sampling rate, parabolic interpolation can be performed on three sampling points near the peak. The extreme moment of the sub-sampling level can be obtained by fitting a quadratic polynomial, thus obtaining a high-precision time delay estimate of the sub-sampling level. This can typically improve the time resolution by 3-5 times, further enhancing the positioning accuracy.

[0068] Furthermore, the peak confidence level for each channel is calculated using the following expression:

[0069] In the formula, The maximum peak amplitude of the generalized cross-correlation function; This is the average of the absolute values ​​of all sampling points of the generalized cross-correlation function; For the first The peak confidence level of the channel indicates that the higher the value, the more prominent the delay peak and the higher the reliability of the estimation result.

[0070] The confidence threshold is typically set between 3 and 5; values ​​below this threshold are considered unreliable. Peak width is also checked; excessively wide peaks indicate severe multipath overlap, reducing time delay accuracy. Physical plausibility is also verified: based on the engine block's maximum geometric dimensions and material sound velocity, the theoretical maximum propagation time delay is calculated. If the calculated arrival time difference exceeds this range, it is considered invalid. Channels that simultaneously meet the confidence requirements and physical range are marked as valid channels; otherwise, they are marked as low-confidence channels, and their weight is significantly reduced in subsequent positioning processes, or they are directly excluded from the positioning equations to avoid erroneous observations interfering with the positioning results.

[0071] Example 3 In this embodiment, the step of using the time difference of arrival algorithm to invert and obtain the spatial coordinates of the knock source within the cylinder specifically includes: Based on the elastic modulus and density of the cylinder material, the corrected sound velocity at the current temperature is calculated according to the sound velocity-temperature function of the material; Using the spatial coordinates of the detonation source as unknowns and the known installation coordinates of each acoustic emission sensor as geometric constraints, multiple sets of hyperboloid constraint equations are established based on the corrected sound velocity and the arrival time difference, forming an overdetermined set of equations. The target function for positioning is established by using the peak confidence scores of each channel as weights, and the weighted least squares iterative solution is used to obtain the estimated location of the detonation source in the current working cycle. Calculate the positioning residual for the current loop. When the positioning residual exceeds a preset threshold, mark the position estimation result of the loop as a low-confidence result.

[0072] Specifically, the calibration of the material's sound velocity-temperature function is a prerequisite for implementing dynamic correction. The propagation velocity of longitudinal waves in a solid is determined by both the material's elastic modulus and density, with the basic formula being:

[0073] In the formula, This refers to the current temperature of the cylinder block material. For temperature Elastic modulus of the lower cylinder block material; The density of the cylinder block material is approximately constant because the density change due to thermal expansion is minimal in engineering applications. For temperature Longitudinal wave velocity in a solid medium.

[0074] For commonly used cylinder block materials such as cast iron and aluminum alloy, the elastic modulus decreases linearly with increasing temperature. Therefore, the overall sound velocity decreases approximately linearly with increasing temperature. The sound velocity function is preferably obtained through online calibration rather than directly applying the material's theoretical value: using spark plug ignition as a standard sound source at a known location, the arrival time difference is measured under different operating conditions to deduce the actual sound velocity, and an operating condition-sound velocity MAP table is established, which more closely reflects the actual material properties of engine castings.

[0075] The above can also be calibrated through sample experiments. A metal sample of the same material as the engine block is cut and placed in a high and low temperature test chamber. Multiple temperature points are set within the range of -40℃ to 200℃. The longitudinal wave velocity at each temperature is measured using the ultrasonic pulse transmission method, obtaining multiple sets of temperature-velocity data pairs. Linear fitting is then performed on the data points to obtain a directly usable sound velocity-temperature expression. During actual operation, the cylinder block temperature is collected in real time. Substitute the calibrated function to calculate the current corrected sound velocity, and replace the sound velocity parameter in the positioning equation.

[0076] Then, let the three-dimensional coordinates of the detonation source be unknowns. , No. The known installation coordinates of the sensors are The reference sensor coordinates are After installation, the sensor's coordinate values ​​can be accurately calibrated using a coordinate measuring machine. For the first... Based on the physical relationship that the distance difference between the sound source and the two sensors equals the speed of sound multiplied by the time difference of arrival, a hyperboloid constraint equation is established for sensor 1 and reference sensor 1. The expression is:

[0077] In the formula, Represents the Euclidean distance in three-dimensional space, which is the straight-line distance between two points; The corrected speed of sound at the current temperature; For the first The arrival time difference between the channel and the reference channel.

[0078] It should be noted that the set of points whose distance difference to the two focal points is constant constitutes a hyperboloid, and the true knock source must lie on this hyperboloid. Each effective measurement channel corresponds to an independent equation, and N effective channels can establish N-1 equations, forming an overdetermined nonlinear equation system. The overdetermined characteristics can be optimized to offset some random measurement errors, yielding results that are more stable than the definite solution equations. Four non-coplanar sensors can yield three independent equations, theoretically solving for two spatial intersection points. Combining this with the physical boundaries of the cylinder, invalid solutions outside the cylinder can be discarded, resulting in a unique knock source location.

[0079] Then, the Gauss-Newton iterative method is used to solve the nonlinear least squares problem. The localization objective function is constructed using the peak confidence scores of each channel as weights, and its expression is:

[0080] In the formula, For the first The weight coefficients of each observation channel are positively correlated with the peak confidence level of the corresponding channel; the higher the confidence level, the greater the weight. The objective function is defined as the weighted sum of squared residuals of all observation equations.

[0081] The above goal is to find the objective function that makes the objective function... The smallest coordinate point, which is also the smallest weighted sum of squared residuals of all observation equations, corresponds to the following expression:

[0082] In the formula, This represents finding the independent variable that minimizes the objective function. The possible values ​​of ; For the first Estimates of the detonation source location for each working cycle.

[0083] Specifically, first, set the initial values ​​for the iteration. Usually, the geometric center of the cylinder is taken as the initial coordinates. Alternatively, the planar positioning results of the three sensors can be used as the initial values ​​to speed up convergence and reduce the number of iterations. Then, perform a Taylor expansion of the constraint equations at the current coordinates, retain the first-order terms, and linearize the nonlinear equations to obtain a system of linear equations. Next, solve the weighted system of linear equations to obtain the coordinate correction. Finally, add the correction to the current coordinates to complete one iteration.

[0084] By repeating the linearization and solution process, the coordinate correction modulus is kept below the convergence threshold (typically 0.1 mm) or the maximum number of iterations (typically 20) is reached. The coordinates at this point represent the estimated detonation source location for the current working cycle. If convergence is not achieved even after reaching the maximum number of iterations, the result is directly classified as low confidence. Weights are proportional to the peak confidence level; channel weights with a confidence level of 0 are reset to zero, equivalent to being removed from the equation set.

[0085] Then, positioning residual calculation and reliability verification filtering are performed. The formula for calculating positioning residual is as follows:

[0086] In the formula, The total number of effective sensor channels; For the first The positioning residuals of each work cycle characterize the overall degree of agreement between the positioning results and all observations.

[0087] The smaller the residual value, the higher the agreement between the positioning result and all the observations, and the more reliable the result. If the residual is too large, it indicates that there are significant contradictions in the observations of each channel, and it is likely to be affected by strong noise interference or the sound velocity model has a large deviation, resulting in a distorted positioning result.

[0088] It should be noted that the residual threshold is set according to the positioning accuracy requirements, usually between 2-5mm. When the residual is less than the threshold, it is considered a valid positioning result and stored in the subsequent statistical window; when the residual is greater than the threshold, it is marked as a low-confidence result and discarded, and will not participate in subsequent multi-period statistics to avoid outliers contaminating the hotspot identification results.

[0089] In this embodiment, after obtaining the estimated location of the detonation source in the current working cycle, multi-period sliding window clustering is also performed, specifically including: The detonation source locations estimated from the most recent M effective working cycles are used to form a sliding window, where M ranges from 5 to 10. Based on the cylinder geometry and positioning error, the neighborhood radius and minimum number of neighborhood points are set. Within the sliding window, a density clustering method based on Euclidean distance threshold is used to remove isolated noise points and select the main clusters. The weighted centroid of the main cluster is calculated using the peak confidence level or the detonation index corresponding to each point within the main cluster as weights. The weighted centroid is output as the final knock position and mapped to the corresponding geometric region of the cylinder; If there is no main cluster within the sliding window that meets the point requirement, then a location confidence deficiency flag is output.

[0090] Specifically, the most recent The positioning results of each effective work cycle form a sliding window, expressed as:

[0091] In the formula, This is the length of the sliding window, which is the number of effective work cycles contained within the window; a typical value is 5-10. For the first The set of sliding windows corresponding to each cycle, including the most recent The results of the detonation source location estimation for one effective cycle.

[0092] The above window length can be adaptively adjusted according to the rotational speed. At high speeds, the cycle period is short, so a larger M value can be used; at low speeds, the cycle period is long, so a smaller M value can be used to ensure that the physical time corresponding to the window is roughly constant, taking into account both statistical stability and operating condition tracking.

[0093] By setting up a separate circular queue buffer in the controller's memory to store the corresponding window data, after each working cycle of positioning calculation, if the cycle result is determined to be a valid positioning, it is written to the latest position in the buffer, while the oldest data is removed, allowing the window to slide forward cycle by cycle. If the current cycle results in a low-confidence result, it is not written to the buffer, and the window retains its original data. Furthermore, in the buffer, each data entry stores not only the 3D coordinates but also the peak confidence and detonation index of the corresponding cycle, which are used for subsequent weighted calculations and main cluster selection.

[0094] Furthermore, density clustering logic based on Euclidean distance threshold is adopted. First, the neighborhood radius is set according to the cylinder geometry and single-cycle positioning error. Typical values ​​are 5-10 mm, and usually 2-3 times the single-cycle positioning mean square error is used to ensure that positioning points with normal random fluctuations can still be classified into the same cluster. At the same time, a minimum number of neighborhood points is set. The typical value is 2-3. For any positioning point within the window... Its neighborhood set is defined as:

[0095] In the formula, For the first in the sliding window One positioning point; For the first The neighborhood set of a given point contains all points whose Euclidean distance from that point is not greater than the neighborhood radius.

[0096] A point is considered a core point when the number of points in its neighborhood meets the following condition:

[0097] In the formula, Indicates the number of elements in the set; The minimum number of neighboring points is preset. Then, starting from any unvisited core point, all density-reachable points are recursively traversed and grouped into the same cluster; this process is repeated until all core points are assigned to their corresponding clusters. Finally, all isolated points not assigned to any cluster are marked as noise points and directly discarded. Thus, this embodiment does not require pre-specifying the number of clusters and can automatically identify knock accumulation areas of arbitrary shapes, adapting to irregular hotspot distributions within the cylinder.

[0098] After clustering, by statistically analyzing the number of points contained in each cluster and the average detonation index, the cluster with the most points and the highest average detonation index is selected as the master cluster, corresponding to the most important and stable detonation hotspot region. If multiple clusters have similar number of points and detonation indices, all are output and sequentially labeled as hotspot number one, hotspot number two, etc., to facilitate calibration engineers in handling multiple abnormal regions simultaneously. The formula for calculating the weighted centroid of the master cluster is:

[0099] In the formula, The set of all localization points within the primary cluster; For the first The weight corresponding to each positioning point can be either the peak confidence level or the knock index. The weighted centroid coordinates of the main cluster are the final output locations of the detonation source.

[0100] The calculated centroid coordinates are further mapped to the preset geometric region of the cylinder. By pre-dividing the coordinate boundaries of each region in the cylinder CAD model, it is divided into four categories: spark plug region, intake valve region, exhaust valve region, and cylinder wall region, and the corresponding region code is output. Of course, for scenarios with higher positioning accuracy, it can be further subdivided into smaller geometric blocks.

[0101] If there is no main cluster within the sliding window that meets the minimum number of points, it is determined that the location confidence is insufficient, and the corresponding status flag is output. Only the knock risk warning function is retained, and the specific spatial location is not output to avoid misleading the control strategy due to incorrect positioning.

[0102] In this embodiment, the step of triggering the engine ECU to adjust the ignition timing and hydrogen injection strategy specifically includes: Based on the cylinder region where the knock source is located, an ignition angle delay command for the corresponding cylinder is sent to the engine ECU. Alternatively, it can send direct injection pressure and hydrogen injection timing adjustment commands to the engine ECU to change the air-fuel mixture equivalence ratio in the knocking region.

[0103] Specifically, the system communicates with the engine ECU via the CAN bus, and pre-defined single-cylinder ignition angle correction messages are used. The message fields include information such as cylinder number, delay angle value, and effective cycle number.

[0104] When pre-knock signs are detected in a cylinder, the delay level is determined based on the knock index: mild knock delay 1-2°CA, moderate knock delay 3-5°CA, and severe knock delay 6-8°CA. To avoid frequent ignition angle fluctuations leading to unstable combustion, an adjustment rate limit is set, ensuring that the ignition angle change per cycle does not exceed 1°CA, guaranteeing a smooth transition. The delay parameters are encapsulated into a command message and sent to the ECU. The ECU takes effect immediately in the current or next cycle, adjusting the ignition timing of the corresponding cylinder.

[0105] Meanwhile, this embodiment also includes a delayed recovery mechanism. If no knock precursor is detected for 3-5 consecutive cycles, the delay angle is gradually reduced, with each recovery being 0.5-1°CA. The recovery rate is slower than the delay rate to prevent repeated knocking until the original ignition timing is returned, thus avoiding a decrease in power performance and thermal efficiency due to continuous delay.

[0106] The direct injection pressure and hydrogen injection timing adjustment strategies are based on targeted adjustments to the location of knock, rather than a uniform global change. This approach suppresses knock while minimizing the impact on overall combustion efficiency. For hydrogen internal combustion engines using multi-stage injection, when the knock source is concentrated on the exhaust valve side, the ECU appropriately delays the hydrogen injection timing during the main injection phase. This allows the peak hydrogen concentration to avoid the exhaust valve hotspot area, reducing the air-fuel mixture equivalence at the hotspot. It can also appropriately increase the pre-injection ratio to complete partial combustion earlier, reducing the intensity of the main combustion phase. When the knock source is concentrated on the cylinder wall side, the direct injection pressure is appropriately reduced, and the injection duration is extended to enhance in-cylinder turbulent mixing, improve mixture uniformity, and eliminate locally rich areas. Furthermore, the adjustment range is determined based on the knock level; typically, the hydrogen injection timing adjustment range is ±5°CA, and the direct injection pressure adjustment range is ±10% of the rated pressure. After knock is eliminated, the injection parameters are gradually restored to their calibrated values ​​to ensure engine power output and emissions performance.

[0107] It should be noted that the system uses built-in hierarchical control logic to automatically select the optimal control combination based on the knock risk level, and the strategy emphasis differs under different operating conditions. When the knock tendency is strong under low-speed, high-load conditions, ignition delay is the primary method, supplemented by hydrogen injection adjustment, prioritizing suppression. Under medium-to-high speed conditions, the knock tendency is weak, so hydrogen injection adjustment is the primary method, minimizing changes to ignition timing to reduce power loss. Under low-level knock risk, hydrogen injection strategy is prioritized, without altering ignition timing, to maximize engine power and economy. Under medium-level risk, a small ignition delay combined with hydrogen injection parameter adjustments is used, balancing response speed and performance loss. Under high-level knock risk, a large ignition delay and hydrogen injection adjustment are applied simultaneously, and torque limiting can be added if necessary, prioritizing engine structural safety.

[0108] Furthermore, the grading thresholds and adjustment ranges of the control strategy can be customized through bench calibration. The system can also incorporate a self-learning mechanism to record the knock suppression effect after each adjustment, automatically optimize the control parameters corresponding to knock at different locations, and adapt to hydrogen internal combustion engine models with different displacements and power levels.

[0109] Example 3 The fourth embodiment of the present invention provides a knock warning and location system for a hydrogen internal combustion engine, applied to the method described in the above technical solution, the system comprising: The signal acquisition module 10 is used to arrange at least three acoustic emission sensors on the cylinder head and cylinder block of the hydrogen internal combustion engine to form an acoustic emission sensor array, and to acquire the ultrasonic stress wave signal generated during the hydrogen combustion process. The synchronous sampling module 20 is used to perform high-speed synchronous sampling of the output signals of each of the acoustic emission sensors, and to extract the sampling data within the corresponding crankshaft angle window according to the engine working cycle. Signal processing module 30 is used to perform bandpass filtering on the intercepted signals of each channel, extract high-frequency ultrasonic features and calculate the knock index, and determine the knock risk level of the current cylinder based on the knock index. The positioning and calculation module 40 is used to acquire the engine cylinder block temperature parameters in real time, dynamically correct the sound velocity parameters based on the material sound velocity-temperature function, and use the arrival time difference algorithm to invert the spatial coordinates of the knock source in the cylinder. The closed-loop control module 50 is used to output a knock warning signal and the location information of the knock source to the engine ECU when the knock index exceeds a preset threshold, thereby triggering the engine ECU to adjust the ignition timing and hydrogen injection strategy.

[0110] Compared with existing technologies, the advantages of using the hydrogen internal combustion engine knock warning and positioning system shown in this embodiment are as follows: This embodiment employs an acoustic emission sensor array to acquire high-frequency ultrasonic stress wave signals propagating from the cylinder block's metal substrate. Combined with bandpass filtering to extract the knock characteristic frequency band, it can capture knock precursor signals in the early stages of knock development, before macroscopic structural vibrations have formed. This effectively avoids interference from low-frequency mechanical noises such as piston knocking and valve vibration. Compared to traditional vibration and acoustic non-invasive detection schemes, it significantly extends the warning lead time, improves detection accuracy, and is well-suited to the characteristics of rapid hydrogen combustion and knock development, providing a time basis for proactive knock suppression. Simultaneously, by using a time-of-arrival algorithm to invert the spatial coordinates of the knock source and dynamically correct them, it can offset the sound velocity drift caused by temperature changes under operating conditions, ensuring positioning accuracy across all operating conditions. It can precisely locate the specific area where knock occurs, providing data support for cylinder cooling channel optimization, hydrogen injection timing adjustment, and combustion chamber structure improvement, meeting the needs of directional knock control in hydrogen internal combustion engines.

[0111] Example 5 A fifth embodiment of the present invention provides a readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the steps of the method described in any of the above embodiments.

[0112] Example 6 A sixth embodiment of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method described in any of the above embodiments.

[0113] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0114] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A method for early warning and location of knocking in a hydrogen internal combustion engine, characterized in that, The method includes: At least three acoustic emission sensors are arranged on the cylinder head and cylinder block of the hydrogen internal combustion engine to form an acoustic emission sensor array to collect ultrasonic stress wave signals generated during hydrogen combustion. The output signals of each acoustic emission sensor are sampled at high speed and synchronously, and the sampled data within the corresponding crankshaft angle window is extracted according to the engine working cycle; The intercepted signals from each channel are subjected to bandpass filtering to extract high-frequency ultrasonic features and calculate the knock index. The knock risk level of the current cylinder is determined based on the knock index. Real-time engine block temperature parameters are acquired, sound velocity parameters are dynamically corrected based on the material sound velocity-temperature function, and the spatial coordinates of the knock source in the cylinder are obtained by using the time difference of arrival algorithm. When the knock index exceeds a preset threshold, a knock warning signal and the location information of the knock source are output to the engine ECU, triggering the engine ECU to adjust the ignition timing and hydrogen injection strategy.

2. The method for early warning and location of knocking in a hydrogen internal combustion engine according to claim 1, characterized in that, The acoustic emission sensor array is configured with four non-coplanar acoustic emission sensors, forming a non-coplanar tetrahedral array. The specific steps for arranging the four acoustic emission sensors include: A spark plug-side sensor is arranged on the top of the cylinder head, close to the spark plug base, as a time reference channel for the acoustic emission sensor array. An intake valve side sensor is arranged on the outside of the intake valve passage in the cylinder head, and the sound distance between the intake valve side sensor and a potential ignition source is controlled to be no more than 30mm. An exhaust valve side sensor is placed in the cylinder head bridge area directly above the exhaust valve, corresponding to the high-incidence area of ​​spontaneous combustion of the end mixture; A cylinder wall side sensor is arranged in the middle section of the cylinder block bushing, and the distance between the cylinder wall side sensor and the other three acoustic emission sensors in the Z-axis direction is controlled to be no less than 40mm. The geometric precision factor (GDOP) of the non-coplanar tetrahedral array is controlled to be less than 3.

3. The method for early warning and location of knocking in a hydrogen internal combustion engine according to claim 2, characterized in that, The steps of performing high-speed synchronous sampling, cyclic truncation, bandpass filtering, and calculating the detonation index on the output signals of each acoustic emission sensor specifically include: The same sampling clock is used to synchronously sample the signals of each channel, the sampling frequency is not less than 1MHz, and the phase error between multiple channels is not greater than 0.5μs; A second-order Butterworth bandpass filter is used to filter the sampled signal. The filter passband range is 20kHz-500kHz, and the corresponding time domain signal is output. The detonation index is constructed based on the energy gradient of the filtered high-frequency ultrasonic signal.

4. The method for early warning and location of knocking in a hydrogen internal combustion engine according to claim 3, characterized in that, In the step of inverting the spatial coordinates of the detonation source using the aforementioned time difference of arrival algorithm, the calculation of the time difference of arrival specifically includes: Using the spark plug-side sensor as a reference channel, the filtered discrete signals of each channel are captured within a preset crankshaft angle window; Fast Fourier transform is performed on the discrete signals of the reference channel and the other channels respectively to obtain the corresponding frequency domain signals; Construct the phase-weighted cross-power spectrum between each channel and the reference channel, and perform an inverse Fourier transform on the phase-weighted cross-power spectrum to obtain the generalized cross-correlation function; The time delay corresponding to the maximum peak value of the generalized cross-correlation function is taken as the arrival time difference of the corresponding channel relative to the reference channel; Calculate the peak confidence level of each channel. When the peak confidence level is lower than a preset threshold or the arrival time difference exceeds the maximum propagation delay range limited by the engine geometry and material sound speed, the corresponding channel is marked as a low confidence channel, its weight is reduced, or it is removed.

5. The method for early warning and location of knocking in a hydrogen internal combustion engine according to claim 1, characterized in that, The step of using the time difference of arrival algorithm to invert and obtain the spatial coordinates of the knock source within the cylinder specifically includes: Based on the elastic modulus and density of the cylinder material, the corrected sound velocity at the current temperature is calculated according to the sound velocity-temperature function of the material; Using the spatial coordinates of the detonation source as unknowns and the known installation coordinates of each acoustic emission sensor as geometric constraints, multiple sets of hyperboloid constraint equations are established based on the corrected sound velocity and the arrival time difference, forming an overdetermined set of equations. The target function for positioning is established by using the peak confidence scores of each channel as weights, and the weighted least squares iterative solution is used to obtain the estimated location of the detonation source in the current working cycle. Calculate the positioning residual for the current loop. When the positioning residual exceeds a preset threshold, mark the position estimation result of the loop as a low-confidence result.

6. The method for early warning and location of knocking in a hydrogen internal combustion engine according to claim 5, characterized in that, After obtaining the estimated detonation source location for the current work cycle, multi-period sliding window clustering is also performed, specifically including: The detonation source locations estimated from the most recent M effective working cycles are used to form a sliding window, where M ranges from 5 to 10. Based on the cylinder geometry and positioning error, the neighborhood radius and minimum number of neighborhood points are set. Within the sliding window, a density clustering method based on Euclidean distance threshold is used to remove isolated noise points and select the main clusters. The weighted centroid of the main cluster is calculated using the peak confidence level or the detonation index corresponding to each point within the main cluster as weights. The weighted centroid is output as the final knock position and mapped to the corresponding geometric region of the cylinder; If there is no main cluster within the sliding window that meets the point requirement, then a location confidence deficiency flag is output.

7. The method for early warning and location of knocking in a hydrogen internal combustion engine according to claim 6, characterized in that, The steps for triggering the engine ECU to adjust the ignition timing and hydrogen injection strategy specifically include: Based on the cylinder region where the knock source is located, an ignition angle delay command for the corresponding cylinder is sent to the engine ECU. Alternatively, it can send direct injection pressure and hydrogen injection timing adjustment commands to the engine ECU to change the air-fuel mixture equivalence ratio in the knocking region.

8. A knock warning and positioning system for a hydrogen internal combustion engine, characterized in that, The system, applicable to the method of any one of claims 1-7, comprises: The signal acquisition module is used to arrange at least three acoustic emission sensors on the cylinder head and cylinder block of the hydrogen internal combustion engine to form an acoustic emission sensor array, and to acquire the ultrasonic stress wave signal generated during hydrogen combustion. The synchronous sampling module is used to perform high-speed synchronous sampling of the output signals of each of the acoustic emission sensors, and to extract the sampling data within the corresponding crankshaft angle window according to the engine working cycle; The signal processing module is used to perform bandpass filtering on the intercepted signals from each channel, extract high-frequency ultrasonic features and calculate the knock index, and determine the knock risk level of the current cylinder based on the knock index. The positioning and calculation module is used to acquire the engine cylinder block temperature parameters in real time, dynamically correct the sound velocity parameters based on the material sound velocity-temperature function, and use the arrival time difference algorithm to invert the spatial coordinates of the knock source in the cylinder. The closed-loop control module is used to output a knock warning signal and the location information of the knock source to the engine ECU when the knock index exceeds a preset threshold, thereby triggering the engine ECU to adjust the ignition timing and hydrogen injection strategy.

9. A readable storage medium having computer instructions stored thereon, characterized in that, When executed by a processor, the computer instructions implement the steps of the method as described in any one of claims 1-7.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method as described in any one of claims 1-7.

Images