A knock sensor signal optimization method in complex environment
By obtaining the cylinder block layout data and sensor layout network of multi-cylinder engines, signal optimization and matching are performed, combined with engine operating conditions and feature extraction, the problem of low signal recognition accuracy in complex environments of multi-cylinder engines is solved, and a higher signal is achieved Identify accuracy.
Patent Information
- Application Number
- CN202411448276.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-17
- Publication Date
- 2024-12-20
- Estimated Expiration
- 2044-10-17
AI Technical Summary
In the complex environment of multi-cylinder engines, traditional knock sensor signal optimization methods are difficult to meet complex needs, resulting in reduced signal recognition accuracy.
By acquiring the cylinder block layout data and sensor layout network of multi-cylinder engines, signal optimization scheme mining and aggregation is carried out, signal matching and optimization is performed in combination with engine operating conditions, signal acquisition is used for single-cylinder and twin-cylinder sensors, and signal extraction is used for feature extraction and abnormal identification The network performs signal recognition.
It improves the accuracy and accuracy of knock signals to meet the signal detection requirements in complex environments of multi-cylinder engines.
Smart Images
Figure CN118966028B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of sensor technology, and in particular to a knock sensor signal optimization method in a complex environment. Background Art
[0002] With the continuous development of engine technology, multi-cylinder engines have been widely used in the fields of automobiles, ships, etc. However, the complex structure and diversified operating conditions of multi-cylinder engines make the signal detection and optimization of knock sensors face many challenges. In a complex operating environment, the layout position of the engine cylinder, the sensor layout network, and the variability of the engine operating conditions will affect the acquisition and processing of the knock signal, resulting in a decrease in the accuracy of knock signal recognition. In addition, traditional knock signal optimization methods usually only optimize a single operating condition or a single cylinder sensor, which is difficult to meet the complex needs of multi-cylinder engines. Therefore, how to improve the detection accuracy of knock signals by optimizing the knock sensor layout network and signal processing process in the complex environment of multi-cylinder engines has become a problem that needs to be solved urgently. Summary of the invention
[0003] The present application provides a knock sensor signal optimization method in a complex environment, which solves the technical problem of low knock sensor signal recognition accuracy in a complex environment.
[0004] In view of the above problems, the present application provides a knock sensor signal optimization method in a complex environment.
[0005] The present application provides a knock sensor signal optimization method in a complex environment, the method comprising:
[0006] Acquire cylinder layout data of a target multi-cylinder engine, wherein the cylinder layout data includes multiple layout positions of multiple engine cylinders and a cylinder distribution network; perform knock sensor layout based on the multiple layout positions and the cylinder distribution network to obtain a knock sensor layout network, wherein the knock sensor layout network includes multiple single-cylinder detection knock sensors and multiple dual-cylinder detection knock sensors; perform signal optimization scheme mining with the cylinder layout data and the knock sensor layout network as indexes to obtain multiple sample signal optimization schemes, wherein the multiple sample signal optimization schemes include multiple sample engine operating conditions; use the engine operating condition as an index, perform similar aggregation on the multiple sample signal optimization schemes in combination with the multiple sample engine operating conditions and perform scheme centralization processing on the aggregation results to obtain multiple aggregated signal optimization schemes, wherein the multiple aggregated signal optimization schemes include multiple Aggregate engine operating condition bandwidth; collect operating parameters of the target multi-cylinder engine in a preset detection window to obtain a first engine operating condition, and match the first engine operating condition with the multiple aggregated engine operating condition bandwidths, and obtain a target signal optimization scheme based on the matching results in combination with the multiple aggregated signal optimization schemes; traverse the multiple single-cylinder detection knock sensors and the multiple dual-cylinder detection knock sensors in the preset detection window to collect signals to obtain multiple single-cylinder signal sets and multiple dual-cylinder signal sets; use the target signal optimization scheme to perform signal optimization on the multiple single-cylinder signal sets and the multiple dual-cylinder signal sets to obtain multiple optimized single-cylinder signal sets and multiple optimized dual-cylinder signal sets; perform knock signal recognition on the multiple optimized single-cylinder signal sets and the multiple optimized dual-cylinder signal sets in combination with the knock sensor layout network to obtain target knock signals and target knock positions.
[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:
[0008] First, the cylinder layout data of the target multi-cylinder engine is obtained, wherein the cylinder layout data includes multiple layout positions and cylinder distribution networks of multiple engine cylinders. Next, knock sensors are laid out based on the multiple layout positions and cylinder distribution networks to obtain a knock sensor layout network, wherein the knock sensor layout network includes multiple single-cylinder detection knock sensors and multiple dual-cylinder detection knock sensors. Next, signal optimization scheme mining is performed with the cylinder layout data and the knock sensor layout network as indexes to obtain multiple sample signal optimization schemes, wherein the multiple sample signal optimization schemes include multiple sample engine operating conditions. Further, with the engine operating condition as an index, multiple sample signal optimization schemes are aggregated into the same category in combination with multiple sample engine operating conditions and the aggregation results are centralized to obtain multiple aggregated signal optimization schemes, wherein the multiple aggregated signal optimization schemes include multiple aggregated engine operating condition bandwidths. The operating parameters of the target multi-cylinder engine are collected in the preset detection window to obtain the first engine operating condition, and the first engine operating condition is matched with multiple aggregated engine operating condition bandwidths. According to the matching results, multiple aggregated signal optimization schemes are combined to obtain the target signal optimization scheme; multiple single-cylinder detection knock sensors and multiple dual-cylinder detection knock sensors are traversed in the preset detection window to collect signals, and multiple single-cylinder signal sets and multiple dual-cylinder signal sets are obtained. Then, the target signal optimization scheme is used to optimize the signals of multiple single-cylinder signal sets and multiple dual-cylinder signal sets, and multiple optimized single-cylinder signal sets and multiple optimized dual-cylinder signal sets are obtained. Finally, the knock signal is identified for multiple optimized single-cylinder signal sets and multiple optimized dual-cylinder signal sets in combination with the knock sensor layout network to obtain the target knock signal and the target knock position. The technical problem of low knock sensor signal recognition accuracy in complex environments is solved, and the technical effect of improving the accuracy of knock signal recognition is achieved. BRIEF DESCRIPTION OF THE DRAWINGS
[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0010] Figure 1 A schematic flow chart of a knock sensor signal optimization method in a complex environment provided by an embodiment of the present application;
[0011] Figure 2 A schematic diagram of a process for knock signal identification in a knock sensor signal optimization method under a complex environment provided in an embodiment of the present application. DETAILED DESCRIPTION
[0012] The present application solves the technical problem of low knock sensor signal recognition accuracy in complex environments by providing a knock sensor signal optimization method in complex environments.
[0013] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.
[0014] It should be noted that the terms "including" and "having" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or are inherent to these processes, methods, products or devices.
[0015] Examples, such as Figure 1 As shown, an embodiment of the present application provides a knock sensor signal optimization method in a complex environment, wherein the method includes:
[0016] The cylinder layout data of the target multi-cylinder engine is obtained, wherein the cylinder layout data includes a plurality of layout positions of a plurality of engine cylinders and a cylinder distribution network.
[0017] The cylinder layout data of the target multi-cylinder engine includes multiple layout positions of multiple engine cylinders and cylinder distribution networks; cylinder layout position refers to the physical position of each cylinder in the engine, including the cylinder installation point, angle, arrangement direction, etc. relative to the engine body; cylinder distribution network refers to the overall distribution structure of the cylinders in the engine and the spatial relationship between them. The network describes how multiple cylinders are related, arranged, and connected to other components (such as crankshafts, cylinder heads, combustion chambers, etc.). By obtaining cylinder layout position and cylinder distribution network data, the physical structure layout of the engine can be fully understood, providing data support for subsequent sensor layout, signal acquisition, and optimization.
[0018] Knock sensors are arranged based on the multiple arrangement positions and the cylinder distribution network to obtain a knock sensor arrangement network, wherein the knock sensor arrangement network includes multiple single-cylinder detection knock sensors and multiple dual-cylinder detection knock sensors.
[0019] By analyzing the layout position and distribution network of the cylinder, the optimal layout of the knock sensor is determined, and then the knock sensor is laid out to form a knock sensor layout network. The knock sensor layout network includes multiple single-cylinder detection knock sensors and multiple dual-cylinder detection knock sensors; the single-cylinder detection knock sensor is used to monitor the knock of a specific cylinder, usually directly installed near a single cylinder, and is responsible for collecting the vibration and sound signals of the cylinder; the dual-cylinder detection knock sensor monitors the knock signals of two adjacent cylinders at the same time, and is usually laid between the two cylinders or can cover the position of the two cylinders at the same time.
[0020] Further, knock sensors are arranged based on the multiple arrangement positions and the cylinder distribution network to obtain a knock sensor arrangement network, wherein the knock sensor arrangement network includes multiple single-cylinder detection knock sensors and multiple dual-cylinder detection knock sensors, including:
[0021] Traversing the multiple layout positions to layout single-cylinder detection knock sensors, and obtaining multiple single-cylinder detection knock sensors; performing neighbor identification of multiple engine cylinders based on the cylinder distribution network and the multiple layout positions, and obtaining multiple engine cylinder neighbor combinations, wherein each engine cylinder neighbor combination is a combination of each engine cylinder and the engine cylinder that is closest in position; arranging dual-cylinder detection knock sensors at the middle position of the multiple engine cylinder neighbor combinations, and obtaining multiple single-cylinder detection knock sensors; constructing the knock sensor layout network according to the multiple single-cylinder detection knock sensors and the multiple dual-cylinder detection knock sensors and the corresponding sensor layout positions.
[0022] Specifically, all cylinder layout positions of the target engine are traversed, and a single-cylinder knock detection sensor is deployed for each cylinder position. The layout position of each sensor should be close to the cylinder to ensure that the knock signal of the cylinder can be accurately collected, thereby obtaining multiple single-cylinder knock detection sensors; in the cylinder distribution network, the mutual neighbor relationship of multiple cylinders is identified, that is, by analyzing the spatial distribution of the cylinders, each cylinder is determined to be closest to other cylinders, forming a neighbor combination of multiple cylinders. For example, if the physical distance between two cylinders is the shortest, they can be considered to be a neighbor combination; for each engine cylinder neighbor combination, the middle position of the combination is selected to deploy a dual-cylinder knock detection sensor. The dual-cylinder sensor is placed at the center of the two cylinders and can detect the knock signals of the two cylinders at the same time; after all the single-cylinder and dual-cylinder detection sensors are deployed, a complete knock sensor deployment network is constructed according to their deployment positions. The knock sensor deployment network combines all single-cylinder and dual-cylinder knock detection sensors to ensure that the knock signal monitoring needs of all cylinders are covered. By gradually deploying single-cylinder and dual-cylinder detection sensors and combining the layout position and neighbor recognition of the cylinder body, a complete knock sensor layout network is formed. The knock sensor layout network can effectively capture the knock signals of multi-cylinder engines in complex environments, laying the foundation for improving signal detection accuracy and optimizing processing efficiency.
[0023] Signal optimization scheme mining is performed using the cylinder layout data and the knock sensor layout network as indexes to obtain a plurality of sample signal optimization schemes, wherein the plurality of sample signal optimization schemes include a plurality of sample engine operating conditions.
[0024] Based on the cylinder layout data and the knock sensor layout network, signal optimization schemes are mined from the relevant database to obtain multiple sample signal optimization schemes. Each sample signal optimization scheme contains a specific engine operating condition, reflecting the acquisition of knock signals, sensor response and signal processing strategies under this operating condition. By using the cylinder layout data and the knock sensor layout network as indexes, multiple signal optimization schemes can be effectively mined, covering different engine operating conditions.
[0025] Taking the engine operating condition as an index, the multiple sample signal optimization schemes are aggregated in the same category in combination with the multiple sample engine operating conditions and the aggregation results are centrally processed to obtain multiple aggregate signal optimization schemes, wherein the multiple aggregate signal optimization schemes include multiple aggregate engine operating condition bandwidths.
[0026] Taking the engine operating condition as an index, the sample signal optimization schemes under different operating conditions are analyzed to identify similar or close operating conditions, and the sample signal optimization schemes are clustered using a clustering algorithm (such as K-means clustering, hierarchical clustering, etc.), and similar signal optimization schemes are classified into one category; on the basis of the same type of aggregation, the optimization schemes are centralized, and the most effective signal optimization strategy is retained by removing redundant signal processing steps and parameter settings; after completing the same type of aggregation and centralized processing, multiple aggregated signal optimization schemes are generated, each of which corresponds to an engine operating condition bandwidth. The engine operating condition bandwidth refers to the range of operating condition fluctuations. For example, the bandwidth of a certain operating condition may include operating conditions with a speed of 3000-4000 rpm and a load of 50%-70%.
[0027] Further, taking the engine operating condition as an index, the plurality of sample signal optimization schemes are aggregated in the same category in combination with the plurality of sample engine operating conditions, and the aggregation results are subjected to scheme centralization processing to obtain a plurality of aggregate signal optimization schemes, wherein the plurality of aggregate signal optimization schemes include a plurality of aggregate engine operating condition bandwidths, including:
[0028] A first sample engine operating condition is randomly extracted from the plurality of sample engine operating conditions, and the first sample engine operating condition is added to a first aggregation node storing a twin network identification unit; the plurality of sample engine operating conditions are aggregated in the same category using the first aggregation node, and the sample engine operating conditions whose twin coefficients meet a preset twin coefficient threshold are stored in the first aggregation internal node, and the sample engine operating conditions whose twin coefficients do not meet the preset twin coefficient threshold are stored in the first partition set; a second sample engine operating condition is randomly extracted from the first partition set without replacement, a second aggregation node is constructed, and the first partition set is aggregated in the same category using the second aggregation node to obtain a second aggregation internal node and a second partition set; based on the The second partition set performs multiple similar aggregations until a preset number of divisions is reached to obtain multiple internal aggregation nodes and the Nth partition set, wherein N is the preset number of divisions and N is an integer greater than 1; multiple similar aggregated sample engine condition sets are obtained based on the multiple internal aggregation nodes and the Nth partition set; in combination with a one-to-one mapping relationship between sample engine conditions and sample signal optimization schemes, multiple similar aggregated sample signal optimization scheme sets are obtained based on the multiple similar aggregated sample engine condition sets, wherein the multiple similar aggregated sample signal optimization scheme sets include multiple aggregated engine condition bandwidths; the multiple similar aggregated sample signal optimization scheme sets are centrally processed to obtain the multiple aggregated signal optimization schemes.
[0029] Specifically, an engine operating condition is randomly selected from multiple sample engine operating conditions as the first sample engine operating condition, and the first sample engine operating condition is added to the first aggregation node storing the twin network recognition unit. The first aggregation node is used to cluster similar operating conditions, and the twin network recognition unit is used to calculate the similarity of different operating conditions. The first aggregation node is used to perform cluster analysis on other sample engine operating conditions through the twin network recognition unit, and by calculating the twin coefficient, it is determined which operating conditions have a high similarity with the first sample engine operating condition, and the sample engine operating conditions whose twin coefficients meet the preset threshold are stored in the first aggregation node. The operating conditions are classified into the first aggregation node, and the operating conditions whose twin coefficients do not meet the preset threshold are stored in the first partition set; the second sample engine operating condition is randomly selected from the first partition set without replacement to construct a second aggregation node, and the second aggregation node is used to perform the same type of aggregation on the operating conditions in the first partition set; similarly, the operating conditions that meet the twin coefficient threshold are stored in the second aggregation internal node, and those that do not meet the threshold are placed in the second partition set; based on the second partition set, multiple same type aggregations are continued, and the above steps are repeated until the preset number of partitions N (N is an integer greater than 1) is reached; in this In the first process, each aggregation result is stored in the corresponding internal node of the aggregation, and the working conditions that do not meet the conditions are further divided into the next division set; when the number of divisions reaches the preset value, multiple internal aggregation nodes and the Nth division set can be finally obtained, and these nodes and sets represent multiple sets of similar aggregated sample engine working conditions, reflecting the similarity classification results under different working conditions; each sample engine working condition has a corresponding sample signal optimization plan, and through this one-to-one mapping relationship, the corresponding optimization plan can be associated with the aggregated engine working condition set, so that multiple sets of similar aggregated sample signal optimization plans can be obtained based on the aggregated engine working condition set, and multiple sets of similar aggregated sample signal optimization plans include multiple aggregated engine working condition bandwidths, and the aggregated engine working condition bandwidth refers to the end value set of the working condition in each set; for each set of similar aggregated sample signal optimization plans, the plan is centralized, and the purpose of centralized processing is to remove redundancy, unify the optimization strategy, so that the optimization plan can adapt to more working condition scenarios; after centralized processing, the final multiple aggregated signal optimization plans are obtained, each plan can cover multiple engine working condition bandwidths, ensuring effective signal optimization under different working conditions.
[0030] Furthermore, performing scheme centralization processing on the plurality of similar aggregate sample signal optimization scheme sets to obtain the plurality of aggregate signal optimization schemes includes:
[0031] Based on the plurality of sets of optimization solutions for similar aggregated sample signals, a plurality of centralized processing spaces are constructed, wherein each centralized processing space includes a plurality of particle points, and each particle point corresponds to a optimization solution for similar aggregated sample signals; a plurality of central particle points of the plurality of centralized processing spaces are extracted, and the plurality of central particle points are used as starting points to search in the plurality of centralized processing spaces according to a plurality of centralized step sizes to obtain a plurality of iterative particle points; it is determined whether the concentration density of the plurality of iterative particle points is greater than or equal to the concentration density of the plurality of central particle points, and if so, the plurality of iterative particle points are used as starting points to iterate until a preset number of iterations is met to obtain a plurality of target iterative particle points; if not, the search directions from the plurality of central particle points to the plurality of iterative particle points are added to a disabled direction list, and the plurality of central particle points are continued to be used as starting points to iterate until a preset number of iterations is met to obtain the plurality of target iterative particle points, wherein the disabled direction list includes a disabled number; and the plurality of optimization solutions for similar aggregated sample signals corresponding to the plurality of target iterative particle points are used as the plurality of aggregated signal optimization solutions.
[0032] Specifically, according to a set of optimization solutions for multiple similar aggregated sample signals, multiple centralized processing spaces are constructed, and each centralized processing space represents a signal optimization processing environment in a specific range; in each centralized processing space, multiple particle points are set, and each particle point represents a specific optimization solution for similar aggregated sample signals; multiple central particle points are extracted from each centralized processing space, and these central particle points serve as starting points, representing the key signal optimization solutions in the current centralized processing space, which may be the optimal or near-optimal signal optimization strategy; the central particle point is used as the starting point for iterative search, and the centralized processing space is searched according to the set concentration step size to find multiple iterative particle points; after searching for multiple iterative particle points, their concentration density is calculated, and the concentration density is the degree of aggregation of these particle points, reflecting whether they are close to the optimal solution or the local optimal area; the concentration density of the iterative particle points is compared with the concentration density of the central particle points to determine whether the new particle points have better optimization characteristics. If If the density of iterative particle points is greater than or equal to the central particle point, these particle points are considered to be better signal optimization solutions; if the concentration density of iterative particle points is higher than that of the central particle point, these iterative particle points are used as new starting points and iterations are continued until the preset number of iterations is reached; if the concentration density of iterative particle points is lower than that of the central particle point, the current search direction is considered to be inappropriate. At this time, the search direction is added to the disabled direction list and the number of disabled times is recorded to avoid repeated invalid direction searches; on the basis of the disabled directions, other directions are searched starting from the central particle point until the preset number of iterations is met; after completing the iterative search, multiple target iterative particle points are finally obtained, and the signal optimization solutions corresponding to these particle points have higher optimization performance, reflecting the best or near-optimal signal processing strategy; each target iterative particle point corresponds to a similar aggregate sample signal optimization solution, and the solutions corresponding to these particle points are extracted as the final aggregate signal optimization solution.
[0033] Further, including:
[0034] The multiple similar aggregate sample signal optimization solution sets are enumerated in pairs respectively to obtain multiple similar aggregate sample signal optimization solution group sets; the multiple similar aggregate sample signal optimization solution group sets are identified for similarity using a cosine similarity calculation formula to obtain multiple combined similarity sets; the multiple combined similarity sets are identified for fluctuation variance to obtain multiple fluctuation variances; and the multiple concentrated step sizes are configured based on the multiple fluctuation variances.
[0035] For multiple sets of optimization solutions for similar aggregated sample signals, enumeration processing is performed pair by pair, and the solutions are combined in pairs to form multiple sets of optimization solution groups for similar aggregated sample signals; the cosine similarity calculation formula is used to measure the similarity of each pair of optimization solutions, and multiple combined similarity sets are obtained. The solution groups with similarity values close to 1 indicate that they have high similarity in signal optimization parameters, and values close to 0 indicate large differences; the volatility of similarity is evaluated using fluctuation variance, which reflects the fluctuation of similarity of optimization solution groups for similar aggregated sample signals under different conditions. By calculating the fluctuation variance of each pair of solution groups, multiple fluctuation variance results are obtained. The larger the fluctuation variance, the lower the data similarity, and a larger concentration step size is required at this time. The smaller the fluctuation variance, the higher the data similarity, and a smaller concentration step size is required; according to the size of the fluctuation variance, an appropriate concentration step size is configured to improve the efficiency of centralized processing of the solution.
[0036] The operating parameters of the target multi-cylinder engine are collected in a preset detection window to obtain a first engine operating condition, and the first engine operating condition is matched with the multiple aggregated engine operating condition bandwidths, and a target signal optimization scheme is obtained based on the matching result combined with the multiple aggregated signal optimization schemes.
[0037] A time window is set and data is collected within the time period. In the preset detection window, key operating parameters of the target multi-cylinder engine are collected in real time, including engine speed, load, temperature, air pressure, fuel injection conditions, etc. These parameters constitute the first engine operating condition, that is, the engine operating condition at the current moment; the collected first engine operating condition is matched with multiple aggregated engine operating condition bandwidths to find the aggregated operating condition bandwidth closest to the current operating condition, which is the matching result; based on the matching result, combined with the aggregated signal optimization scheme corresponding to the operating condition bandwidth, the target signal optimization scheme is obtained.
[0038] The plurality of single-cylinder detection knock sensors and the plurality of dual-cylinder detection knock sensors are traversed in the preset detection window to collect signals, and a plurality of single-cylinder signal sets and a plurality of dual-cylinder signal sets are obtained.
[0039] A plurality of single-cylinder detection knock sensors and dual-cylinder detection knock sensors arranged on the engine cylinder block are traversed in a preset detection window to obtain a plurality of single-cylinder signal sets and a plurality of dual-cylinder signal sets.
[0040] The target signal optimization scheme is used to perform signal optimization on the multiple single-cylinder signal sets and the multiple dual-cylinder signal sets to obtain multiple optimized single-cylinder signal sets and multiple optimized dual-cylinder signal sets.
[0041] The target signal optimization scheme includes the best signal processing strategy for the current engine operating state, such as filtering parameters, signal enhancement algorithm, noise suppression, etc.; using the target signal optimization scheme to optimize multiple single-cylinder signal sets and multiple dual-cylinder signal sets can effectively remove noise and enhance useful signals, and finally obtain multiple optimized single-cylinder signal sets and multiple optimized dual-cylinder signal sets.
[0042] The knock sensor layout network is combined to perform knock signal identification on the multiple optimized single-cylinder signal sets and the multiple optimized dual-cylinder signal sets to obtain a target knock signal and a target knock position.
[0043] By combining a knock sensor deployment network and performing knock signal recognition on multiple optimized single-cylinder signal sets and multiple optimized dual-cylinder signal sets, a high-precision target knock signal and knock position can be obtained.
[0044] Furthermore, if Figure 2 As shown, knock signal recognition is performed on the multiple optimized single-cylinder signal sets and the multiple optimized dual-cylinder signal sets in combination with the knock sensor layout network to obtain a target knock signal and a target knock position, including:
[0045] The multiple optimized single-cylinder signal sets and the multiple optimized dual-cylinder signal sets are subjected to time synchronization processing to obtain multiple synchronously optimized single-cylinder signal sets and multiple synchronously optimized dual-cylinder signal sets; feature extraction is performed on the multiple synchronously optimized single-cylinder signal sets and the multiple synchronously optimized dual-cylinder signal sets using a feature extraction network layer to obtain multiple synchronously optimized single-cylinder signal feature sets and multiple synchronously optimized dual-cylinder signal feature sets; abnormality identification is performed on the multiple synchronously optimized single-cylinder signal feature sets using a knock abnormality signal identification network layer to obtain multiple abnormal synchronously optimized single-cylinder signal feature sets; knock signal identification is performed based on the multiple abnormal synchronously optimized single-cylinder signal feature sets and the multiple synchronously optimized dual-cylinder signal feature sets, as well as the knock sensor layout network to obtain the target knock signal and target knock position.
[0046] Specifically, each optimized single-cylinder signal set and dual-cylinder signal set is time-aligned to ensure that the data points in different signal sets can be compared and analyzed on the same time scale; after synchronization processing, multiple synchronously optimized single-cylinder signal sets and multiple synchronously optimized dual-cylinder signal sets are generated to ensure that these sets can simultaneously reflect the changes in the knock signal; using the feature extraction network layer, feature extraction (such as changes in amplitude, frequency, signal waveform, etc.) is performed from multiple synchronously optimized single-cylinder signal sets and multiple synchronously optimized dual-cylinder signal sets to obtain multiple synchronously optimized single-cylinder signal feature sets and multiple synchronously optimized dual-cylinder signal feature sets; knock is an abnormal engine state, which usually exhibits specific abnormal characteristics in the signal. The abnormal signal recognition network layer can accurately identify these abnormal signals and distinguish normal signals from knock signals; using the knock abnormal signal The signal recognition network layer analyzes each synchronously optimized single-cylinder signal feature set, identifies abnormal signals, and forms an abnormal synchronously optimized single-cylinder signal feature set. The knock abnormal signal recognition network layer can detect, for example, a sudden increase in amplitude or abnormal vibration within a specific frequency range. These features usually indicate the occurrence of knock; through the processing of the abnormal recognition network layer, multiple abnormal synchronously optimized single-cylinder signal feature sets are generated, which represent the knock signals that may exist in each single cylinder; based on multiple abnormal synchronously optimized single-cylinder signal feature sets and multiple synchronously optimized dual-cylinder signal feature sets, the knock signal is recognized by using the location and coverage of the knock sensor layout network, the target knock signal is determined, and its frequency, amplitude and occurrence time are determined; the knock sensor layout network is used to accurately locate the occurrence location of the knock by combining the source of the signal and the physical location of the sensor. Through the coordinated work of time synchronization processing, feature extraction network layer and knock abnormal signal recognition network layer, the knock signal can be efficiently and accurately identified from the optimized signal set, and the cylinder or area where the knock occurs can be accurately located.
[0047] Furthermore, the abnormal knock signal recognition network layer is a network layer constructed based on a convolutional neural network and used for engine cylinder knock signal recognition.
[0048] A convolutional neural network is trained based on the engine knock signal dataset to learn the knock signal patterns under various working conditions. During the training process, a supervised learning method is used and the labeled knock signals are used as training samples. The network will continuously adjust the weights of the convolution kernel and the classifier according to the feedback to improve the recognition accuracy. After the training is completed, the abnormal knock signal recognition network layer is obtained.
[0049] Further, including:
[0050] Based on the multiple abnormal synchronously optimized single-cylinder signal feature sets, multiple abnormal knock sensor layout positions are determined; taking the multiple abnormal knock sensor layout positions as the derivation starting point, first-order derivation of dual-cylinder signals is performed in the knock sensor layout network to obtain multiple first-order derived dual-cylinder signal feature sets; respectively determine whether the knock signal strengths of the first-order derived dual-cylinder signal feature sets are all less than or equal to the knock signal strengths of the multiple abnormal synchronously optimized single-cylinder signal feature sets, and if so, use the multiple abnormal synchronously optimized single-cylinder signal feature sets as target knock signals, and add the corresponding multiple abnormal knock sensor layout positions into the target knock positions.
[0051] By analyzing multiple abnormal synchronously optimized single-cylinder signal feature sets, it is possible to determine which sensors' collected signals show knock anomalies; based on the sensor position data in the knock sensor layout network and combined with the abnormal signal characteristics, the sensor position that detects the abnormal signal is determined as the abnormal knock sensor layout position; starting from multiple abnormal knock sensor layout positions, first-order derivative processing of the dual-cylinder signal is performed in the knock sensor layout network. The first-order derivative refers to the expansion from the single-cylinder signal to the adjacent dual-cylinder signal, and its correlation with the abnormal single-cylinder signal is analyzed; through the first-order derivative, multiple first-order derivative dual-cylinder signal feature sets can be obtained. Each dual-cylinder signal feature set corresponds to the knock signal of the adjacent cylinder body, reflecting the linkage effect of the two adjacent cylinder bodies, especially the transmission of the knock signal between the adjacent cylinder bodies. broadcast; compare the knock signal strength of the first-order derived dual-cylinder signal feature set and the knock signal strength of the abnormal synchronously optimized single-cylinder signal feature set respectively; if the knock signal strength of the first-order derived dual-cylinder signal feature set is less than or equal to the corresponding single-cylinder signal strength, it means that the knock signal is mainly concentrated in the single cylinder, and has not expanded or transmitted in the adjacent cylinders; if the first-order derived dual-cylinder signal strength is greater than the single-cylinder signal strength, it indicates that the knock signal may affect multiple cylinders, or be transmitted between adjacent cylinders; when the knock signal strength of the first-order derived dual-cylinder signal feature set is less than or equal to the knock signal strength of the abnormal synchronously optimized single-cylinder signal feature set, the abnormal synchronously optimized single-cylinder signal feature set is confirmed as the target knock signal, and at the same time, the corresponding multiple abnormal knock sensor layout positions are added to the target knock positions.
[0052] In summary, the embodiments of the present application have at least the following technical effects:
[0053] First, the cylinder layout data of the target multi-cylinder engine is obtained, wherein the cylinder layout data includes multiple layout positions and cylinder distribution networks of multiple engine cylinders. Next, knock sensors are laid out based on the multiple layout positions and cylinder distribution networks to obtain a knock sensor layout network, wherein the knock sensor layout network includes multiple single-cylinder detection knock sensors and multiple dual-cylinder detection knock sensors. Next, signal optimization scheme mining is performed with the cylinder layout data and the knock sensor layout network as indexes to obtain multiple sample signal optimization schemes, wherein the multiple sample signal optimization schemes include multiple sample engine operating conditions. Further, with the engine operating condition as an index, multiple sample signal optimization schemes are aggregated into the same category in combination with multiple sample engine operating conditions and the aggregation results are centralized to obtain multiple aggregated signal optimization schemes, wherein the multiple aggregated signal optimization schemes include multiple aggregated engine operating condition bandwidths. The operating parameters of the target multi-cylinder engine are collected in the preset detection window to obtain the first engine operating condition, and the first engine operating condition is matched with multiple aggregated engine operating condition bandwidths. According to the matching results, multiple aggregated signal optimization schemes are combined to obtain the target signal optimization scheme; multiple single-cylinder detection knock sensors and multiple dual-cylinder detection knock sensors are traversed in the preset detection window to collect signals, and multiple single-cylinder signal sets and multiple dual-cylinder signal sets are obtained. Then, the target signal optimization scheme is used to optimize the signals of multiple single-cylinder signal sets and multiple dual-cylinder signal sets, and multiple optimized single-cylinder signal sets and multiple optimized dual-cylinder signal sets are obtained. Finally, the knock signal is identified for multiple optimized single-cylinder signal sets and multiple optimized dual-cylinder signal sets in combination with the knock sensor layout network to obtain the target knock signal and the target knock position. The technical problem of low knock sensor signal recognition accuracy in complex environments is solved, and the technical effect of improving the accuracy of knock signal recognition is achieved.
[0054] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0055] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.
[0056] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.
Claims
1. A knock sensor signal optimization method in a complex environment, characterized in that: The method comprises: Acquire cylinder layout data of a target multi-cylinder engine, wherein the cylinder layout data includes multiple layout positions of multiple engine cylinders and a cylinder distribution network; Arranging knock sensors based on the multiple arrangement positions and the cylinder distribution network to obtain a knock sensor arrangement network, wherein the knock sensor arrangement network includes multiple single-cylinder detection knock sensors and multiple dual-cylinder detection knock sensors; Performing signal optimization scheme mining with the cylinder layout data and the knock sensor layout network as indexes to obtain a plurality of sample signal optimization schemes, wherein the plurality of sample signal optimization schemes include a plurality of sample engine operating conditions; Taking the engine operating condition as an index, combining the multiple sample engine operating conditions, performing similar aggregation on the multiple sample signal optimization solutions and performing solution centralization processing on the aggregation results to obtain multiple aggregated signal optimization solutions, wherein the multiple aggregated signal optimization solutions include multiple aggregated engine operating condition bandwidths; Collecting operating parameters of the target multi-cylinder engine in a preset detection window to obtain a first engine operating condition, matching the first engine operating condition with the multiple aggregated engine operating condition bandwidths, and obtaining a target signal optimization solution based on the matching result and the multiple aggregated signal optimization solutions; Traversing the plurality of single-cylinder detection knock sensors and the plurality of dual-cylinder detection knock sensors in the preset detection window to collect signals, and obtaining a plurality of single-cylinder signal sets and a plurality of dual-cylinder signal sets; Performing signal optimization on the multiple single-cylinder signal sets and the multiple dual-cylinder signal sets using the target signal optimization scheme to obtain multiple optimized single-cylinder signal sets and multiple optimized dual-cylinder signal sets; The knock sensor layout network is combined to perform knock signal identification on the multiple optimized single-cylinder signal sets and the multiple optimized dual-cylinder signal sets to obtain a target knock signal and a target knock position.
2. The knock sensor signal optimization method in a complex environment as claimed in claim 1, characterized in that: The knock sensors are arranged based on the multiple arrangement positions and the cylinder distribution network to obtain a knock sensor arrangement network, wherein the knock sensor arrangement network includes multiple single-cylinder detection knock sensors and multiple dual-cylinder detection knock sensors, including: Traversing the plurality of layout positions to layout the single-cylinder detection knock sensors, and obtaining a plurality of single-cylinder detection knock sensors; Perform neighbor identification of multiple engine cylinders based on the cylinder distribution network and the multiple layout positions to obtain multiple engine cylinder neighbor combinations, wherein each engine cylinder neighbor combination is a combination of each engine cylinder and the engine cylinder that is closest in position; Arranging a dual-cylinder knock sensor at a middle position of the plurality of adjacent combinations of engine cylinders to obtain a plurality of single-cylinder knock sensors; The knock sensor layout network is constructed according to the multiple single-cylinder detection knock sensors and the multiple dual-cylinder detection knock sensors and the corresponding sensor layout positions.
3. The knock sensor signal optimization method in a complex environment as claimed in claim 1, characterized in that: Taking the engine operating condition as an index, the plurality of sample signal optimization schemes are aggregated in the same category in combination with the plurality of sample engine operating conditions, and the aggregation results are subjected to scheme centralization processing to obtain a plurality of aggregate signal optimization schemes, wherein the plurality of aggregate signal optimization schemes include a plurality of aggregate engine operating condition bandwidths, including: Randomly extracting a first sample engine operating condition from the plurality of sample engine operating conditions, and adding the first sample engine operating condition to a first aggregation node storing a twin network identification unit; Using the first aggregation node to aggregate the plurality of sample engine operating conditions into the same category, storing the sample engine operating conditions whose twin coefficients satisfy a preset twin coefficient threshold to the first aggregation internal node, and storing the sample engine operating conditions whose twin coefficients do not satisfy the preset twin coefficient threshold to the first partition set; randomly extracting a second sample engine operating condition from the first partition set without replacement, constructing a second aggregation node, and performing similar aggregation on the first partition set using the second aggregation node to obtain a second aggregation internal node and a second partition set; Performing multiple aggregations of the same type based on the second partition set until a preset number of partitions is reached, to obtain multiple internal aggregation nodes and an Nth partition set, where N is a preset number of partitions and is an integer greater than 1; Obtaining a plurality of similar aggregated sample engine operating condition sets according to the plurality of internal aggregated nodes and the Nth partition set; In combination with a one-to-one mapping relationship between sample engine operating conditions and sample signal optimization solutions, a plurality of similar aggregated sample signal optimization solution sets are obtained according to the plurality of similar aggregated sample engine operating condition sets, wherein the plurality of similar aggregated sample signal optimization solution sets include a plurality of aggregated engine operating condition bandwidths; Centralized processing is performed on the plurality of similar aggregate sample signal optimization solution sets to obtain the plurality of aggregate signal optimization solutions.
4. The knock sensor signal optimization method in a complex environment as claimed in claim 3, characterized in that: Centralizing the schemes for the plurality of similar aggregate sample signal optimization scheme sets to obtain the plurality of aggregate signal optimization schemes includes: Constructing multiple centralized processing spaces based on the multiple sets of optimization solutions for the same type of aggregated sample signals, wherein each centralized processing space includes multiple particle points, and each particle point corresponds to a same type of aggregated sample signal optimization solution; Extracting a plurality of central particle points from the plurality of centralized processing spaces, taking the plurality of central particle points as starting points, searching in the plurality of centralized processing spaces according to a plurality of centralized step lengths, respectively, to obtain a plurality of iterative particle points; Determine whether the concentration density of the multiple iterative particle points is greater than or equal to the concentration density of the multiple central particle points. If so, use the multiple iterative particle points as starting points and perform iterations until a preset number of iterations is met to obtain multiple target iterative particle points. If not, the search directions from the multiple center particle points to the multiple iteration particle points are added to the disabled direction list, and the multiple center particle points are continuously used as the starting point to iterate until the preset number of iterations is met to obtain the multiple target iteration particle points, wherein the disabled direction list includes the disabled number of times; The multiple similar aggregate sample signal optimization schemes corresponding to the multiple target iterative particle points are used as the multiple aggregate signal optimization schemes.
5. The knock sensor signal optimization method in a complex environment as claimed in claim 4, characterized in that: include: Enumerating the plurality of optimization solution sets for the same type of aggregated sample signals in pairs respectively to obtain a plurality of optimization solution group sets for the same type of aggregated sample signals; The cosine similarity calculation formula is used to identify the similarity of multiple sets of optimization schemes for the same type of aggregated sample signals, and multiple combination similarity sets are obtained; Performing fluctuation variance identification on the multiple combination similarity sets to obtain multiple fluctuation variances; The plurality of concentration steps are configured based on the plurality of fluctuation variances.
6. The knock sensor signal optimization method in a complex environment as claimed in claim 1, characterized in that: In combination with the knock sensor layout network, knock signal recognition is performed on the multiple optimized single-cylinder signal sets and the multiple optimized dual-cylinder signal sets to obtain a target knock signal and a target knock position, including: Performing time synchronization processing on the multiple optimized single-cylinder signal sets and the multiple optimized dual-cylinder signal sets to obtain multiple synchronously optimized single-cylinder signal sets and multiple synchronously optimized dual-cylinder signal sets; Using a feature extraction network layer to perform feature extraction on the multiple synchronously optimized single-cylinder signal sets and the multiple synchronously optimized dual-cylinder signal sets, to obtain multiple synchronously optimized single-cylinder signal feature sets and multiple synchronously optimized dual-cylinder signal feature sets; Using a knock abnormality signal recognition network layer to perform abnormality recognition on the multiple synchronously optimized single-cylinder signal feature sets to obtain multiple abnormal synchronously optimized single-cylinder signal feature sets; Based on the multiple abnormal synchronously optimized single-cylinder signal feature sets and the multiple synchronously optimized dual-cylinder signal feature sets, and the knock sensor layout network, knock signal recognition is performed to obtain the target knock signal and target knock position.
7. The knock sensor signal optimization method in a complex environment as claimed in claim 6, characterized in that: include: Determining a plurality of abnormal knock sensor layout positions based on the plurality of abnormal synchronously optimized single-cylinder signal feature sets; Taking the plurality of abnormal knock sensor layout positions as derivation starting points, performing first-order derivation of dual-cylinder signals in the knock sensor layout network to obtain a plurality of first-order derived dual-cylinder signal feature sets; Determine whether the knock signal strengths of the first-order derived dual-cylinder signal feature sets are all less than or equal to the knock signal strengths of the multiple abnormal synchronously optimized single-cylinder signal feature sets. If so, use the multiple abnormal synchronously optimized single-cylinder signal feature sets as target knock signals, and add the corresponding multiple abnormal knock sensor layout positions into the target knock positions.
8. The knock sensor signal optimization method in a complex environment as claimed in claim 6, characterized in that: The abnormal knock signal recognition network layer is a network layer constructed based on a convolutional neural network and used for engine cylinder knock signal recognition.