Intelligent crankshaft vibration layered self-adaptive control method based on bimodal switching architecture

Abstract

The invention provides an intelligent crankshaft vibration layered adaptive control method based on a bimodal switching architecture, and relates to the technical field of engines, and the method comprises the steps: executing instantaneous torque compensation iteration on a brushless DC torque motor in a discrete data direct drive mode; and after the crankshaft system is smoothly switched to a time sequence prediction intervention mode, reading vibration state time sequence data to execute pre-intervention prediction, performing pre-compensation torque fitting by fusing the vibration state time sequence data and ultra-short-term vibration state prediction data, and driving a brushless direct current torque motor to output compensation torque until the crankshaft system is powered off and stops running. The technical problems that in the prior art, online monitoring of the real-time state of a crankshaft system is lacked, passive energy consumption type vibration reduction is conducted by depending on a response lagging mechanical vibration damper, and broadband vibration suppression under the full working condition of an engine cannot be adapted are solved. The technical effect of vibration self-suppression closed-loop intelligent cooperative control from passive energy consumption type vibration reduction to active force type vibration suppression is achieved.

Description

Technical Field

[0001] This invention relates to the field of engine technology, and in particular to a smart crankshaft vibration hierarchical adaptive control method based on a dual-mode switching architecture. Background Technology

[0002] Traditional large engine crankshaft control systems suffer from a lag in response to passive vibration damping mechanisms. They rely on mechanical devices such as silicone oil dampers to passively dissipate vibration energy by converting it into heat energy through viscous fluid shearing. The inherent heat conduction delay in this energy conversion process leads to a significant lag in the system response, which cannot meet the real-time suppression requirements for wideband vibration during dynamic speed and load switching of the engine.

[0003] Existing technologies lack real-time online monitoring of the crankshaft system and rely on mechanical dampers with lag for passive energy-dissipating vibration reduction, resulting in a technical problem that cannot adapt to wide-band vibration suppression under all engine operating conditions. Summary of the Invention

[0004] To address the aforementioned deficiencies or improvement needs of existing technologies, this invention provides an intelligent crankshaft vibration hierarchical adaptive control method based on a dual-mode switching architecture. This method addresses the technical problem that existing technologies lack real-time online monitoring of the crankshaft system's status and rely on passive energy-dissipating vibration reduction using mechanical dampers with lag, resulting in an inability to adapt to wide-band vibration suppression under all engine operating conditions.

[0005] To achieve the above objectives, this invention provides a hierarchical adaptive control method for intelligent crankshaft vibration based on a dual-mode switching architecture, the method comprising:

[0006] After the crankshaft system is powered on, the discrete data direct drive mode is instantaneously triggered. In this mode, the ECU reads the instantaneous vibration data returned by the torsional vibration monitoring sensor array at millisecond intervals and performs instantaneous torque compensation iteration on the brushless DC torque motor integrated at the free end of the crankshaft. After the runtime of the discrete data direct drive mode meets the mode switching time window, the crankshaft system is smoothly switched to the timing prediction intervention mode. In the timing prediction intervention mode, the following steps are performed: Step A: The ECU reads the vibration time-series data returned by the torsional vibration monitoring sensor array at millisecond intervals and performs pre-intervention prediction, outputting ultra-short-term vibration state prediction data; Step B: The vibration time-series data and ultra-short-term vibration state prediction data are fused to perform pre-compensation torque fitting, driving the brushless DC torque motor to output compensation torque; The timing prediction intervention mode continues to run until the crankshaft system is powered off and stops running.

[0007] In one implementation, in the discrete data direct drive mode, after the ECU reads the instantaneous vibration state data returned by the torsional vibration monitoring sensor array at millisecond intervals, it performs instantaneous torque compensation iteration on the brushless DC torque motor integrated at the free end of the crankshaft, and also performs the following processing:

[0008] The ECU reads instantaneous vibration data from the torsional vibration monitoring sensor array at millisecond intervals. This instantaneous vibration data includes instantaneous rotational speed, instantaneous phase, and instantaneous amplitude. The instantaneous vibration data is combined into an instantaneous vibration feature vector. The mean of short-term residual vibration error is locally retrieved, and based on the mean of the short-term residual vibration error and the instantaneous vibration feature vector, the filter weight vector is dynamically updated using the LMS adaptive algorithm, and an updated weight vector is output. An initial compensation torque command is generated based on the updated weight vector and the instantaneous vibration feature vector. This initial compensation torque command includes a torque amplitude and an application phase angle. After driving the brushless DC torque motor to execute the initial compensation torque command for instantaneous torque compensation, the initial residual vibration error is retrieved from the torsional vibration monitoring sensor array. This initial residual vibration error is used as the feedback input for the next control cycle to perform iterative instantaneous torque compensation.

[0009] In one implementation, the following processing is also performed:

[0010] The torsional vibration monitoring sensor array is pre-installed in the crankshaft system and electrically connected to the ECU; wherein the torsional vibration monitoring sensor array includes a magnetoelectric phase sensor and a fiber Bragg grating torque sensor, the magnetoelectric phase sensor is used to collect speed data and phase data, and the fiber Bragg grating torque sensor is used to collect amplitude data.

[0011] In one implementation, the following processing is also performed:

[0012] The ECU is driven to retrieve accumulated vibration state data, wherein the start and end timestamps of the accumulated vibration state data cover the complete operating cycle of the discrete data direct drive mode; the accumulated vibration state data is used to tune the basic state prediction model and construct a state transition prediction model; the vibration state time series data is input into the state transition prediction model to perform pre-intervention prediction and output the ultra-short-term vibration state prediction data.

[0013] In one implementation, the accumulated vibration state data is used to tune the basic state prediction model and construct a state transition prediction model. The following processing is also performed:

[0014] After time-series alignment of the cumulative rotational speed data, cumulative phase data, and cumulative amplitude data in the accumulated vibration state data, data binning is performed to obtain multiple multidimensional data binning intervals for multiple discrete state nodes. Multiple discrete state vectors are constructed based on these multidimensional data binning intervals. P sample discrete state groups are obtained through historical state transition statistics, and a state probability transition matrix is ​​constructed based on these P sample discrete state groups as the basic state prediction model. Adjacent spatiotemporal state transition frequency statistics are performed on the multiple discrete state vectors, and P state transition conditional probability values ​​for the P sample discrete state groups are calculated. The probability matrix parameters of the basic state prediction model are filled using the P state transition conditional probability values ​​to construct the state transition prediction model.

[0015] In one implementation, the vibration state time series data is input into the state transition prediction model to perform pre-intervention prediction, and the ultra-short-term vibration state prediction data is output. The following processing is also performed:

[0016] After converting the vibration state time series data into a real-time state vector, it is input into the state transition prediction model to perform state transition prediction and output a predicted state vector; the predicted state vector is then restored and mapped to output an ultra-short-term predicted speed range, an ultra-short-term predicted phase range, and an ultra-short-term predicted amplitude range; wherein, the ultra-short-term predicted speed range, ultra-short-term predicted phase range, and ultra-short-term predicted amplitude range constitute the ultra-short-term vibration state prediction data.

[0017] In one implementation, the vibration state time-series data and ultra-short-term vibration state prediction data are fused to perform pre-compensation torque fitting, driving the brushless DC torque motor to output compensation torque, and the following processing is also performed:

[0018] A first associated compensation instruction is fitted based on the vibration state time-series data; a second associated compensation instruction is fitted based on the ultra-short-term vibration state prediction data; real-time engine operating conditions are retrieved, and compensation confidence weights are dynamically allocated according to the real-time engine operating conditions; the first associated compensation instruction and the second associated compensation instruction are weighted and fused using the compensation confidence weights, and a pre-compensation torque instruction is output; the brushless DC torque motor is driven to execute the pre-compensation torque instruction to perform instantaneous torque compensation.

[0019] In one implementation, based on the ultra-short-term vibration state prediction data, a second correlation compensation instruction is fitted, and the following processing is also performed:

[0020] The crankshaft vibration interference extreme values ​​are extracted from the ultra-short-term predicted speed range, ultra-short-term predicted phase range, and ultra-short-term predicted amplitude range to obtain the predicted maximum interference speed, predicted maximum interference phase, and predicted maximum interference amplitude; based on the predicted maximum interference speed, predicted maximum interference phase, and predicted maximum interference amplitude, the second correlation compensation command is fitted and output.

[0021] In one implementation, the following processing is also performed:

[0022] The real-time engine operating condition is matched with the real-time operating condition category; the real-time operating condition category is used as the search key to query the switching mapping rule and obtain the mode switching time window.

[0023] One or more technical solutions provided in this invention have at least the following technical effects or advantages:

[0024] The method provided in this embodiment of the invention instantaneously triggers a discrete data direct drive mode after the crankshaft system is powered on and started. In the discrete data direct drive mode, the ECU reads the instantaneous vibration state data returned by the torsional vibration monitoring sensor array at millisecond intervals and performs instantaneous torque compensation iteration on the brushless DC torque motor integrated at the free end of the crankshaft. After the runtime of the discrete data direct drive mode meets the mode switching time window, the crankshaft system is smoothly switched to a timing prediction intervention mode. In the timing prediction intervention mode, the ECU reads the vibration state timing data returned by the torsional vibration monitoring sensor array at millisecond intervals and performs pre-intervention prediction, outputting ultra-short-term vibration state prediction data. The timing data and ultra-short-term vibration state prediction data are fused to perform pre-compensation torque fitting, driving the brushless DC torque motor to output compensation torque. The timing prediction intervention mode continues to run until the crankshaft system is powered off and stops running. By implementing layered adaptive control of crankshaft vibration based on a dual-mode switching architecture, the technical effect of significantly widening the engine's operating speed range, efficiently suppressing abnormal torsional vibration, and realizing the transformation from passive energy-consuming vibration reduction to active force-driven vibration suppression is achieved. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 A schematic diagram of the intelligent crankshaft vibration hierarchical adaptive control method based on a dual-mode switching architecture provided by the present invention is shown.

[0027] Figure 2 The diagram illustrates the process of predicting ultra-short-term vibration state prediction data in the intelligent crankshaft vibration hierarchical adaptive control method based on a dual-mode switching architecture provided by the present invention. Detailed Implementation

[0028] This invention provides an intelligent crankshaft vibration hierarchical adaptive control method based on a dual-mode switching architecture, which addresses the technical problem that existing technologies lack real-time online monitoring of the crankshaft system and rely on mechanical dampers with lag for passive energy-dissipating vibration reduction, resulting in an inability to adapt to wide-frequency vibration suppression under all engine operating conditions.

[0029] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0030] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0031] Unless otherwise expressly stated, throughout the specification and claims, the term "comprising" or its variations such as "including" or "comprises" shall be understood to include the stated elements or components without excluding other elements or other components.

[0032] The flowchart of the intelligent crankshaft vibration hierarchical adaptive control method based on a dual-mode switching architecture provided in this embodiment of the invention is shown below. Figure 1 The method includes:

[0033] During the crankshaft system assembly stage, a torsional vibration monitoring sensor array is pre-installed and electrically connected to the ECU (Electronic Control Unit) to enable the ECU to acquire amplitude, phase, and speed data in real time.

[0034] The torsional vibration monitoring sensor array specifically includes a magnetoelectric phase sensor and a fiber Bragg grating torque sensor. The specific functions of the two sensors are as follows:

[0035] The magnetoelectric phase sensor is specifically responsible for real-time acquisition of crankshaft speed and rotational phase angle data, while the fiber optic grating torque sensor independently undertakes the task of monitoring torsional vibration amplitude, forming a multi-physical quantity collaborative sensing network. The physical characteristics of the acquired speed data, phase data, and amplitude data will be described later in this embodiment.

[0036] P100: After the crankshaft system is powered on, the discrete data direct drive mode is triggered instantaneously.

[0037] It should be understood that in this embodiment, the dual-mode switching architecture specifically includes a discrete data direct-drive mode and a time-series predictive intervention mode. The discrete data direct-drive mode is used to sense the vibration state in real time and perform millisecond-level torque compensation iterations, while the time-series predictive intervention mode is used to predict the vibration evolution trend and perform forward-looking intervention control.

[0038] P200: In the discrete data direct drive mode, after the ECU reads the instantaneous vibration state data returned by the torsional vibration monitoring sensor array in millisecond cycles, it performs instantaneous torque compensation iteration on the brushless DC torque motor integrated at the free end of the crankshaft.

[0039] In one implementation, in the discrete data direct drive mode, after the ECU reads the instantaneous vibration state data returned by the torsional vibration monitoring sensor array at millisecond intervals, it performs instantaneous torque compensation iteration on the brushless DC torque motor integrated at the free end of the crankshaft. The method step P200 provided by the present invention includes:

[0040] P210: The ECU reads the instantaneous vibration state data returned by the torsional vibration monitoring sensor array at millisecond intervals, wherein the instantaneous vibration state data includes instantaneous rotational speed, instantaneous phase, and instantaneous amplitude.

[0041] P220: Combine the instantaneous vibration state data into an instantaneous vibration feature vector.

[0042] P230: Locally call the mean of short-term residual vibration error, and based on the mean of short-term residual vibration error and the instantaneous vibration feature vector, dynamically update the filter weight vector through the LMS adaptive algorithm, and output the updated weight vector.

[0043] P240: Generate an initial compensation torque command based on the updated weight vector and the instantaneous vibration feature vector, wherein the initial compensation torque command includes the torque amplitude and the action phase angle.

[0044] P250: After driving the brushless DC torque motor to execute the initial compensation torque command to perform instantaneous torque compensation, retrieve the initial residual vibration error from the torsional vibration monitoring sensor array.

[0045] P260: The initial residual vibration error is used as the feedback input for the next control cycle to perform instantaneous torque compensation iteration.

[0046] After the crankshaft system is powered on, the discrete data direct drive mode is instantaneously triggered. In this mode, the ECU reads the instantaneous vibration data returned by the torsional vibration monitoring sensor array at millisecond intervals, and then performs instantaneous torque compensation iteration on the brushless DC torque motor integrated at the free end of the crankshaft. The specific implementation is as follows:

[0047] The ECU reads instantaneous vibration data from the torsional vibration monitoring sensor array at fixed millisecond intervals. This data includes the instantaneous crankshaft speed and rotational phase angle collected by the magnetoelectric phase sensor, and the instantaneous torsional vibration amplitude value collected by the fiber optic torque sensor. For example, the speed data is 1200 revolutions per minute, the phase angle is 45 degrees, and the amplitude value is 0.15 millimeters.

[0048] The instantaneous vibration data are combined into a three-dimensional instantaneous vibration feature vector to form a mathematical representation describing the current vibration state, which serves as the core input parameters of the adaptive algorithm. For example, [1200rpm, 45°, 0.15mm] constitutes a complete vibration state sample.

[0049] The ECU retrieves the mean short-term residual vibration error from local memory, which reflects the vibration suppression residual level of the previous control cycle.

[0050] By combining the current instantaneous vibration feature vector and the mean short-term residual vibration error, the filter weight parameters are dynamically adjusted using the LMS adaptive algorithm. The algorithm iteratively updates the weight vector based on the gradient direction of the product of the mean error and the feature vector, thereby improving the compensation accuracy in the next cycle. For example, when the mean residual amplitude is 0.05 mm, the weight vector is updated along the negative gradient direction to enhance the high-frequency vibration suppression capability.

[0051] The updated weight vector and the instantaneous vibration feature vector are used for matrix calculation to generate an initial compensation torque command. This initial compensation torque command includes two key physical parameters: first, the torque amplitude calculated based on the rotational speed and amplitude, to ensure it is equal to the vibration torque amplitude; second, the action phase angle determined based on the phase angle, to ensure a strict 180-degree phase difference with the vibration torque. For example, inputting the feature vector [1500rpm, 60°, 0.2mm] generates the command {amplitude: 80Nm, phase angle: 240°}.

[0052] After the brushless DC torque motor outputs precise electromagnetic torque according to the initial compensation torque command, the ECU immediately retrieves the actual residual vibration error data from the torsional vibration monitoring sensor array, which is the initial residual vibration error.

[0053] The initial residual vibration error specifically includes the amplitude residual measured by the fiber optic grating torque sensor and the phase offset detected by the magnetoelectric phase sensor. For example, after performing 80 Nm compensation, the residual amplitude is 0.03 mm and the phase deviation is 5 degrees.

[0054] The initial residual vibration error is used as a closed-loop feedback signal input to the next millisecond control cycle. It should be understood that the residual vibration error data is updated in each cycle and participates in a new round of LMS weight update calculation, forming a continuously optimized adaptive control closed loop. For example, a 0.03 mm amplitude residual and a 5-degree phase deviation will trigger a fine-tuning of the weight vector, causing the compensation command in the next cycle to approach the theoretical optimal value to achieve gradual zeroing of the vibration.

[0055] In this embodiment, the LSM algorithm is used to perform real-time vibration state perception and closed-loop torque compensation iteration in discrete data direct drive mode, achieving the technical effect of actively suppressing abnormal torsional vibration at the millisecond level.

[0056] P300: After the runtime of the discrete data direct drive mode meets the mode switching time window, smoothly switch the crankshaft system to the timing prediction intervention mode.

[0057] In one implementation, the method provided by the present invention further includes:

[0058] P310: Match the real-time operating condition category according to the real-time engine operating condition.

[0059] P320: Using the real-time operating condition category as the search key, query the switching mapping rules to obtain the mode switching time window.

[0060] Specifically, in this embodiment, the current operating condition category is identified based on the real-time collected engine operating parameters. Continuous quantities such as speed fluctuation range and load change rate are mapped to discrete operating condition labels through preset rules to determine and output the real-time operating condition category.

[0061] The determined real-time operating condition category is used as the index key value to query the switching mapping rule table pre-stored in the ECU and obtain the corresponding mode switching time window.

[0062] This embodiment sets the mode switching time based on real-time operating conditions, achieving a full-condition adaptive matching control mode and ensuring the technical effect of continuous vibration suppression reliability.

[0063] When the continuous running time of the discrete data direct drive mode reaches the time threshold of the mode switching time window, the ECU performs a smooth and non-disruptive switching operation, transitioning the control of the crankshaft system from the discrete data direct drive mode to the timing prediction intervention mode. The switching process is ensured by the phase continuity of the control commands to avoid mechanical shock to the crankshaft system.

[0064] P400: In the time series prediction intervention mode, perform the following steps:

[0065] Step A: The ECU reads the vibration state timing data returned by the torsional vibration monitoring sensor array at millisecond intervals, performs pre-intervention prediction, and outputs ultra-short-term vibration state prediction data.

[0066] Step B: The vibration state time series data and ultra-short-term vibration state prediction data are fused together to perform pre-compensation torque fitting, driving the brushless DC torque motor to output compensation torque.

[0067] The timing prediction intervention mode continues to run until the crankshaft system is powered off and stops operating.

[0068] Specifically, it should be understood that in the timing prediction intervention mode, the ECU periodically executes millisecond-level data acquisition and predictive control processes. The timing prediction intervention mode, as an advanced evolution stage of the discrete data direct drive mode, continues to run until the crankshaft system is powered off and shut down, achieving active vibration suppression throughout the entire life cycle.

[0069] The time-series prediction intervention mode cyclically executes steps A to B. Specifically, in step A, the ECU continuously reads the vibration state time-series data stream returned by the torsional vibration monitoring sensor array at millisecond intervals, and inputs the real-time speed sequence, phase angle sequence, and amplitude sequence into the state transition prediction model to perform pre-intervention prediction. Through the ultra-short-term prediction of the vibration evolution trend by the state transition prediction model, ultra-short-term vibration state prediction data containing speed range, phase range, and amplitude range are output.

[0070] In step B, the current real-time vibration state time series data and ultra-short-term vibration state prediction data are integrated, and a pre-compensation torque command is generated through a dynamic weight allocation strategy. The real-time data provides the current vibration state benchmark, and the prediction data contributes evolution trend information. The two work together to drive the brushless DC torque motor to output precise reverse compensation torque, thereby achieving forward-looking vibration suppression control.

[0071] Repeat steps A through B until the crankshaft system is powered off and stops operating.

[0072] This embodiment achieves the technical effect of significantly widening the engine operating speed range, efficiently suppressing abnormal torsional vibration, and realizing the vibration self-suppression closed-loop intelligent collaborative control from passive energy-consuming vibration reduction to active force-based vibration suppression by using a dual-mode switching architecture for layered adaptive control of crankshaft vibration.

[0073] In one implementation, see Figure 2 The method provided by this invention further includes:

[0074] P411: Drive the ECU to retrieve the accumulated vibration state data, wherein the start and end timestamps of the accumulated vibration state data cover the complete operating cycle of the discrete data direct drive mode.

[0075] P412: The accumulated vibration state data is used to adjust the parameters of the basic state prediction model and construct the state transition prediction model.

[0076] P413: Input the vibration state time series data into the state transition prediction model to perform pre-intervention prediction and output the ultra-short-term vibration state prediction data.

[0077] In one implementation, the accumulated vibration state data is used to tune the basic state prediction model and construct a state transition prediction model. Step P412 of the method provided by this invention further includes:

[0078] P4121: After performing time-series alignment on the cumulative rotational speed data, cumulative phase data, and cumulative amplitude data in the cumulative vibration state data, data binning is performed to obtain multiple multidimensional data binning intervals for multiple discrete state nodes.

[0079] P4122: Construct multiple discrete state vectors based on the multiple multidimensional data binning intervals.

[0080] P4123: P discrete state groups are obtained by performing historical state transition statistics, and a state probability transition matrix is ​​constructed based on the P discrete state groups as the basic state prediction model.

[0081] P4124: Perform adjacent spatiotemporal state transition frequency statistics on the multiple discrete state vectors, and calculate the P state transition conditional probability values ​​of the P sample discrete state groups.

[0082] P4125: The probability matrix parameters of the basic state prediction model are filled using the P state transition conditional probability values ​​to construct the state transition prediction model.

[0083] In one implementation, the vibration state time series data is input into the state transition prediction model to perform pre-intervention prediction, and the ultra-short-term vibration state prediction data is output. Step P413 of the method provided by this invention further includes:

[0084] P4131: After converting the vibration state time series data into a real-time state vector, input it into the state transition prediction model to perform state transition prediction and output the predicted state vector.

[0085] P4132: Restore and map the predicted state vector to output the ultra-short-term predicted speed range, ultra-short-term predicted phase range, and ultra-short-term predicted amplitude range.

[0086] The ultra-short-term predicted rotational speed range, ultra-short-term predicted phase range, and ultra-short-term predicted amplitude range constitute the ultra-short-term vibration state prediction data.

[0087] Specifically, the ECU is driven to retrieve accumulated vibration state data, which fully covers all historical operating cycles of the discrete data direct drive mode from startup to switching, and its timestamp range starts from the moment the direct drive mode is activated and ends at the moment the mode switching is triggered.

[0088] The accumulated rotational speed data, accumulated phase data, and accumulated amplitude data in the accumulated vibration state data are time-series aligned to ensure that the three types of data strictly correspond at the same time. After alignment, data discretization and binning are performed. Based on a preset threshold, the continuous data is divided into multiple discrete intervals to form multidimensional data binning intervals representing different operating state nodes, ultimately resulting in multiple multidimensional data binning intervals for multiple discrete state nodes.

[0089] Based on the divided multidimensional data bin intervals, corresponding discrete state vectors are generated. Each vector uniquely represents a combination of state nodes in a specific speed range, phase range, and amplitude range, thus realizing the discretized mathematical representation of continuous vibration state.

[0090] By analyzing the transition patterns of state nodes in historical data, P discrete state groups with causal relationships are extracted. Each group contains a state vector before transition and a state vector after transition. A state probability transition matrix is ​​constructed based on this sample set. The row vectors of the matrix represent the initial state, the column vectors represent the target state, and the matrix elements represent the transition probabilities. This matrix constitutes the core framework of the basic state prediction model. The constructed state probability transition matrix is ​​used as the basic state prediction model.

[0091] The frequency of state transitions between adjacent time steps of multiple discrete state vectors is statistically analyzed. Specifically, the conditional probability value of the transition from the initial state to the target state in each sample discrete state group is calculated, and finally, P conditional probability values ​​of state transitions corresponding to P groups of samples are obtained.

[0092] The calculated P state transition conditional probability values ​​are filled into the probability transition matrix of the basic state prediction model. Each probability value is written into the intersection of the starting state row and the target state column of the matrix to complete the assignment of probability matrix parameters, thereby constructing a state transition prediction model that can quantify and predict state evolution.

[0093] The real-time collected vibration state time series data is input into the constructed state transition prediction model to perform pre-intervention prediction. The future ultra-short-term vibration state is deduced through the state transition logic inside the model, and finally the ultra-short-term vibration state prediction data containing the speed change range, phase change range and amplitude change range is output.

[0094] Specifically, in this embodiment, the vibration state time series data is converted into a real-time state vector. The vector includes speed interval labels, phase interval labels, and amplitude interval labels. The real-time state vector is input into the state transition prediction model to perform state transition deduction. The model outputs the predicted state vector for future times based on the probability transition matrix.

[0095] Physical quantity restoration mapping is performed on the discrete interval labels contained in the predicted state vector. The speed interval labels are converted into specific speed range values ​​to output the ultra-short-term predicted speed interval, the phase interval labels are converted into angle range values ​​to output the ultra-short-term predicted phase interval, and the amplitude interval labels are converted into amplitude range values ​​to output the ultra-short-term predicted amplitude interval.

[0096] This embodiment achieves the technical effect of effectively and accurately predicting future vibration trends, providing highly reliable reference data for subsequent vibration intervention and suppression analysis.

[0097] In one implementation, the vibration state time-series data and ultra-short-term vibration state prediction data are fused to perform pre-compensation torque fitting, driving the brushless DC torque motor to output compensation torque. Step P400 of the method provided by this invention further includes:

[0098] P421: Fit the first associated compensation command based on the vibration state time series data.

[0099] P422: Fit a second correlation compensation command based on the ultra-short-term vibration state prediction data.

[0100] P423: Retrieve real-time engine operating conditions and dynamically allocate compensation confidence weights based on the real-time engine operating conditions.

[0101] P424: The first associated compensation instruction and the second associated compensation instruction are weighted and fused using the compensation confidence weight to output the pre-compensation torque instruction.

[0102] P425: Drive the brushless DC torque motor to execute the pre-compensation torque command for instantaneous torque compensation.

[0103] In one implementation, based on the ultra-short-term vibration state prediction data, a second associated compensation instruction is fitted. Step P422 of the method provided by this invention further includes:

[0104] P4221: Extract the extreme values ​​of crankshaft vibration interference from the ultra-short-term predicted speed range, ultra-short-term predicted phase range, and ultra-short-term predicted amplitude range to obtain the predicted maximum interference speed, predicted maximum interference phase, and predicted maximum interference amplitude.

[0105] P4222: Based on the predicted maximum interference speed, predicted maximum interference phase, and predicted maximum interference amplitude, the second correlation compensation command is fitted and output.

[0106] Specifically, mathematical correlation analysis is performed on the instantaneous values ​​of rotational speed, phase, and amplitude contained in the real-time collected vibration state time-series data to generate the first correlation compensation command that reflects the vibration suppression requirements at the current moment.

[0107] It should be understood that in this embodiment, any torque command such as the initial compensation torque command and the first associated compensation command is generated in real time through the LMS algorithm described in step P200 above.

[0108] Based on the ultra-short-term vibration state prediction data output by the prediction module, including the rotational speed variation range, phase variation range, and amplitude variation range, a forward-looking analysis is performed to generate a second associated compensation command for future vibration suppression needs. Specifically:

[0109] In one implementation, based on the ultra-short-term vibration state prediction data, a second associated compensation instruction is fitted. Step P422 of the method provided by this invention further includes:

[0110] Extreme value analysis is performed on the ultra-short-term predicted speed range to extract the predicted maximum disturbance speed value, extreme value analysis is performed on the ultra-short-term predicted phase range to extract the predicted maximum disturbance phase value, and extreme value analysis is performed on the ultra-short-term predicted amplitude range to extract the predicted maximum disturbance amplitude value.

[0111] Based on the extracted predicted maximum interference speed value, predicted maximum interference phase value, and predicted maximum interference amplitude value, the second correlation compensation command is fitted and output through the LMS algorithm described in step P200 above.

[0112] The engine's current operating condition parameters are retrieved. Based on the system dynamic characteristics represented by these operating condition parameters, the weight allocation ratio of the first associated compensation command and the second associated compensation command in the comprehensive decision-making is determined in real time, and the compensation confidence weight is output. In this embodiment, the numerical mapping relationship between the operating condition and the compensation confidence weight is not strictly limited and can be set according to the user's actual application.

[0113] The first associated compensation command and the second associated compensation command are weighted and fused using the proportional parameter of the compensation confidence weight to output the final pre-compensation torque command acting on the actuator. This command includes information on the compensation torque amplitude and phase angle.

[0114] The brushless DC torque motor is driven to output instantaneous compensation torque to the free end of the crankshaft according to the amplitude and phase angle requirements of the pre-compensation torque command.

[0115] This embodiment achieves the technical effect of significantly improving vibration suppression response speed and adaptability to operating conditions.

[0116] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A hierarchical adaptive control method for intelligent crankshaft vibration based on a dual-mode switching architecture, characterized in that, include: After the crankshaft system is powered on and started, the discrete data direct drive mode is triggered instantaneously; In the discrete data direct drive mode, after the ECU reads the instantaneous vibration state data returned by the torsional vibration monitoring sensor array in millisecond cycles, it performs instantaneous torque compensation iteration on the brushless DC torque motor integrated on the free end of the crankshaft. After the runtime of the discrete data direct drive mode meets the mode switching time window, the crankshaft system is smoothly switched to the timing prediction intervention mode. In the time-series prediction intervention mode, the following steps are performed: Step A: The ECU reads the vibration state timing data returned by the torsional vibration monitoring sensor array at millisecond intervals, performs pre-intervention prediction, and outputs ultra-short-term vibration state prediction data; Step B: Integrate the vibration state time series data and ultra-short-term vibration state prediction data to perform pre-compensation torque fitting, and drive the brushless DC torque motor to output compensation torque; The timing prediction intervention mode continues to run until the crankshaft system is powered off and stops operating.

2. The intelligent crankshaft vibration hierarchical adaptive control method based on a dual-mode switching architecture as described in claim 1, characterized in that, In the discrete data direct drive mode, after the ECU reads the instantaneous vibration state data returned by the torsional vibration monitoring sensor array at millisecond intervals, it performs instantaneous torque compensation iteration on the brushless DC torque motor integrated at the free end of the crankshaft, including: The ECU reads the instantaneous vibration state data returned by the torsional vibration monitoring sensor array at millisecond intervals, wherein the instantaneous vibration state data includes instantaneous rotational speed, instantaneous phase, and instantaneous amplitude; The instantaneous vibration state data are combined into an instantaneous vibration feature vector; The mean of short-term residual vibration error is called locally, and the filter weight vector is dynamically updated through the LMS adaptive algorithm based on the mean of short-term residual vibration error and the instantaneous vibration feature vector, and the updated weight vector is output. An initial compensation torque command is generated based on the updated weight vector and the instantaneous vibration feature vector, wherein the initial compensation torque command includes the torque amplitude and the action phase angle; After driving the brushless DC torque motor to execute the initial compensation torque command to perform instantaneous torque compensation, the initial residual vibration error is retrieved from the torsional vibration monitoring sensor array; The initial residual vibration error is used as the feedback input for the next control cycle to perform instantaneous torque compensation iteration.

3. The intelligent crankshaft vibration hierarchical adaptive control method based on a dual-mode switching architecture as described in claim 1, characterized in that, Also includes: The torsional vibration monitoring sensor array is pre-installed in the crankshaft system and electrically connected to the ECU; The torsional vibration monitoring sensor array includes a magnetoelectric phase sensor and a fiber Bragg grating torque sensor. The magnetoelectric phase sensor is used to collect rotational speed data and phase data, and the fiber Bragg grating torque sensor is used to collect amplitude data.

4. The intelligent crankshaft vibration hierarchical adaptive control method based on a dual-mode switching architecture as described in claim 1, characterized in that, Also includes: The ECU is driven to retrieve accumulated vibration state data, wherein the start and end timestamps of the accumulated vibration state data cover the complete operating cycle of the discrete data direct drive mode; The accumulated vibration state data is used to tune the parameters of the basic state prediction model and construct a state transition prediction model. The vibration state time series data is input into the state transition prediction model to perform pre-intervention prediction, and the ultra-short-term vibration state prediction data is output.

5. The intelligent crankshaft vibration hierarchical adaptive control method based on a dual-mode switching architecture as described in claim 4, characterized in that, The accumulated vibration state data is used to tune the basic state prediction model and construct a state transition prediction model, including: After aligning the cumulative rotational speed data, cumulative phase data, and cumulative amplitude data in the cumulative vibration state data according to time sequence, data binning is performed to obtain multiple multidimensional data binning intervals for multiple discrete state nodes. Multiple discrete state vectors are constructed based on the multiple multidimensional data binning intervals; P discrete state groups are obtained by performing historical state transition statistics, and a state probability transition matrix is ​​constructed based on the P discrete state groups as the basic state prediction model. Perform adjacent spatiotemporal state transition frequency statistics on the multiple discrete state vectors, and calculate the P state transition conditional probability values ​​of the P sample discrete state groups; The state transition prediction model is constructed by filling the probability matrix parameters of the basic state prediction model with the P state transition conditional probability values.

6. The intelligent crankshaft vibration hierarchical adaptive control method based on a dual-mode switching architecture as described in claim 4, characterized in that, The vibration state time series data is input into the state transition prediction model to perform pre-intervention prediction, and the ultra-short-term vibration state prediction data is output, including: After converting the vibration state time series data into a real-time state vector, it is input into the state transition prediction model to perform state transition prediction and output the predicted state vector. The predicted state vector is restored and mapped to output the ultra-short-term predicted speed range, ultra-short-term predicted phase range, and ultra-short-term predicted amplitude range. The ultra-short-term predicted rotational speed range, ultra-short-term predicted phase range, and ultra-short-term predicted amplitude range constitute the ultra-short-term vibration state prediction data.

7. The intelligent crankshaft vibration hierarchical adaptive control method based on a dual-mode switching architecture as described in claim 6, characterized in that, The vibration state time-series data and ultra-short-term vibration state prediction data are fused to perform pre-compensation torque fitting, driving the brushless DC torque motor to output compensation torque, including: A first associated compensation instruction is fitted based on the vibration state time series data; A second correlation compensation command is fitted based on the ultra-short-term vibration state prediction data; Retrieve real-time engine operating conditions and dynamically allocate compensation confidence weights based on the real-time engine operating conditions; The first associated compensation instruction and the second associated compensation instruction are weighted and fused using the compensation confidence weight to output the pre-compensation torque instruction; The brushless DC torque motor is driven to execute the pre-compensation torque command to perform instantaneous torque compensation.

8. The intelligent crankshaft vibration hierarchical adaptive control method based on a dual-mode switching architecture as described in claim 7, characterized in that, Fitting a second correlation compensation instruction based on the ultra-short-term vibration state prediction data includes: The crankshaft vibration interference extreme values ​​are extracted from the ultra-short-term predicted speed range, ultra-short-term predicted phase range, and ultra-short-term predicted amplitude range to obtain the predicted maximum interference speed, predicted maximum interference phase, and predicted maximum interference amplitude. Based on the predicted maximum interference speed, predicted maximum interference phase, and predicted maximum interference amplitude, the second correlation compensation command is fitted and output.

9. The intelligent crankshaft vibration hierarchical adaptive control method based on a dual-mode switching architecture as described in claim 7, characterized in that, Also includes: Match the real-time operating condition category according to the real-time engine operating condition; Using the real-time operating condition category as the search key, query the switching mapping rules to obtain the mode switching time window.

Images