A large nickel-copper alloy forging surface defect rotating detection system

Abstract

The present application relates to large-scale forging surface defect rotating detection technical field, specifically to a kind of large-scale nickel-copper alloy forging surface defect rotating detection system, including main detection unit, for output sensor coordinate system under surface point vector;Dynamic pose perception unit, for measuring the relative pose of forging surface relative to the sensor base;Space-time data fusion unit, for real-time estimation of sensor base real pose, and the real pose of forging relative to base;Dynamic reference reconstruction engine, for output the surface point vector, real-time calculation the unique coordinate of the surface point vector in the local model coordinate system of forging;Generate distortionless global surface model;System state feedback unit, for real-time calculation dynamic reference jitter index;The system state feedback unit, actively send speed command to rotating drive system;The present application solves the technical conflict between severe mechanical vibration and micron-level detection accuracy.

Description

Technical Field

[0001] This invention relates to the field of rotary detection technology for surface defects in large forgings, specifically a rotary detection system for surface defects in large nickel-copper alloy forgings. Background Technology

[0002] In the field of surface defect inspection of large forgings, especially for large asymmetric forgings, high-speed rotational inspection is a necessary means to improve efficiency.

[0003] However, asymmetric forgings inevitably generate severe mechanical vibrations during rotation due to unbalanced mass, with vibration amplitudes reaching the millimeter or even centimeter level. At the same time, the defect detection task itself requires micrometer-level measurement accuracy. This severe macroscopic mechanical vibration and the micrometer-level detection accuracy requirement create a sharp technical conflict. Vibration causes severe contamination of the measurement reference, making it difficult for traditional detection methods to accurately register data in dynamic environments and achieve distortion-free global surface model reconstruction. Therefore, how to achieve high-precision and high-efficiency surface defect detection under strong mechanical vibration is a key technical problem that urgently needs to be solved in this field. Summary of the Invention

[0004] To solve the above-mentioned technical problems, the present invention provides a rotary detection system for surface defects in large nickel-copper alloy forgings. Specifically, the technical solution of the present invention includes:

[0005] The main detection unit is used to collect microscopic defects and morphology data of the surface of the rotating forging and output the surface point vector in the sensor coordinate system.

[0006] The dynamic pose sensing unit is used to synchronously measure the dynamic pose of the sensor base in the world coordinate system and to measure the relative pose of the forging surface with respect to the sensor base.

[0007] The spatiotemporal data fusion unit is used to fuse the dynamic pose and the relative pose measured by the dynamic pose sensing unit, and combine them with a preset rotational dynamics model to estimate the real pose of the sensor base and the real pose of the forging relative to the base in real time.

[0008] The dynamic reference reconstruction engine is used to combine the true pose of the sensor base estimated by the spatiotemporal data fusion unit and the true pose of the forging relative to the base, as well as the surface point vector output by the main detection unit, and calculate the unique coordinates of the surface point vector in the local model coordinate system of the forging in real time based on a preset dynamic coordinate transformation model; the dynamic reference reconstruction engine is also used to generate a distortion-free global surface model based on the unique coordinates.

[0009] The system state feedback unit is used to calculate the dynamic reference jitter index in real time based on the state vector estimated by the spatiotemporal data fusion unit; the system state feedback unit is also used to actively send speed adjustment commands to the rotation drive system based on the comparison relationship between the dynamic reference jitter index and the preset resonance threshold.

[0010] Preferably, the dynamic pose sensing unit includes:

[0011] An inertial measurement unit is rigidly mounted on the base of the main detection unit and is used to measure the inertial motion of the base in the world coordinate system in real time to generate base posture measurement values.

[0012] A non-contact displacement sensor array is also mounted on the base to measure the relative distance between the base and multiple fixed points on the surface of the forging in a non-contact manner, so as to calculate the relative pose measurement value.

[0013] Preferably, the spatiotemporal data fusion unit is specifically used for:

[0014] A filter model based on state estimation is adopted;

[0015] The base pose measurement value generated by the inertial measurement unit and the relative pose measurement value calculated by the non-contact displacement sensor array are used together as the observation value;

[0016] Based on the preset rotational dynamics model as the state transition function, the system state vector is predicted;

[0017] The system state vector is updated based on the observed values ​​to estimate the true pose of the sensor base and the true pose of the forging relative to the base in real time.

[0018] Preferably, the state transition function is specifically used for:

[0019] The acceleration term in the system state vector is modeled as a nonlinear function related to the current rotation angle and rotation speed to improve the state prediction accuracy of the filter during the sensor data update interval.

[0020] Preferably, the surface point vector output by the main detection unit further includes a hardware timestamp; the spatiotemporal data fusion unit is further used for:

[0021] In response to the hardware timestamp;

[0022] Using the prediction step of the filter model, the true pose of the sensor base, which is strictly aligned with the hardware timestamp, and the true pose of the forging relative to the base are calculated and output at high frequency to resolve the differences in data sampling rate and hardware latency between the main detection unit and the dynamic pose sensing unit.

[0023] Preferably, the dynamic benchmark reconstruction engine calculates the unique coordinates, specifically for:

[0024] By combining the true pose of the sensor base with the preset sensor static calibration matrix, the first true position of the surface point vector in the world coordinate system is calculated;

[0025] By combining the true pose of the sensor base with the true pose of the forging relative to the base, the dynamic coordinate transformation matrix from the local model coordinate system to the world coordinate system of the forging is calculated.

[0026] Based on the first true position and the inverse of the dynamic coordinate transformation matrix, the unique coordinates of the surface point vector in the local model coordinate system of the forging are calculated.

[0027] Preferably, the system state feedback unit calculates the dynamic reference jitter index, specifically for:

[0028] The linear acceleration component and the angular acceleration component of the base are extracted from the state vector estimated by the spatiotemporal data fusion unit.

[0029] Based on the preset forging characteristic radius, the angular acceleration component of the base is normalized in dimension and converted into an equivalent tangential acceleration.

[0030] Based on preset weighting coefficients, the linear acceleration component of the base and the equivalent tangential acceleration are weighted and summed to generate the dynamic reference jitter index.

[0031] Preferably, the system status feedback unit issues the speed regulation command, specifically for:

[0032] Real-time monitoring of the dynamic benchmark jitter index;

[0033] When the dynamic reference jitter index exceeds the preset resonance threshold, a speed reduction or speed adjustment command is sent to the rotary drive system so that the system operating speed jumps out of the mechanical resonance frequency point.

[0034] When the dynamic reference jitter index does not exceed the preset resonance threshold, the current operating state of the rotation drive system is maintained.

[0035] Preferably, the non-contact displacement sensor array includes redundantly arranged displacement sensors;

[0036] The dynamic pose sensing unit is also used for:

[0037] A random sampling consensus algorithm is used to automatically remove sensor data that is temporarily invalidated due to surface reflection or oil contamination in the displacement sensor, so as to maintain the robustness of the relative pose measurement calculation.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] 1. This system constructs a dynamic sensing and active compensation architecture; it no longer attempts to physically suppress vibration, but instead resolves the technical conflict between severe mechanical vibration and micron-level detection accuracy by actively measuring vibration and correcting defect data in real time.

[0040] 2. This system precisely decouples the dual vibrations of the sensor base and the forging target by combining an inertial measurement unit and a non-contact displacement sensor; combined with a dynamic benchmark reconstruction engine, it can calculate unique coordinates from contaminated data in real time and generate a distortion-free global surface model.

[0041] 3. This system actively monitors the instability trend of the system by calculating the dynamic benchmark jitter index in real time; when the jitter exceeds the resonance threshold, it can automatically issue a speed adjustment command to make the system jump out of the dangerous resonance frequency point, thus ensuring equipment safety and the continuity of detection.

[0042] 4. This system employs redundant displacement sensors and a random sampling consistency algorithm; this enables the system to maintain measurement robustness even when some sensors temporarily fail due to interference such as oil stains and reflections common in industrial environments, thus avoiding system downtime and improving reliability. Attached Figure Description

[0043] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0044] Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0046] Example 1:

[0047] Please see Figure 1 A rotating inspection system for surface defects in large nickel-copper alloy forgings, comprising:

[0048] The main detection unit is used to collect microscopic defects and morphology data of the surface of the rotating forging and output the surface point vector in the sensor coordinate system.

[0049] The dynamic pose sensing unit is used to synchronously measure the dynamic pose of the sensor base in the world coordinate system and to measure the relative pose of the forging surface with respect to the sensor base.

[0050] The spatiotemporal data fusion unit is used to fuse the dynamic pose and the relative pose measured by the dynamic pose sensing unit, and combine them with a preset rotational dynamics model to estimate the real pose of the sensor base and the real pose of the forging relative to the base in real time.

[0051] The dynamic reference reconstruction engine is used to combine the true pose of the sensor base estimated by the spatiotemporal data fusion unit and the true pose of the forging relative to the base, as well as the surface point vector output by the main detection unit, and calculate the unique coordinates of the surface point vector in the local model coordinate system of the forging in real time based on a preset dynamic coordinate transformation model; the dynamic reference reconstruction engine is also used to generate a distortion-free global surface model based on the unique coordinates.

[0052] The system state feedback unit is used to calculate the dynamic reference jitter index in real time based on the state vector estimated by the spatiotemporal data fusion unit; the system state feedback unit is also used to actively send speed adjustment commands to the rotation drive system based on the comparison relationship between the dynamic reference jitter index and the preset resonance threshold.

[0053] This embodiment provides a rotating detection system for surface defects in large nickel-copper alloy forgings. The system addresses the problem of severe mechanical vibration caused by unbalanced mass during the rotation of large asymmetric forgings by constructing a fully coupled architecture of dynamic sensing, active compensation, and real-time reconstruction. Instead of attempting to physically suppress or algorithmically tolerate vibration, the system actively measures vibration and uses the vibration information as part of a dynamic coordinate system to correct the spatial registration of defect data in real time.

[0054] In this embodiment, the system includes five core units:

[0055] The main detection unit aims to acquire microscopic data of the forging surface at high speed. In this embodiment, it is a high-resolution 3D line laser profilometer used to acquire microscopic defects and morphological data of the rotating forging surface and output surface point vectors. ; the surface point vector This refers to the position of the sensor in its own coordinate system. The three-dimensional coordinates of the surface points measured at any time are derived from direct measurements by the sensor;

[0056] Dynamic pose sensing unit: Its purpose is to decouple and measure the dual vibrations in the system—namely, the vibration of the observer sensor base itself and the vibration of the target forging relative to the observer; in this embodiment, the unit uses a hybrid sensing array to simultaneously measure the dynamic pose of the sensor base in the world coordinate system, hereinafter referred to as the base pose. And used to measure the relative pose of the forging surface with respect to the sensor base (hereinafter referred to as relative pose). ;

[0057] The spatiotemporal data fusion unit aims to fuse multiple asynchronous, multi-rate, and noisy pose measurements from the dynamic pose sensing unit into a unified, smooth, high-frequency true pose estimate that is strictly aligned with the timestamp of the main detection unit. In this embodiment, this unit employs a state estimation-based filter to fuse the dynamic pose and relative pose measurements, and combines them with a preset rotational dynamics model to estimate the true pose of the sensor base in real time. and the true position of the forging relative to the base ;

[0058] Dynamic baseline reconstruction engine: Its purpose is to utilize the fused true pose to reconstruct the vector of each surface point acquired by the main detection unit. A real-time coordinate transformation is performed, converting the coordinates from the vibrating sensor coordinate system to the stationary local coordinate system of the forging. In this embodiment, the engine incorporates the actual pose of the sensor base estimated by the spatiotemporal data fusion unit. The true position of the forging relative to the base and the surface point vector output by the main detection unit. Based on a preset dynamic coordinate transformation model (see Embodiment 6 for details), the unique coordinates of the surface point vector in the local model coordinate system of the forging are calculated in real time. The engine is also used to generate a distortion-free global surface model based on the aggregated unique coordinates.

[0059] The system state feedback unit aims to achieve closed-loop feedback from the measurement domain to the control domain, actively avoiding system instability. In this embodiment, this unit relies on the state vector estimated by the spatiotemporal data fusion unit. In particular, the acceleration component is used to calculate the dynamic reference jitter index in real time. This unit is also used to base the dynamic reference jitter index on... With preset resonance threshold Based on the comparison relationship, the system actively sends speed adjustment commands to the rotary drive system to avoid the system operating at the mechanical resonance frequency point;

[0060] Through the collaborative work of the above five units, this invention constructs a complete technical closed loop of dynamic measurement, dynamic fusion, dynamic reconstruction, and dynamic feedback; it solves the sharp technical conflict between millimeter-level or even centimeter-level mechanical vibration and micrometer-level detection accuracy in high-speed rotation detection of large asymmetric forgings; this solution no longer relies on an idealized stable benchmark, but achieves micrometer-level defect reconstruction accuracy and industrial-grade high-speed detection efficiency simultaneously under strong vibration background through active measurement and mathematical compensation.

[0061] Example 2:

[0062] The dynamic pose sensing unit includes:

[0063] An inertial measurement unit is rigidly mounted on the base of the main detection unit and is used to measure the inertial motion of the base in the world coordinate system in real time to generate base posture measurement values.

[0064] A non-contact displacement sensor array is also mounted on the base to measure the relative distance between the base and multiple fixed points on the surface of the forging in a non-contact manner, so as to calculate the relative pose measurement value.

[0065] In this embodiment, the dynamic pose sensing unit described in Embodiment 1 is further specified to realize a base-target dual vibration decoupling model. This unit specifically includes an inertial measurement unit (IMU) and a non-contact displacement sensor array. The purpose of the inertial measurement unit is to measure the motion of the observer, i.e., the sensor base itself. In this embodiment, it is a six-axis IMU, rigidly mounted on the base support of the main detection unit, such as a line laser profilometer. It is used to measure the inertial motion, i.e., acceleration of the base in the world coordinate system in real time. and angular velocity After integration using a standard inertial navigation algorithm and incorporating gravity compensation, the base station attitude measurement values ​​are generated. ; This refers to the original measurement of the base's posture, which has the advantages of good real-time performance and high frequency, but the disadvantage of cumulative drift. The purpose of the non-contact displacement sensor array is to measure the relative motion of the target, i.e., the forging, relative to the observer, i.e., the base. In this embodiment, the array consists of at least three, for example, four or five, laser displacement sensors to achieve redundancy, which are also mounted on the base. They are used to non-contactly measure multiple at least three fixed points on the surface of the base and the forging. relative distance between By solving the system of geometric constraint equations, for example ,in The known position of the sensor on the base is used to calculate the relative pose measurement value. ; This refers to the raw measurement of relative pose. Its advantage is that there is no cumulative drift, but its disadvantages are that the noise is relatively large and the sampling rate may be lower than that of IMU.

[0066] This specific combination of IMU and displacement sensor array represents a non-obvious differential-inertial decoupling design; it anticipates and solves the problem of contamination of the measurement reference caused by the vibration of the base itself in traditional measurements; and it measures the inertial motion of the observer's base through the IMU. Then, combined with the displacement sensor array to measure the relative motion of the target forging relative to the observer, This scheme yielded two independent measurements with perfectly complementary characteristics: high frequency but drift vs. low frequency but accuracy. This combination provides the core input data required for the subsequent spatiotemporal data fusion unit to completely decouple the dual vibrations of the system.

[0067] Example 3:

[0068] The spatiotemporal data fusion unit is specifically used for:

[0069] A filter model based on state estimation is adopted;

[0070] The base pose measurement value generated by the inertial measurement unit and the relative pose measurement value calculated by the non-contact displacement sensor array are used together as the observation value;

[0071] Based on the preset rotational dynamics model as the state transition function, the system state vector is predicted;

[0072] The system state vector is updated based on the observed values ​​to estimate the true pose of the sensor base and the true pose of the forging relative to the base in real time.

[0073] Based on the sensor configuration of Embodiment 2, the spatiotemporal data fusion unit in this embodiment is specifically used to perform the fusion of asynchronous, multi-rate, and multi-source data; its core is to adopt a filter model based on state estimation, which is preferably an extended Kalman filter (EKF) in this embodiment.

[0074] The EKF's operating logic is as follows: the system will use the base pose measurement values ​​generated by the inertial measurement unit. High frequency, with drift, and the relative pose measurement value calculated by the non-contact displacement sensor array. Low frequency and no drift are the observations required for the EKF update step. Meanwhile, EKF uses the preset rotational dynamics model as the state transition function. For the system state vector Make predictions. It refers to a high-dimensional vector that contains the system's... All dynamic characteristics at any given time, such as the base's position, velocity, acceleration, and 18 state variables including angle, angular velocity, and angular acceleration; and based on the observed values. For the state vector Update and correct;

[0075] Through a prediction-update cycle, the EKF model is able to estimate the optimal state vector in real time. And from this, the smooth, high-frequency, drift-free true pose of the sensor base, synchronized with the main detector at 10kHz, can be extracted. and the true position of the forging relative to the base ;

[0076] Instead of simple linear interpolation or independent filters, an EKF model based on state estimation is employed. This design intelligently fuses sensor data from two IMUs with complementary noise characteristics—drift and white noise from displacement gauges—utilizing a state transition function. The physical model constraints enable estimation accuracy far superior to traditional filtering; it solves the problem of huge differences in sampling rate, delay and noise characteristics of multi-source sensor data, and provides high-quality, spatiotemporally consistent pose truth values ​​for subsequent dynamic reconstruction.

[0077] Example 4:

[0078] The state transition function is specifically used for:

[0079] The acceleration term in the system state vector is modeled as a nonlinear function related to the current rotation angle and rotation speed to improve the state prediction accuracy of the filter during the sensor data update interval.

[0080] To further improve the accuracy of EKF in the sensor data update gap in Example 3, this example modifies the state transition function. Specific optimizations were performed;

[0081] Traditional state transition functions are typically based on standard kinematic models, such as uniform or uniformly accelerated models. These models assume that motion is general and do not understand the source of vibration. However, in the specific scenario of this invention, the main vibration of the system originates from the rotation of the unbalanced mass, and its vibration characteristics, such as acceleration versus rotation angle, are significant. and rotational speed Strong correlation;

[0082] Therefore, the state transition function in this embodiment Specifically used for:

[0083] The system state vector acceleration term in Modeled as relative to the current rotation angle and speed Related nonlinear functions; for example, acceleration terms. It is no longer viewed as a random walk, but is modeled as a function with physical meaning, such as... ;in, and These represent the unbalanced mass of the forging and its centroid radius, respectively. These represent the total mass of the forging; they are all preset physical model parameters of the system. and The current rotational speed and angle are derived from encoder input.

[0084] This is an innovative model based on physical constraints; it embeds prior knowledge of rotational dynamics into the state transition function. The EKF prediction process is no longer a blind kinematic calculation, but a prediction based on profound physical insights; this allows the EKF to make predictions between two observations. For example, the system state can be accurately calculated even during the intervals between updates from a 1kHz displacement sensor; this high-precision state prediction interpolation capability is crucial for providing synchronized pose for a 10kHz master detection unit.

[0085] Example 5:

[0086] The surface point vector output by the main detection unit also includes a hardware timestamp;

[0087] The spatiotemporal data fusion unit is also used for:

[0088] In response to the hardware timestamp;

[0089] Using the prediction step of the filter model, the true pose of the sensor base, which is strictly aligned with the hardware timestamp, and the true pose of the forging relative to the base are calculated and output at high frequency to resolve the differences in data sampling rate and hardware latency between the main detection unit and the dynamic pose sensing unit.

[0090] To address the spatiotemporal alignment problem of multi-rate data in the system described in Embodiment 3, this embodiment introduces a high-precision time synchronization mechanism;

[0091] One specific technical challenge that this solution aims to address is that the sampling period of the main detection unit, such as a 10kHz IMU or displacement meter, is 0.1ms, which is much smaller than the sampling period of the dynamic pose sensing unit, such as a 1kHz IMU or displacement meter, and different sensors have different hardware delays. Any tiny misalignment of timestamps, such as using the pose 1ms ago to compensate for a measurement point 0.1ms ago, will lead to the complete failure of micron-level reconstruction.

[0092] The surface point vector output by the main detection unit It also includes a high-precision hardware timestamp. The timestamp originates from a unified system clock source, such as the IEEE 1588-based PTP precise time protocol. The measurements from the dynamic pose sensing unit also carry this unified timestamp. The spatiotemporal data fusion unit, i.e., the EKF, responds to the hardware timestamp sent by the main detection unit. Furthermore, the fusion unit utilizes the prediction step of the EKF filter model as described in Example 4, which has high physical fidelity, from the previous state... High-frequency calculation to At any given time, and output the timestamp from the hardware. Strictly aligned sensor base true pose and the true position of the forging relative to the base ;

[0093] By combining hardware timestamp (PTP) and EKF state prediction, this solution addresses the fundamental differences between data sampling rate and hardware latency; the system does not use the most recent measurement pose, but instead uses the current measurement point... Precise timestamp The predicted pose and This ensures that each frame of micron-level data is compensated using its precise pose at the moment of occurrence, achieving sub-millisecond spatiotemporal data alignment.

[0094] Example 6:

[0095] The dynamic baseline reconstruction engine calculates the unique coordinates, specifically for:

[0096] By combining the true pose of the sensor base with the preset sensor static calibration matrix, the first true position of the surface point vector in the world coordinate system is calculated;

[0097] By combining the true pose of the sensor base with the true pose of the forging relative to the base, the dynamic coordinate transformation matrix from the local model coordinate system to the world coordinate system of the forging is calculated.

[0098] Based on the first true position and the inverse of the dynamic coordinate transformation matrix, the unique coordinates of the surface point vector in the local model coordinate system of the forging are calculated.

[0099] This embodiment details how the dynamic benchmark reconstruction engine calculates the unique coordinates in Embodiment 1. The core mathematical model; this model is a fully dynamic chain coordinate transformation, which completely abandons the rigid transformation that depends on the position of the ideal encoder;

[0100] Objective: To calculate the points measured by the sensor. Coordinates in the forging's own, static, local model coordinate system V-coordinate system ;

[0101] in, Surface point vector, measured value in sensor coordinate system S; Source: Main detection unit

[0102] The true pose of the sensor base, i.e., the transformation matrix from the base coordinate system B to the world coordinate system W. Source: Spatiotemporal Data Fusion Unit

[0103] The true pose of the forging relative to the base, i.e., the transformation matrix from the forging's local coordinate system V to the base's coordinate system B. Source: Spatiotemporal Data Fusion Unit

[0104] The sensor static calibration matrix, i.e., the static transformation from the sensor coordinate system S to the base coordinate system B, is obtained through offline calibration; Source: System preset parameters.

[0105] The unique coordinates to be solved are the coordinates of the point in the local coordinate system V of the forging.

[0106] Derivation process:

[0107] In this embodiment, the dynamic reference reconstruction engine is combined with the true pose of the sensor base. With sensor static calibration matrix Calculate the surface point vector The first true position in the world coordinate system W ;

[0108] Derivation chain A measurement chain:

[0109] Substitute parameters:

[0110] At the same time, the engine combines the true pose of the sensor base. The true position of the forging relative to the base Calculate the dynamic coordinate transformation matrix from the local model coordinate system V to the world coordinate system W of the forging. ;

[0111] Derivation chain B model chain:

[0112] Substitute parameters:

[0113] point The position in the world coordinate system W is also... :

[0114]

[0115] The engine is based on the first real location. and dynamic coordinate transformation matrix The inverse matrix is ​​used to solve for the unique coordinates. ;

[0116] Solve the equations A and B together:

[0117] Solve Core formula:

[0118]

[0119] The core formula of this dynamic coordinate transformation model is the core mathematical model of this invention; it is the first mathematically rigorous coupling of all dynamic components of the base vibration. Relative vibration of workpieces Sensor measurement Using this formula, the system can extract data from raw data that has been severely contaminated by vibration. In this process, all vibration effects are stripped away algebraically and in real time, and a unique, static value is calculated. Coordinates are used to achieve distortion-free global model reconstruction.

[0120] Example 7:

[0121] The system status feedback unit calculates the dynamic reference jitter index, specifically for:

[0122] The linear acceleration component and the angular acceleration component of the base are extracted from the state vector estimated by the spatiotemporal data fusion unit.

[0123] Based on the preset forging characteristic radius, the angular acceleration component of the base is normalized in dimension and converted into an equivalent tangential acceleration.

[0124] Based on preset weighting coefficients, the linear acceleration component of the base and the equivalent tangential acceleration are weighted and summed to generate the dynamic reference jitter index.

[0125] This embodiment details the calculation of the dynamic reference jitter index by the system state feedback unit in Embodiment 1. Methods;

[0126] Objective: To define a single scalar with a clear physical meaning for real-time quantification of the instability of the current measurement reference, i.e., the sensor base, especially high-frequency jitter and the tendency to approach resonance; acceleration, rather than displacement, is the most dangerous physical quantity during resonance.

[0127] This unit obtains the state vector estimated in real time by the spatiotemporal data fusion unit EKF. In the process, the linear acceleration components of the base are extracted. and base angular acceleration components ;

[0128] A three-dimensional vector The unit is m / s²; source: calculated by the spatiotemporal data fusion unit module 3.

[0129] A three-dimensional vector The unit is rad / s²; source: calculated by the spatiotemporal data fusion unit module 3.

[0130] Due to the different dimensions, the system is based on the preset characteristic radius of the forging. angular acceleration components of the base Perform dimensional normalization;

[0131] : A scalar, such as the maximum radius of gyration of a forging, in meters; its source is: pre-extracted from the CAD design model of the forging, and used as input for system initialization parameters;

[0132] Normalization calculation: converting angular acceleration into equivalent tangential acceleration The unit of this term becomes rad / s²⋅m, which is physically equivalent to m / s², and... The dimensions are consistent;

[0133] Based on preset weighting coefficients and For the linear acceleration components of the base With the equivalent tangential acceleration A weighted summation is performed to generate a dynamic baseline jitter index. ;

[0134] Dimensionless weighting coefficients are used to balance the contributions of linear and angular acceleration to the total vibration index; their source is: calibration performed through offline experiments under different vibration conditions to optimize... Sensitivity to resonance trends; calculation formulas, for example, weighted sums of L2 norms:

[0135]

[0136] in The L2 norm represents the magnitude of a vector. The final unit is m / s², which is a physically defined equivalent total acceleration.

[0137] An original definition of a jitter index with a definite physical dimension of m / s² was established. This index, through dimensional normalization, scientifically integrates the combined contributions of linear acceleration and angular acceleration to system instability, making it a more sensitive and accurate resonance trend sentinel parameter than monitoring displacement or velocity alone.

[0138] Example 8:

[0139] The system status feedback unit issues the speed regulation command, specifically for:

[0140] Real-time monitoring of the dynamic benchmark jitter index;

[0141] When the dynamic reference jitter index exceeds the preset resonance threshold, a speed reduction or speed adjustment command is sent to the rotary drive system so that the system operating speed jumps out of the mechanical resonance frequency point.

[0142] When the dynamic reference jitter index does not exceed the preset resonance threshold, the current operating state of the rotation drive system is maintained.

[0143] Based on the dynamic benchmark jitter index calculated in Example 7 In this embodiment, the system state feedback unit executes adaptive resonance avoidance closed-loop control logic, and the specific way it issues speed regulation commands is as follows:

[0144] This unit is set with a preset resonance threshold. ; This refers to the maximum permissible vibration index of the system, measured in m / s². Its value is a pre-set safety value based on the mechanical rigidity limit of the equipment or offline vibration table simulation analysis; the system monitors in real time. The value;

[0145] When the dynamic benchmark jitter index Exceeding the preset resonance threshold When this occurs, it indicates that the system is approaching or in a dangerous mechanical resonance zone, at which point both data quality and equipment safety are threatened; the system status feedback unit immediately sends a speed reduction or speed adjustment command to the rotary drive system, such as the turntable servo motor controller, G05D; for example, the command... in It is a speed reduction factor, such as 5%, to make the system operating speed quickly jump out of the current mechanical resonance frequency point;

[0146] When the dynamic benchmark jitter index The preset resonance threshold was not exceeded. When the system is in a safe and stable vibration range, the system state feedback unit maintains the current operating state of the rotary drive system, allowing it to continue operating at high speed to ensure efficiency.

[0147] A method was implemented to measure the domain G01N, via... The evaluation reaches the control domain G05D, and innovative real-time closed-loop feedback is achieved through speed regulation commands. The synergistic effect of this passive compensation embodiment 6 and active avoidance embodiment 8 (1+1>2) ensures that the system can achieve the highest accuracy through compensation in the non-resonance region, and can ensure that the system does not become unstable when approaching the resonance region through speed regulation. This enables the system to operate intelligently at the highest permissible safe speed, achieving a balance between efficiency and robustness.

[0148] Example 9:

[0149] The non-contact displacement sensor array includes redundantly configured displacement sensors;

[0150] The dynamic pose sensing unit is also used for:

[0151] A random sampling consensus algorithm is used to automatically remove sensor data that is temporarily invalidated due to surface reflection or oil contamination in the displacement sensor, so as to maintain the robustness of the relative pose measurement calculation.

[0152] To improve the robustness of the dynamic pose sensing unit in Embodiment 2 under harsh industrial environments, this embodiment incorporates redundant design.

[0153] One specific technical challenge this solution aims to address is that non-contact displacement sensors, such as laser displacement gauges, are easily affected by factors like surface reflections from forgings, oil stains, and water mist in industrial settings, leading to temporary data loss or jumps. Furthermore, if only a mathematically minimum of three sensors are used, the temporary failure of any one sensor will affect the relative pose measurement value. The solution crashed;

[0154] The non-contact displacement sensor array includes redundantly configured displacement sensors; for example, it may be configured with four or five laser displacement sensors instead of the minimum of three.

[0155] The geometric solver of the dynamic pose sensing unit is also used to employ the Random Sample Consensus (RANSAC) algorithm. When one or more sensor data in the displacement sensor temporarily fail due to surface reflection or oil contamination, the RANSAC algorithm will automatically remove the failed sensor data, i.e., outliers, during the iterative solution process.

[0156] This algorithm uses only a subset of consistent interior point sensors to compute relative pose measurements. This ensures the robustness of the relative pose measurement calculation.

[0157] By combining hardware redundancy (>3 sensors) with robust RANSAC algorithms, the anti-interference capability of the dynamic pose sensing unit is greatly enhanced. The system can tolerate the instantaneous failure of one or more sensors without service interruption, avoiding catastrophic downtime caused by local oil contamination or glare, and significantly improving the reliability and availability of the system in real industrial environments.

[0158] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A large nickel-copper alloy forging surface defect rotating detection system, characterized in that, The system comprises: a main detection unit configured to collect micro-defect and topography data of a rotating forging surface and output a surface point vector in a sensor coordinate system; a dynamic pose perception unit configured to synchronously measure a dynamic pose of a sensor base in a world coordinate system and measure a relative pose of the forging surface relative to the sensor base; a space-time data fusion unit configured to fuse the dynamic pose and the relative pose measured by the dynamic pose perception unit, combine a preset rotation dynamics model, and estimate a real pose of the sensor base and a real pose of the forging relative to the base in real time; a dynamic reference reconstruction engine configured to combine the real pose of the sensor base and the real pose of the forging relative to the base estimated by the space-time data fusion unit, the surface point vector output by the main detection unit, and a preset dynamic coordinate transformation model, and calculate unique coordinates of the surface point vector in a forging local model coordinate system in real time; the dynamic reference reconstruction engine is further configured to generate a global surface model without distortion based on the unique coordinates; a system state feedback unit configured to calculate a dynamic reference jitter index in real time according to a state vector estimated by the space-time data fusion unit; the system state feedback unit is further configured to actively send a speed regulation instruction to a rotating drive system based on a comparison relationship between the dynamic reference jitter index and a preset resonance threshold value; the dynamic pose perception unit comprises: an inertial measurement unit rigidly mounted on a base of the main detection unit, configured to measure inertial motion of the base in the world coordinate system in real time to generate a base pose measurement value; a non-contact displacement sensor array also mounted on the base, configured to non-contactly measure relative distances between the base and a plurality of fixed points on the forging surface to solve a relative pose measurement value; the dynamic reference reconstruction engine calculates the unique coordinates, specifically for: combining the real pose of the sensor base and a preset sensor static calibration matrix to calculate a first real position of the surface point vector in the world coordinate system; combining the real pose of the sensor base and the real pose of the forging relative to the base to calculate a dynamic coordinate transformation matrix from the forging local model coordinate system to the world coordinate system; based on the first real position and an inverse matrix of the dynamic coordinate transformation matrix, solving the unique coordinates of the surface point vector in the forging local model coordinate system; the system state feedback unit calculates the dynamic reference jitter index, specifically for: extracting a base linear acceleration component and a base angular acceleration component from the state vector estimated by the space-time data fusion unit; based on a preset forging feature radius, performing dimensionless normalization on the base angular acceleration component to convert it into an equivalent tangential linear acceleration; based on a preset weight coefficient, performing weighted summation on the base linear acceleration component and the equivalent tangential linear acceleration to generate the dynamic reference jitter index.

2. The surface defect rotating detection system for large nickel-copper alloy forgings according to claim 1, characterized in that, The space-time data fusion unit is specifically configured to: adopt a filter model based on state estimation. The base pose measurement generated by the inertial measurement unit and the relative pose measurement calculated by the non-contact displacement sensor array are taken as observation values together; The system state vector is predicted based on the preset rotation dynamics model as a state transition function; The system state vector is updated based on the observation values to estimate the real pose of the sensor base and the real pose of the forging relative to the base in real time.

3. The surface defect rotating detection system for large nickel-copper alloy forgings according to claim 2, characterized in that, The state transition function is specifically used for: Modeling the acceleration term in the system state vector as a nonlinear function related to the current rotation angle and rotation speed to improve the state prediction accuracy of the filter during the sensor data update gap.

4. The surface defect rotating detection system for large nickel-copper alloy forgings according to claim 2, characterized in that, The surface point vector output by the main detection unit also includes a hardware timestamp; The space-time data fusion unit is also used for: Responding to the hardware timestamp; Using the prediction step of the filter model to calculate and output the real pose of the sensor base and the real pose of the forging relative to the base strictly aligned with the hardware timestamp to solve the difference in data sampling rate and hardware delay between the main detection unit and the dynamic pose perception unit.

5. The surface defect rotating detection system for large nickel-copper alloy forgings according to claim 1, characterized in that, The system state feedback unit sends the speed regulation instruction, which is specifically used for: Monitoring the dynamic reference jitter index in real time; When the dynamic reference jitter index exceeds the preset resonance threshold, send a speed reduction or regulation instruction to the rotation driving system to make the system operating speed jump out of the mechanical resonance frequency point; When the dynamic reference jitter index does not exceed the preset resonance threshold, maintain the current rotation driving system operating state.

6. The surface defect rotating detection system for large nickel-copper alloy forgings according to claim 1, characterized in that, The non-contact displacement sensor array includes redundantly arranged displacement sensors; The dynamic pose perception unit is also used for: Using the random sampling consensus algorithm to automatically exclude the failed sensor data when the sensor data in the displacement sensor is temporarily invalid due to surface reflection or oil pollution to maintain the robustness of the relative pose measurement calculation.

Images