CN-121185214-B - Spectral confocal follow-up detection method based on autoregressive feedforward prediction compensation

CN121185214BCN 121185214 BCN121185214 BCN 121185214BCN-121185214-B

Abstract

The invention discloses a spectral confocal follow-up detection method based on autoregressive feedforward prediction compensation, which comprises the steps of building a three-dimensional measurement system, adjusting a spectral confocal sensor to a working range, calculating a sample height value in a current detection area, storing a plurality of historical height value data, calculating deviation between current sensor reading and the sample height value in real time, predicting future height change by using an autoregressive model, generating feedforward compensation control quantity according to the predicted height, and adding the feedforward compensation control quantity into a control signal to generate a final control signal. By using the invention, accurate dynamic follow-up detection is realized. The invention can be widely applied to the field of precision detection.

Inventors

ZHANG LANYU
GUO DONGTAO
ZHENG ZHUOJUN
GAO JIAN

Assignees

广东工业大学

Dates

Publication Date: 20260505
Application Date: 20250909

Claims (4)

1. The spectral confocal follow-up detection method based on autoregressive feedforward prediction compensation is characterized by comprising the following steps of: Acquiring height data and constructing a historical height data set; setting a constant height and making a difference with the height data to obtain a height deviation; predicting based on the historical height data set and an autoregressive model to obtain a predicted height; Generating a feedforward predictive compensation control amount based on the predicted altitude and generating a final control output signal; the step of acquiring height data and constructing a historical height data set specifically comprises the following steps: Acquiring height data based on the sensor; acquiring an absolute position value fed back by the high-precision grating ruler; adding the height data and the absolute position value to obtain the height of the sample; Recording the actual heights of the surfaces of a plurality of samples to obtain a historical height data set; The expression of the final control output signal is as follows: Wherein, the Is the current height difference; PID proportion; Is an integral coefficient; Is a differential coefficient; compensating the control quantity for feedforward prediction; The autoregressive model is represented as follows: Wherein, the The regression coefficient is represented as a function of the regression coefficient, For the error term, i represents the sequence number of the autoregressive term, Representing the order of the autoregressive model, Representing the estimated height of the predicted time t +1, Representing the true measured height at time t-i + 1.
2. The method for spectral confocal follow-up detection based on auto-regressive feedforward predictive compensation according to claim 1, further comprising: three-dimensional point cloud data is generated.
3. The method for spectral confocal follow-up detection based on auto-regressive feedforward prediction compensation according to claim 1, wherein said predicting based on said historical height data set and auto-regressive model to obtain a predicted height comprises the steps of: Defining an expression of an autoregressive model; Determining regression coefficients of the autoregressive model by using a least square method with the aim of minimizing the sum of squares of errors based on the expression and the historical height data set; and predicting based on the autoregressive model after determining the regression coefficient to obtain the predicted height.
4. The method according to claim 1, wherein the step of generating a feedforward predictive compensation control amount based on the predicted height and generating a final control output signal comprises: Generating a feedforward prediction compensation control quantity according to the prediction height and the feedforward gain coefficient; and introducing the feedforward predictive compensation control quantity as a feedforward signal, constructing a composite controller and generating a final control output signal.

Description

Spectral confocal follow-up detection method based on autoregressive feedforward prediction compensation Technical Field The invention relates to the field of precision detection, in particular to a spectral confocal follow-up detection method based on autoregressive feedforward prediction compensation. Background With the requirements of micro-nano manufacturing, wafer detection, optical surface profile measurement and other fields on three-dimensional detection precision, stability and scanning efficiency, the spectral confocal sensor is widely applied to three-dimensional shape detection scenes due to the advantages of non-contact, high precision, high response frequency and the like. However, current commercial high-precision spectral confocal sensors generally suffer from small Z-axis measurement ranges, typically between hundreds of microns and 1 millimeter. When there is a large relief or overall height change in the facing surface of the sample, the sensor is prone to measurement interruption, data discontinuity, and even sensor failure, significantly affecting system stability and data integrity. The follow-up detection system of the existing spectral confocal sensor generally adopts a Z-axis follow-up detection strategy based on current height feedback, and the Z-axis actuator is driven to carry out position compensation adjustment only by depending on the height change measured at the current moment. However, this detection method ignores that the actual height of the sample being measured is moving continuously with the XY stage during the execution of the follower in the Z-axis, and in dynamic variation, this results in a change in the Z-axis follower that always lags behind the actual height of the sample. When detecting a complex curved surface workpiece with obvious height fluctuation, the Z-axis follower mechanism is difficult to adjust to the current ideal measurement position in time, so that the following error is increased, and even the sensor is out of the effective range, so that the detection fails. Disclosure of Invention In view of this, in order to solve the technical problem that the existing follow-up detection method relies on the height measured at the current moment to drive, and hysteresis exists, and then the following error increases, in a first aspect, the invention provides a spectral confocal follow-up detection method based on autoregressive feedforward prediction compensation, which comprises the following steps: building a three-dimensional measurement system, and adjusting the spectral confocal sensor to be within a working range; calculating a sample height value in a current detection area, and storing a plurality of historical height value data; Calculating the deviation between the current sensor reading and the sample height value in real time; Predicting future altitude changes by using an autoregressive model; generating a feedforward compensation control amount according to the predicted height; The feedforward compensation control quantity is added to the control signal to generate a final control signal. Based on the scheme, the invention provides the spectral confocal follow-up detection method based on the autoregressive feedforward prediction compensation, a feedforward prediction mechanism based on the autoregressive model is innovatively introduced on the basis of the traditional follow-up detection based on the current height feedback, the real-time modeling and prediction are carried out on the height change trend of a detection sample, the prediction result is used as feedforward compensation quantity to be superimposed into a Z-axis control instruction, the prospective follow-up detection of the Z-axis is realized, the sensor is ensured to be always in an optimal measurement interval, the following precision is improved, and the problem of the sensor overranging is effectively avoided. The method can realize stable and accurate dynamic follow-up detection of the spectral confocal sensor in a complex large-range scene, and effectively expands the application boundary of the spectral confocal technology in complex large-range workpiece measurement. Drawings FIG. 1 is a flow chart of steps of a spectral confocal follow-up detection method based on autoregressive feedforward predictive compensation according to the present invention; FIG. 2 is a schematic diagram of a follow-up detection strategy in accordance with an embodiment of the present invention. Detailed Description The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the appli