A method for ultra-precision polishing of calcium fluoride optical elements

Abstract

The present application relates to the field of optical processing technology and provides an ultra-precision polishing method for a calcium fluoride optical element, comprising the following steps: S10: initializing the polishing process parameters of the calcium fluoride optical element; S20: starting the polishing equipment to polish the calcium fluoride optical element and simultaneously activating the multidimensional sensor network to collect multidimensional sensor data of the polishing area in real time; S30: using a sliding window to calculate the sliding window median and the median absolute deviation normalized value, and checking for abnormal points in the multidimensional sensor data; S40: during the polishing process, dynamically correcting the polishing process parameters based on the distribution of abnormal points in the multidimensional sensor data; S50: when the surface roughness reaches the target value and the surface shape accuracy meets the standard, terminating the polishing to obtain the polished target calcium fluoride. Through multi-sensor fusion and intelligent feedback control, the present application solves the core problems of traditional calcium fluoride polishing, such as poor real-time performance, difficult to control the coupling of multiple physical quantities, and delayed abnormal response.

Description

Technical Field

[0001] The invention belongs to the technical field of optical processing, and in particular relates to an ultra-precision polishing method for a calcium fluoride optical element. Background Art

[0002] Calcium fluoride is an important optical material, widely used in ultraviolet and deep ultraviolet optical systems, such as photolithography lenses and laser optical components. Existing calcium fluoride polishing methods include mechanical polishing, chemical mechanical polishing, magnetorheological polishing, and ion beam polishing.

[0003] However, due to calcium fluoride's low hardness, high brittleness, and large thermal expansion coefficient, traditional polishing methods face numerous challenges during processing. For example, traditional polishing relies on offline detection (such as white light interferometry and atomic force microscopy), which makes it impossible to adjust process parameters in real time, resulting in poor processing consistency. Furthermore, factors such as pressure, temperature, vibration, and the chemical environment interact during the polishing process, making comprehensive monitoring difficult with traditional methods, which can easily lead to partial over- or under-polishing. Summary of the Invention

[0004] In view of the above-mentioned defects of the prior art, the present invention proposes an ultra-precision polishing method for calcium fluoride optical elements. The technical solution designed by the present invention includes the following steps:

[0005] S10: polishing process parameters of initial calcium fluoride optical elements;

[0006] S20: starting the polishing equipment to polish the calcium fluoride optical element, and simultaneously activating the multi-dimensional sensor network to collect multi-dimensional sensor data of the polishing area in real time;

[0007] S30: using a sliding window to calculate the sliding window median and the normalized value of the median absolute deviation to check for abnormal points in the multi-dimensional sensor data;

[0008] S40: During the polishing process, dynamically correcting the polishing process parameters based on the abnormal point distribution of the multi-dimensional sensing data;

[0009] S50: When the surface roughness reaches the target value and the surface accuracy meets the standard, polishing is terminated to obtain the polished target calcium fluoride.

[0010] Preferably, the multidimensional sensor network in S20 comprises a fusion of multiple types of sensors to monitor key physical quantities of the polishing process in real time, and the sensors include but are not limited to mechanical sensors, optical sensors, thermal sensors, motion control sensors and chemical sensors.

[0011] Preferably, the S30 includes:

[0012] S3001: Dividing the multidimensional sensing data into continuous subsequence A and subsequence B in chronological order for cross-validation of polishing stability;

[0013] S3002: Calculate the sliding window median and the normalized median absolute deviation for subsequence A and subsequence B using a sliding window, and check for abnormal points in subsequence A and subsequence B based on the sliding window median and the normalized median absolute deviation.

[0014] S3003: If the difference in the proportion of outliers between subsequence A and subsequence B is greater than a preset difference threshold, extend the global window and re-detect until the difference in the proportion of subsequence A and subsequence B is less than the preset difference threshold, and output the distribution of outliers.

[0015] Preferably, the sliding window median in S3002 is calculated as follows:

[0016] M n =Med(m n-a , m n-a+1 ,…,m n ,…,m n+a )

[0017] Where M n is the data point m in the subsequence n is the median value of the sliding window at the center of the window, and a is the data point m covered in the sliding window n The number of preceding and following data points.

[0018] Preferably, the median absolute deviation normalized value in S3002 is calculated as follows:

[0019]

[0020] Where, α n is the median absolute deviation normalized value.

[0021] Preferably, the checking of outliers of subsequences A and B based on the sliding window median and the normalized value of the median absolute deviation includes:

[0022] Preset abnormal threshold β, if |m n -M n |>α n β, then determine the data point m in the subsequence n For abnormal points.

[0023] Preferably, the S40 includes:

[0024] The sources of abnormal points are divided according to the sensor type, the correlation between the sensor data belonging to the abnormal point source and the polishing process parameters is calculated, a weight is assigned to each sensor belonging to the abnormal point source, the adjustment amount of the polishing process parameters is calculated and the polishing process parameters of the polishing process are corrected.

[0025] Preferably, the S40 further includes:

[0026] Before calculating the correlation between the sensor data belonging to the abnormal point source and the polishing process parameters, the sensor data belonging to the abnormal point source and the polishing process parameters are dimensionlessly processed.

[0027] Beneficial effects:

[0028] 1. This application uses heterogeneous sensor collaborative monitoring to break through the limitations of a single data dimension, accurately capture the multi-dimensional characteristics of the polishing state, and improve the sensitivity of anomaly detection;

[0029] 2. This application determines process stability by checking abnormal points and their proportional differences, extends the window mechanism to avoid misadjustment, significantly reduces noise interference, and ensures that parameter corrections are based on reliable data. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a flow chart of a preferred embodiment of the present invention.

[0031] Figure 2 1 is a flow chart of abnormal point detection according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0032] The embodiments of the present invention are described in detail below. The following embodiments are implemented based on the technical solutions of the present invention, and provide detailed implementation methods and specific operating procedures. However, the protection scope of the present invention is not limited to the following embodiments.

[0033] The present invention designs a method for ultra-precision polishing of calcium fluoride optical elements, such as Figure 1 As shown, the technical solution includes the following steps, specifically including:

[0034] S10: polishing process parameters of initial calcium fluoride optical elements;

[0035] S20: starting the polishing equipment to polish the calcium fluoride optical element, and simultaneously activating the multi-dimensional sensor network to collect multi-dimensional sensor data of the polishing area in real time;

[0036] S30: using a sliding window to calculate the sliding window median and the normalized value of the median absolute deviation to check for abnormal points in the multi-dimensional sensor data;

[0037] S40: During the polishing process, dynamically correcting the polishing process parameters based on the abnormal point distribution of the multi-dimensional sensing data;

[0038] S50: When the surface roughness reaches the target value and the surface accuracy meets the standard, polishing is terminated to obtain the polished target calcium fluoride.

[0039] Preferably, the multidimensional sensor network in S20 is composed of a fusion of multiple types of sensors to monitor key physical quantities of the polishing process in real time. The sensors include but are not limited to mechanical sensors, optical sensors, thermal sensors, motion control sensors and chemical sensors.

[0040] Specifically, the mechanical sensors include dynamic pressure sensors and six-dimensional force sensors, which are respectively used to detect the pressure distribution in the polishing area and the posture and force balance of the polishing head; the optical sensors include online white light interferometers and laser displacement sensors, which are used to measure the surface roughness, surface accuracy and thickness changes of the workpiece in real time; the thermal sensors include infrared thermal imagers and thermocouples, which are respectively used to detect the temperature field distribution in the polishing area and the embedded polishing pad bottom layer; the motion control sensors include encoders and vibration sensors, which are respectively used to detect the speed feedback of the polishing head and the workpiece table and the abnormal vibration spectrum; the chemical sensors are used to detect pH electrodes and conductivity meters.

[0041] Preferably, if Figure 2 As shown, S30 includes:

[0042] S3001: Divide the multidimensional sensor data into continuous subsequence A and subsequence B in chronological order for cross-validation of polishing stability;

[0043] S3002: Calculate the sliding window median and the normalized median absolute deviation for subsequence A and subsequence B using a sliding window, and check for abnormal points in subsequence A and subsequence B based on the sliding window median and the normalized median absolute deviation.

[0044] S3003: If the difference in the proportion of outliers between subsequence A and subsequence B is greater than a preset difference threshold, extend the global window and re-detect until the difference in the proportion of subsequence A and subsequence B is less than the preset difference threshold, and output the distribution of outliers.

[0045] Specifically, the multidimensional sensor network is based on a fixed time t a Collect multi-dimensional sensor data of the polishing area, and the time mark of the single sensor data of the polishing area is i is the number of preset fixed time, is the initial time of the sensor data; the sensor data is divided into two parts according to the time mark, the first half is subsequence A, and the second half is subsequence B; before using the sliding window calculation, the preset global window length is T global, then the time tags of subsequence A and subsequence B are expressed as and Where t0 is the initial time of subsequence A, and the preset sliding window length is T window If the global window needs to be extended for re-detection, the total length of the new window is T global +ΔT, then the time tags of subsequence A and subsequence B are expressed as and Where ΔT is the preset extended global window time.

[0046] Furthermore, since an outlier in a single subsequence may be a temporary fluctuation rather than a systemic issue, a significant difference in the ratio of outliers between two subsequences indicates process instability. Direct corrections can lead to miscalibration, as this may be a temporary disturbance. Comparing the stability of the two subsequences ensures that corrections are based on more reliable data and avoid noise interference. When testing for a small difference in their ratios, it indicates good stability over the entire time period, facilitating subsequent parameter correction based on the outliers. However, a large difference indicates potential instability, necessitating a longer sampling window and retesting to ensure that the subsequences A and B used for correction represent stable states and avoid misjudgments.

[0047] Preferably, the sliding window median in S3002 is calculated as follows:

[0048] M n =Med(m n-a , m n-a+1 ,…,m n ,…,m n+a )

[0049] Where M n is the data point m in the subsequence n is the median value of the sliding window at the center of the window, and a is the data point m covered in the sliding window n The number of preceding and following data points.

[0050] Preferably, the median absolute deviation normalized value in S3002 is calculated as follows:

[0051]

[0052] Where, α n is the median absolute deviation normalized value.

[0053] Preferably, checking for outliers in subsequences A and B based on the sliding window median and the median absolute deviation normalized value includes:

[0054] Preset abnormal threshold β, if |m n -M n |>α n β, then determine the data point m in the subsequence n For abnormal points.

[0055] Specifically, the median and MAD are used to resist non-Gaussian distribution and outliers, which solves the failure problem of the traditional mean-variance method in polishing data. In addition, the sliding window mechanism dynamically tracks the changes in process status and avoids the lag of the fixed threshold method. It is particularly suitable for ultra-precision machining of brittle materials such as calcium fluoride.

[0056] Preferably, S40 includes:

[0057] The sources of abnormal points are divided according to the sensor type, the correlation between the sensor data belonging to the abnormal point source and the polishing process parameters is calculated, a weight is assigned to each sensor belonging to the abnormal point source, the adjustment amount of the polishing process parameters is calculated and the polishing process parameters of the polishing process are corrected.

[0058] Preferably, S40 further includes:

[0059] Before calculating the correlation between the sensor data belonging to the abnormal point source and the polishing process parameters, the sensor data belonging to the abnormal point source and the polishing process parameters are dimensionlessly processed.

[0060] Specifically, the correlation between the sensor data belonging to the outlier source and the polishing process parameters is calculated using the following formula:

[0061]

[0062] Where, ρ ij is the correlation between the i-th sensor and the j-th polishing process parameter, S i (t) is the sensor data sequence of the i-th sensor at time t, is the normal reference value of sensor i, G j (t) is the set value sequence of the j-th polishing process parameter, is the reference setting value of the j-th polishing process parameter, t is the sampling time, and T is the total number of sampling points.

[0063] Assign a weight to each sensor that is a source of anomalies using the following formula:

[0064]

[0065] Where, ω i is the weight distribution coefficient of the i-th type of sensor, N is the total number of sensor types, E ik is the cross-correlation energy between sensors i and k, τ is the time lag, and L is the maximum time lag.

[0066] Calculate the adjustment amount of polishing process parameters, the formula is as follows:

[0067]

[0068] Where, ΔP j is the final parameter adjustment for the jth polishing process parameter, α is the dynamic adjustment coefficient, is the influence factor of sensor i on parameter j.

[0069] Specifically, when assigning weights to each sensor that is the source of an outlier, it is necessary to limit i and k to unequal; in addition, intelligent dynamic adjustment of polishing parameters is achieved through multi-sensor data fusion and correlation weight allocation. Compared with traditional static parameter control or single-dimensional feedback methods, it can simultaneously optimize multiple parameters such as pressure, temperature, speed, pH value, etc., solve the strong coupling nonlinearity problem in calcium fluoride polishing, establish a closed-loop relationship between process parameters-sensor data-material removal model, and realize true multivariable decoupling control.

[0070] The above describes in detail the preferred embodiments of the present invention. It should be understood that numerous modifications and variations based on the concepts of the present invention are possible by those skilled in the art without inventive effort. Therefore, any technical solution that can be derived by those skilled in the art through logical analysis, reasoning, or limited experimentation based on the concepts of the present invention and the prior art should be within the scope of protection defined by the claims.

Claims

1. An ultra-precision polishing method for calcium fluoride optical elements, characterized in that: The following steps are involved: S10: polishing process parameters of initial calcium fluoride optical elements; S20: starting the polishing equipment to polish the calcium fluoride optical element, and simultaneously activating the multi-dimensional sensor network to collect multi-dimensional sensor data of the polishing area in real time; S30: using a sliding window to calculate the sliding window median and the normalized value of the median absolute deviation to check for abnormal points in the multi-dimensional sensor data; S40: During the polishing process, dynamically correcting the polishing process parameters based on the abnormal point distribution of the multi-dimensional sensing data; S50: When the surface roughness reaches the target value and the surface accuracy meets the standard, polishing is terminated to obtain the polished target calcium fluoride; The S40 includes: Classify the sources of abnormal points according to sensor type, calculate the correlation between the sensor data belonging to the abnormal point source and the polishing process parameters, assign a weight to each sensor belonging to the abnormal point source, calculate the adjustment amount of the polishing process parameters and correct the polishing process parameters during the polishing process; The calculation of the correlation between the sensor data belonging to the outlier source and the polishing process parameters is as follows: Where, ρ ij is the correlation between the i-th sensor and the j-th polishing process parameter, S i (t) is the sensor data sequence of the i-th sensor at time t, is the normal reference value of sensor i, G j (t) is the set value sequence of the j-th polishing process parameter, is the reference setting value of the jth polishing process parameter, t is the sampling time, and T is the total number of sampling points; The weight is assigned to each sensor that is the source of the outlier. The formula is as follows: Where, ω i is the weight distribution coefficient of the i-th type of sensor, N is the total number of sensor types, E ik is the cross-correlation energy between sensors i and k, τ is the time lag, and L is the maximum time lag; The formula for calculating the adjustment amount of the polishing process parameters is as follows: Where ΔP j is the final parameter adjustment for the jth polishing process parameter, α is the dynamic adjustment coefficient, is the influence factor of sensor i on parameter j.

2. The ultra-precision polishing method for a calcium fluoride optical element according to claim 1, characterized in that: The multi-dimensional sensor network in S20 is composed of a fusion of multiple types of sensors to monitor key physical quantities of the polishing process in real time. The sensors include but are not limited to mechanical sensors, optical sensors, thermal sensors, motion control sensors and chemical sensors.

3. The ultra-precision polishing method for a calcium fluoride optical element according to claim 1, characterized in that: The S30 includes: S3001: Dividing the multidimensional sensing data into continuous subsequence A and subsequence B in chronological order for cross-validation of polishing stability; S3002: Calculate the sliding window median and the normalized median absolute deviation for each of subsequences A and B using a sliding window, and check for abnormal points in subsequences A and B based on the sliding window median and the normalized median absolute deviation. S3003: If the difference in the proportion of outliers between subsequence A and subsequence B is greater than a preset difference threshold, extend the global window and re-detect until the difference in the proportion of subsequence A and subsequence B is less than the preset difference threshold, and output the distribution of outliers.

4. The ultra-precision polishing method for a calcium fluoride optical element according to claim 3, characterized in that: The sliding window median in S3002 is calculated as follows: M n =Med(m n-a ,m n-a+1 ,…,m n ,…,m n+a ) Where M n is the data point m in the subsequence n is the median value of the sliding window at the center of the window, and a is the data point m covered in the sliding window n The number of preceding and following data points.

5. The ultra-precision polishing method for a calcium fluoride optical element according to claim 4, characterized in that: The median absolute deviation normalized value in S3002 is as follows: Where, α n is the median absolute deviation normalized value.

6. The ultra-precision polishing method for a calcium fluoride optical element according to claim 5, characterized in that: The checking of abnormal points of subsequence A and subsequence B based on the sliding window median and median absolute deviation normalized value includes: Preset abnormal threshold β, if |m n -M n |>α n β, then determine the data point m in the subsequence n For abnormal points.

7. The ultra-precision polishing method for a calcium fluoride optical element according to claim 1, characterized in that: The S40 further includes: Before calculating the correlation between the sensor data belonging to the abnormal point source and the polishing process parameters, the sensor data belonging to the abnormal point source and the polishing process parameters are dimensionlessly processed.

Images