CN111988011A - Anti-divergence method for filter - Google Patents
Anti-divergence method for filter Download PDFInfo
- Publication number
- CN111988011A CN111988011A CN202010756820.9A CN202010756820A CN111988011A CN 111988011 A CN111988011 A CN 111988011A CN 202010756820 A CN202010756820 A CN 202010756820A CN 111988011 A CN111988011 A CN 111988011A
- Authority
- CN
- China
- Prior art keywords
- filter
- deviation
- value
- average value
- sliding window
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03H—IMPEDANCE NETWORKS, e.g. RESONANT CIRCUITS; RESONATORS
- H03H9/00—Networks comprising electromechanical or electro-acoustic devices; Electromechanical resonators
- H03H9/46—Filters
Landscapes
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Feedback Control In General (AREA)
Abstract
The invention relates to a method for preventing divergence of a filter, which is characterized in that a sliding window with a certain width for storing the output and input deviation of the filter is established, and the sliding window is initialized when the filter is initialized; and respectively sliding the current deviation value and the first deviation value into and out of the sliding window in each period, sequentially sliding the intermediate deviation value, calculating the absolute value of the average deviation of the sliding window, judging that the filter diverges when the result is greater than a certain specified threshold, resetting the filter to work again, and normally working the filter when the result is within the specified threshold so that the filter is in a closed-loop controlled state all the time. The invention adds a monitoring measure for the filter, and the filter is reset in time when diverging, so that the filter is always in a controlled state.
Description
Technical Field
The invention belongs to the field of digital filters, and the method is suitable for occasions applying the digital filters.
Background
Digital filters are widely used in modern circuits to filter/suppress interference and retain useful signals through complex algorithms, however, because of various reasons, the filters are unstable, and the problem of uncontrollable divergence often occurs in practical operation, so engineers always have many ways to put an end to this problem, and refer to related patents, and the solution is as follows:
publication (publication) No. (CN108550371A) patent "echo cancellation method for fast settling of intelligent voice interaction device" while preventing divergence of adaptive filter by limiting the increment range of filter coefficients during tracking "describes that the adaptive filter can be prevented from being used for tracking.
The patent of publication (publication) No. (CN108663068A) a SVM adaptive kalman filtering method applied in initial alignment states that "aiming at the problems that a noise model is difficult to quantize and a system has sudden change, thereby causing filter divergence and unstable numerical value, proposes to introduce an adaptive factor generated by a support vector machine, to improve the robustness of an algorithm, and to improve the tracking capability of state and parameter sudden change".
The method is that a feedback loop is constructed through observation noise in Kalman filtering, a model noise variance matrix is adjusted on line in the feedback loop by applying a machine learning theory, and an estimation mean square error matrix and a filtering gain matrix of a filter are changed, so that the filter is prevented from diverging, and the filtering precision is improved.
The disclosure (publication) of patent "a method for tightly combined adaptive filtering against GPS outliers" in publication (CN103983996A "adopts a fuzzy logic controller to monitor the filter residual sequence to estimate the measured noise strength on line to adaptively adjust the measured noise matrix parameter value of the filter, so as to improve the filtering accuracy and prevent the filter from diverging".
Publication (publication) No. (CN108646191A) patent "a method for estimating state of charge of battery based on DAFEKF" said "using time-varying attenuation factor to suppress memory length of filter, … …, while adaptively adjusting process noise and measurement noise covariance, preventing phenomena such as … … filter divergence".
The above patents all dynamically modify filter parameters to avoid filter divergence, most algorithms are complex and should have some effect, but the action mechanism cannot ensure that divergence is eradicated immediately after divergence occurs.
Disclosure of Invention
Technical problem to be solved
In order to avoid the defects of the prior art, the invention provides a filter anti-divergence method.
Technical scheme
A method for preventing divergence of a filter, comprising: setting a sliding window with a certain width for storing the deviation between the output and the input of the filter, and initializing the sliding window when the filter is initialized; respectively sliding the current deviation value and the first deviation value into and out of the sliding window in each period, sequentially sliding the middle deviation value, calculating the absolute value of the average deviation of the sliding window, and when the result is greater than a certain specified threshold, judging that the filter diverges, and resetting the filter to work again; when the filter is within the specified threshold, the filter works normally, so that the filter is always in a closed-loop controlled state.
The specified threshold refers to a local filter monitoring threshold, and is obtained by adopting a statistical method: for batch products, a certain product acquires sample files 1-m respectively containing N sampling points according to use conditions, each sampling point comprises a filter input sampling value and a filter output sampling value, each i of the sampling points of each sample file is divided into 10, the deviation value of the filter output and input of each sampling point is calculated and stored in a corresponding segmented file, then the average value of the i deviation values is obtained, and the average value is respectively written into a local deviation average value file for the product and a global deviation average value file for the batch products until all the sample files are processed; and traversing the local deviation average value file and the global deviation average value file to respectively obtain a local deviation average value maximum value and a global deviation average value maximum value, respectively multiplying the two maximum values by a margin coefficient which is more than 1, and taking the absolute value to respectively obtain a local filter monitoring threshold and a global filter monitoring threshold of the product and the batch product under the use condition.
The number of i-10 may also be 10 × j, where j is a positive integer with j being equal to or greater than 1 and equal to or less than 5 and N being equal to or less than 5 j.
The margin coefficient is 2.5.
Advantageous effects
The invention provides a method for preventing divergence of a filter, which adds a monitoring measure for the filter, and resets in time when the filter diverges, so that the filter is always in a controlled state. Compared with the prior art, the invention has the following beneficial effects:
1. the method is simple;
2. the divergence is eliminated in time;
3. the reliability of the filter is greatly improved.
Drawings
FIG. 1 is a schematic diagram of the present invention
FIG. 2 is a diagram of an embodiment: (a) leveling the lower embodiment, and (b) erecting the lower embodiment with rough leveling.
Detailed Description
The invention will now be further described with reference to the following examples and drawings:
a simple and easy filter anti-divergence method is characterized in that:
setting a sliding window with a certain width (1.. n, n is the width of the sliding window, n is more than or equal to 1) for storing the deviation between the output and the input of the filter, and initializing the sliding window when the filter is initialized; respectively sliding the current deviation value and the first deviation value into and out of the sliding window in each period, sequentially sliding the intermediate deviation value, calculating the absolute value of the average deviation of the sliding window, and when the result is greater than a certain specified threshold, judging that the filter diverges, resetting the filter to work again, and simultaneously (or not) resetting the sliding window; when the filter is within the specified threshold, the filter works normally, so that the filter is always in a closed-loop controlled state.
The threshold determination may be obtained by statistical methods, and one filter monitoring threshold statistical algorithm is referred to as follows: for a batch product, a certain product may obtain sample files 1 to m respectively including N sampling points according to a use condition 1, each sampling point includes filter input and output sampling values, each i of the sampling points of each sample file is divided by 10 (or other sampling numbers i, i is 10 × j, j is a positive integer with j being equal to or greater than 1 and equal to or less than 5 and j being equal to or less than 5) to obtain a deviation value of the filter output and input of each sampling point, and the deviation value is stored in a corresponding segment file, then an average value is obtained for the i deviation values, and the average value is written into a local deviation average value file for the product and a global deviation average value file for the batch product, if all the sample files are processed. Traversing the local deviation average value file and the global deviation average value file to respectively obtain a local deviation average value maximum value and a global deviation average value maximum value, respectively multiplying the two maximum values by a margin coefficient which is more than 1, respectively obtaining the absolute values to respectively obtain a local filter monitoring threshold 1 and a global filter monitoring threshold 1 under the use condition 1 of the product and the batch product, and respectively storing the local filter monitoring threshold and the global filter monitoring threshold into corresponding files. The local and global filter monitoring threshold acquisition method under other use conditions of the product is similar.
Example 1:
an embodiment of the invention is schematically shown in figure 2 a.
Referring to fig. 2a, a sliding window with a certain width of 64 is established to store the deviation between the output and the input of the filter, and the sliding window is initialized when the filter is initialized; respectively sliding the current deviation value and the first deviation value into and out of the sliding window in each period, sequentially sliding the middle deviation value, calculating the absolute value of the average deviation of the sliding window, and when the result is greater than the dispersion threshold under leveling, judging that the filter disperses, resetting the filter to work again, and simultaneously (or not) resetting the sliding window; when the dispersion threshold is within the leveling lower dispersion threshold, the filter normally works, so that the filter is always in a closed-loop controlled state.
The threshold determination can be obtained by adopting a statistical method, and the statistical algorithm of the filter monitoring threshold is as follows: for a batch product, a certain product can obtain sample files 1-3 respectively containing 207 sampling points according to leveling use conditions, each sampling point comprises filter input and output sampling values, each sampling point of each sample file is divided into 10 sampling points, the deviation value of the filter output and input of each sampling point is calculated and stored in a corresponding segmented file, then the average value of the 10 deviation values is obtained, and the average value is respectively written into a local deviation average value file for the product and a global deviation average value file for the batch product, if all the sample files are processed. And traversing the local deviation average value file and the global deviation average value file to respectively obtain a local deviation average value maximum value and a global deviation average value maximum value, respectively multiplying the two maximum values by a margin coefficient of 2.5, respectively obtaining absolute values to respectively obtain local and global filter monitoring thresholds aiming at the product and the batch products under the leveling use condition, and respectively storing the local and global filter monitoring thresholds into corresponding files.
Example 2:
referring to fig. 2b, a sliding window with a certain width of 64 is established to store the deviation between the output and the input of the filter, and the sliding window is initialized when the filter is initialized; respectively sliding the current deviation value and the first deviation value into and out of the sliding window in each period, sequentially sliding the middle deviation value, calculating the absolute value of the average deviation of the sliding window, and when the result is greater than the divergence threshold under the erection of the coarse adjustment level, judging that the filter diverges, resetting the filter to work again, and simultaneously (or not) resetting the sliding window; when the divergence threshold is within the setting coarse adjustment level, the filter works normally, so that the filter is always in a closed-loop controlled state.
The method for obtaining the monitoring threshold of the local and global filters of the product under the use condition of erection rough leveling is similar to the previous embodiment.
Claims (4)
1. A method for preventing divergence of a filter, comprising: setting a sliding window with a certain width for storing the deviation between the output and the input of the filter, and initializing the sliding window when the filter is initialized; respectively sliding the current deviation value and the first deviation value into and out of the sliding window in each period, sequentially sliding the middle deviation value, calculating the absolute value of the average deviation of the sliding window, and when the result is greater than a certain specified threshold, judging that the filter diverges, and resetting the filter to work again; when the filter is within the specified threshold, the filter works normally, so that the filter is always in a closed-loop controlled state.
2. The method as claimed in claim 1, wherein the specified threshold is a local filter monitoring threshold obtained by a statistical method: for batch products, a certain product acquires sample files 1-m respectively containing N sampling points according to use conditions, each sampling point comprises a filter input sampling value and a filter output sampling value, each i of the sampling points of each sample file is divided into 10, the deviation value of the filter output and input of each sampling point is calculated and stored in a corresponding segmented file, then the average value of the i deviation values is obtained, and the average value is respectively written into a local deviation average value file for the product and a global deviation average value file for the batch products until all the sample files are processed; and traversing the local deviation average value file and the global deviation average value file to respectively obtain a local deviation average value maximum value and a global deviation average value maximum value, respectively multiplying the two maximum values by a margin coefficient which is more than 1, and taking the absolute value to respectively obtain a local filter monitoring threshold and a global filter monitoring threshold of the product and the batch product under the use condition.
3. The method according to claim 2, wherein i-10 is a positive integer with j being 1 ≦ j ≦ 5 and N being 5j ≦ 10 xj.
4. A method as claimed in claim 2, wherein said margin coefficient is 2.5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010756820.9A CN111988011B (en) | 2020-07-31 | 2020-07-31 | Anti-divergence method for filter |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010756820.9A CN111988011B (en) | 2020-07-31 | 2020-07-31 | Anti-divergence method for filter |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111988011A true CN111988011A (en) | 2020-11-24 |
| CN111988011B CN111988011B (en) | 2023-01-03 |
Family
ID=73444830
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010756820.9A Active CN111988011B (en) | 2020-07-31 | 2020-07-31 | Anti-divergence method for filter |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111988011B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050094889A1 (en) * | 2003-10-30 | 2005-05-05 | Samsung Electronics Co., Ltd. | Global and local statistics controlled noise reduction system |
| WO2007142111A1 (en) * | 2006-06-08 | 2007-12-13 | Nec Corporation | Noise erasing device and method, and noise erasing program |
| US20110019094A1 (en) * | 2009-07-21 | 2011-01-27 | Francois Rossignol | System and method for random noise estimation in a sequence of images |
| US20150285916A1 (en) * | 2014-04-07 | 2015-10-08 | Honeywell International Inc. | Systems and methods for a code carrier divergence high-pass filter monitor |
| US20160019777A1 (en) * | 2014-07-18 | 2016-01-21 | Google Inc. | Systems and methods for intelligent alarming |
| CN106130508A (en) * | 2016-06-13 | 2016-11-16 | 电子科技大学 | Digital multimeter noise-reduction method based on FIR filter |
| CN108631753A (en) * | 2018-05-15 | 2018-10-09 | 西安空间无线电技术研究所 | A kind of integration compensation digital filter design method |
| WO2020078399A1 (en) * | 2018-10-17 | 2020-04-23 | 深圳锐越微技术有限公司 | Filtering method and device for filter, filter and storage medium |
-
2020
- 2020-07-31 CN CN202010756820.9A patent/CN111988011B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050094889A1 (en) * | 2003-10-30 | 2005-05-05 | Samsung Electronics Co., Ltd. | Global and local statistics controlled noise reduction system |
| WO2007142111A1 (en) * | 2006-06-08 | 2007-12-13 | Nec Corporation | Noise erasing device and method, and noise erasing program |
| US20100183165A1 (en) * | 2006-06-08 | 2010-07-22 | Nec Corporation | Noise cancelling device and method, and noise cancelling program |
| US20110019094A1 (en) * | 2009-07-21 | 2011-01-27 | Francois Rossignol | System and method for random noise estimation in a sequence of images |
| US20150285916A1 (en) * | 2014-04-07 | 2015-10-08 | Honeywell International Inc. | Systems and methods for a code carrier divergence high-pass filter monitor |
| US20160019777A1 (en) * | 2014-07-18 | 2016-01-21 | Google Inc. | Systems and methods for intelligent alarming |
| CN106130508A (en) * | 2016-06-13 | 2016-11-16 | 电子科技大学 | Digital multimeter noise-reduction method based on FIR filter |
| CN108631753A (en) * | 2018-05-15 | 2018-10-09 | 西安空间无线电技术研究所 | A kind of integration compensation digital filter design method |
| WO2020078399A1 (en) * | 2018-10-17 | 2020-04-23 | 深圳锐越微技术有限公司 | Filtering method and device for filter, filter and storage medium |
Non-Patent Citations (2)
| Title |
|---|
| JAMES L. COYTE等: ""A Drift Detecting Anti-Divergent EKF for Online Biodynamic Model Identification"", 《2015 IEEE 17TH INTERNATIONAL CONFERENCE ON HIGH PERFORMANCE COMPUTING AND COMMUNICATIONS, 2015 IEEE 7TH INTERNATIONAL SYMPOSIUM ON CYBERSPACE SAFETY AND SECURITY, AND 2015 IEEE 12TH INTERNATIONAL CONFERENCE ON EMBEDDED SOFTWARE AND SYSTEMS》 * |
| 张园等: ""机动目标跟踪的一种防发散RBUKF算法"", 《指挥仿真与控制》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111988011B (en) | 2023-01-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR900005546B1 (en) | Adaptive process control system | |
| US5166873A (en) | Process control device | |
| CA2315637C (en) | Method of predicting overshoot in a control system response | |
| CN112418051B (en) | State estimation method for nonlinear dynamic system under non-Gaussian noise | |
| Oussalah et al. | Adaptive Kalman filter for noise identification | |
| CN108983610B (en) | Robust self-adaptive anti-interference control method | |
| CN109946979B (en) | Self-adaptive adjusting method for sensitivity function of servo system | |
| CN111913175A (en) | Water surface target tracking method with compensation mechanism under transient failure of sensor | |
| CN108512528B (en) | Ratio control and normalization LMP filtering method under a kind of CIM function | |
| Khosravian et al. | State estimation for nonlinear systems with delayed output measurements | |
| CN110555398A (en) | Fault diagnosis method for determining first arrival moment of fault based on optimal filtering smoothness | |
| CN111988011B (en) | Anti-divergence method for filter | |
| CN112946641B (en) | Data filtering method based on relevance of Kalman filtering innovation and residual error | |
| CN110048694A (en) | Random Fourier's feature core least mean square algorithm based on argument step-length | |
| CN108804721B (en) | Oil pumping machine fault diagnosis method based on self-adaptive unscented Kalman filtering and RBF neural network | |
| US20240094690A1 (en) | Generalized estimator (ge), and generalized disturbance rejection controller (gdrc) and design method thereof | |
| CN108462481A (en) | Ratio LMP filtering methods based on parameter adjustment under a kind of μ rule function | |
| CN108512529A (en) | Ratio control under a kind of μ rule function and normalization LMP filtering methods | |
| KR101572241B1 (en) | Control system with robust control capability | |
| Kim et al. | Controller design for time domain specifications | |
| CN110723594A (en) | Rope tying tension control device and method | |
| CN111222214A (en) | Improved strong tracking filtering method | |
| FILIPOVIĆ et al. | On robustified adaptive minimum-variance controller | |
| CN116031604B (en) | Automatic debugging method of microwave filter based on response feature extraction | |
| CN110542412B (en) | Self-adaptive dynamic and static closed-loop control method for nuclear magnetic resonance gyroscope |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |