noise and number of sensors
play

noise and number of sensors Giovanni Capellari Eleni Chatzi - PowerPoint PPT Presentation

Optimal sensor placement through Bayesian experimental design: effect of measurement noise and number of sensors Giovanni Capellari Eleni Chatzi Stefano Mariani 3 rd International Electronic Conference on Sensors and Aplications, 10-15 November


  1. Optimal sensor placement through Bayesian experimental design: effect of measurement noise and number of sensors Giovanni Capellari Eleni Chatzi Stefano Mariani 3 rd International Electronic Conference on Sensors and Aplications, 10-15 November 2016

  2. Motivation Structural Health Monitoring can be conceptually divided in three stages: in our work, we will focus on the design of the sensor network SHM system design ๐’† Data collection ๐’› Parameters estimation ๐œพ Decision making

  3. Motivation Optimal The usefulness of the sensor network depends on the SHM system number, type and location of the sensors. Therefore, design we need a method to quantify the information obtained by the acquisition system. measurement error Estimates SHM system Uncertainty Identifiability cost configuration # sensors

  4. Optimal sensor placement: deterministic methods The existing approaches does not take into account the measurement noise, i.e. the sensors accuracy. EFI EVP KE Sensitivity to damage M. Meo , G. Zumpano , (2005), M. Bruggi , S. Mariani , (2013), Leyder , C., Ntertimanis , V., Chatzi , E., Frangi , A. (2015).

  5. Optimal sensor placement: Bayesian framework In a Bayesian sense, the optimal spatial configuration ๐’† โˆ— of the sensor network can be found by maximizing the Shannon information gain. In order to compute it, we use a Monte Carlo approximation. Expected gain in Shannon information ๐’† โˆ— = arg max ๐’†โˆˆ๐‘ฌ ๐‘‰(๐’†) ๐‘‰ ๐’† = เถฑ เถฑ ๐‘ฃ ๐’†, ๐’›, ๐œพ ๐‘ž ๐œพ, ๐’› ๐’† ๐‘’๐œพ๐‘’๐’› ๐’ ฮ˜ Monte Carlo sampling Prior: ๐œพ~๐‘ž ๐œพ Likelihood: ๐’›~๐‘ž(๐’›|๐œพ, ๐’†) ๐‘œ ๐‘๐‘ฃ๐‘ข ๐‘œ ๐‘—๐‘œ 1 1 ln ๐‘ž ๐’› ๐‘— ๐œพ ๐‘— , ๐’† ๐‘ž ๐’› ๐‘— ๐œพ ๐‘˜ , ๐’† ๐‘‰ ๐’† โ‰ˆ เท โˆ’ ln เท ๐‘œ ๐‘๐‘ฃ๐‘ข ๐‘œ ๐‘—๐‘œ ๐‘—=1 ๐‘˜=1 X. Huan , Y. M. Marzouk , (2013).

  6. Model evaluation The measurements are related to the mechanical parameters to be estimated through a FEM-based forward model. The sensor accuracy is taken into account through a fictitious measurement noise. โ€ข Evaluation of the likelihood ๐‘ž ๐’› ๐‘— ๐œพ ๐‘˜ , ๐’† = ๐‘ž ๐› ๐’› ๐‘— โˆ’ ๐‘ฏ ๐œพ ๐‘˜ , ๐’† โ€ข Forward model ๐’› = ๐‘ฏ ๐œพ, ๐’† + ๐› Measurement noise ๐‘ฏ ๐œพ, ๐’† = ๐‘ด ๐’† ๐‘ณ(๐œพ) โˆ’1 ๐‘ฎ ๐‘œ ๐œพ ๐น ๐‘— ๐‘ณ ๐œพ = เท ๐น โˆ’ 1 ๐‘ณ ๐‘ฃ๐‘œ๐‘’ โˆ’ ๐‘ณ ๐‘— Observation matrix ๐‘—=1

  7. Optimization In order to reduce the computational cost of the forward model, a cheaper surrogate model is built. โ€ข Surrogate model: polynomial chaos expansion ๐œพ ๐‘— ~๐‘ž ๐œพ , ๐’† ๐‘— ~๐’ฑ ๐“” ๐‘ˆ ๐’† ๐‘— ๐‘ˆ ๐’€ ๐‘— = ๐œพ ๐‘— ๐’› ๐‘„๐ท๐น = ๐‘ ๐’€ = ฯƒ ๐œทโˆˆโ„• ๐‘ ๐‘ง ๐›ฝ ๐œ” ๐›ฝ ๐’€ ๐’› ๐‘— ๐บ๐น = ๐‘ฏ ๐œพ ๐‘— , ๐’† ๐‘— โ€ข Optimization: Covariance Matrix Adaptation Evolution Strategy (CMA-ES) ๐’ โˆˆ โ„ ๐‘œ ๐’† , ๐‘ซ โˆˆ โ„ ๐‘œ ๐’†ร— ๐‘œ ๐’† 1 . ๐’† ๐‘— ~๐’ + ๐œ๐’ช ๐‘— ๐Ÿ, ๐‘ซ 2. ๐’ and ๐‘ซ are updated through cumulation 3. Check the tolerance on ๐‘‰ ๐’† N. Hansen , S.D. Mรผller , P. Koumoutsakos , (2003).

  8. Bayesian OSP framework Training surrogate model Maximizing information Sample input variables Sample design variable ๐œพ ๐‘— ~๐‘ž ๐œพ , ๐’† ๐‘— ~๐’ฑ ๐“” ๐’† ๐‘š ๐‘ˆ ๐’† ๐‘— ๐‘ˆ ๐’€ ๐‘— = ๐œพ ๐‘— MC approximation ๐‘‰(๐’† ๐‘š ) System response ๐‘ฏ ๐บ๐น ๐œพ ๐‘— , ๐’† ๐‘— Update ๐’† ๐‘š โ†’ ๐’† ๐‘š+1 (CMA-ES) PCE surrogate ๐‘ฏ ๐บ๐น ๐œพ ๐‘— , ๐’† ๐‘— โ‰… ๐‘ฏ ๐‘„๐ท๐น ๐œพ ๐‘— , ๐’† ๐‘— Check tolerance on ๐‘‰ ๐’† ๐‘š โˆ’ ๐‘‰ ๐’† ๐‘š+1 Optimal configuration ๐’† โˆ—

  9. Application: simply supported plate 10x10 mesh: 726 d.o.f. Displacement measurements 4 zones: ๐œพ = ๐น 1 , ๐น 2 , ๐น 3 , ๐น 4

  10. Application: simply supported plate Choice of prior distribution ๐‘ž ๐œพ Optimal position of ๐‘œ ๐‘ก = 4 sensors, results of 10 algorithm runs ๐œพ = ๐น 1 ๐น 2 ๐น 3 ๐น 4 ๐‘‚ ๐‘ก : # sensors ๐‘‚ ๐‘„๐ท๐น : # PCE samples ๐‘‚ ๐‘ก = 4, ๐‘‚ ๐‘„๐ท๐น = 10 4 , ๐‘ž : PCE polynomial degree ๐‘ž = 10 , ๐‘‚ ๐‘๐ท = 5 ยท 10 3 ๐‘‚ ๐‘๐ท : # MC samples ๐‘ž ๐œพ ~๐’ฑ 2 ๐น ๐‘ž ๐œพ ~๐’ฑ 0, ๐น 3 , ๐น

  11. Application: simply supported plate Effect of ฯƒ ๐› Contour of the objective function with one sensor for each possible location on the plate with different standard deviations of the measurement noise. 2 ๐œ—~๐’ช 0, ๐œ ๐œ— ๐‘‚ ๐‘ก : # sensors ๐œพ = ๐น 2 ๐‘‚ ๐‘„๐ท๐น : # PCE samples ๐‘‚ ๐‘ก = 1, ๐‘‚ ๐‘„๐ท๐น = 10 4 , ๐‘ž : PCE polynomial degree ๐‘ž = 10 , ๐‘‚ ๐‘๐ท = 5 ยท 10 3 ๐‘‚ ๐‘๐ท : # MC samples 0.6936 2.9 0.697 0.6935 0.6965 0.6934 2.8 0.696 0.6933 2.7 0.6932 0.6955 0.6931 0.695 2.6 0.693 0.6945 0.6929 2.5 0.694 0.6928 0.6935 2.4 0.6927 0.693 0.6926 ฯƒ ๐œ— = 10 โˆ’3 m ฯƒ ๐œ— = 10 โˆ’4 m ฯƒ ๐œ— = 10 โˆ’5 m

  12. Application: simply supported plate Effect of ฯƒ ๐› and number of sensors Contour of the objective function with one sensor for different standard deviations and number of sensors. 2 ๐œ—~๐’ช 0, ๐œ ๐œ— ๐‘‚ ๐‘ก : # sensors ๐œพ = ๐น 2 ๐‘‚ ๐‘„๐ท๐น : # PCE samples ๐‘‚ ๐‘„๐ท๐น = 10 4 , ๐‘ž : PCE polynomial degree ๐‘ž = 10 , ๐‘‚ ๐‘๐ท = 5 ยท 10 3 ๐‘‚ ๐‘๐ท : # MC samples

  13. Conclusions โ€ข Optimal sensor placement and SHM system design โ€ข Take into account: Bayesian optimal experimental design - Measurements uncertainties - Number of sensors โ€ข Maximization of expected information gain between prior and posterior โ€ข Use of surrogate model (PCE) for MC approximation and stochastic optimization (CMA-ES) methods for computational speed-up โ€ข Future developments: larger number of sensors, larger number of parameters, application to complex cases

  14. References Bruggi , M., and Mariani , S. (2013). โ€œOptimization of sensor placement to detect damage in flexible plates.โ€ Engineering Optimization , 45(6), 659 โ€“ 676. Capellari , G., Eftekhar Azam , S., Mariani , S. (2016). โ€œTowards real -time health monitoring of structural systems via recursive Bayesian filtering and reduced order modelling.โ€ International Journal of Sustainable Materials and Structural Systems , In Press. Hansen , N., Mรผller , S. D., Koumoutsakos , P. (2003). โ€œReducing the time complexity of the derandomized evolution strategy with Covariance Matrix Adaptation (CMA- ES).โ€ Evolutionary Computation , 11(1), 1-18. Huan , X., and Marzouk , Y. M. (2013). โ€œSimulation -based optimal Bayesian experimental design for nonlinear systems.โ€ Journal of Computational Physics , 232(1), 288 โ€“ 317. Leyder , C., Ntertimanis , V., Chatzi , E., Frangi , A. (2015). โ€œOptimal sensors placement for the modal identification of an innovative timber structure.โ€ Proceedings of the 1 st International Conference on Uncertainty Quantification in Computational Sciences and Engineering , 467-476. Lindley , D. V. (1972). Bayesian Statistics, A Review , Society for Industrial and Applied Mathematics, SIAM. Marelli , S., and Sudret , B. (2015). UQLab User Manual , Chair of Risk, Safety & Uncertainty Quantification, ETH Zรผrich. Capellari , G., Chatzi , C., Mariani , S. (2016). An optimal sensor placement method for SHM based on Bayesian experimental design and Polynomial Chaos Expansion Proceedings of the VII European Congress on Computational Methods in Applied Sciences and Engineering . Meo , M., and Zumpano , G. (2005). โ€œOn the optimal sensor placement techniques for a bridge structure.โ€ Engineering Structures , 27, 1488-1497. Ryan , K. J. (2003). โ€œEstimating expected information gains for experimental designs with application to the random fatigue-limit model .โ€ Journal of Computational and Graphical Statistics , 12(3), 585-603. Papadimitriou , C., (2004). โ€œOptimal sensor placement methodology for parametric identification of structural systems.โ€ Journal of Sound and Vibration, 278 (4), 923-947.

Recommend


More recommend