ENAC / How do structures really behave? Why does it matter?... to save money! Author James-A. Goulet Supervision Ian F. C. Smith EDCE 2011 / Modelling uncertainties p i j k L m n o x 0 y 0 z 0 PROBABILITY-BASED OBJECTIVE FUNCTION 0 = THE ‘’MOST-LIKELY‘’ MODEL IS OFTEN NOT THE RIGHT MODEL SEVERAL MODELS EXPLAIN MEASUREMENTS 1 STEP Objective function Measuring and Modelling 2 STEP Developping an objective function 3 STEP Interpreting measurements Measurement uncertainties + MODEL FILTERING MODEL FITTING Accelerometers Inclinometers Transducer Strain guages etc... (Model calibration does not work!) 0 Objective function Prediction-measure (Sensor location #i) Prediction-measure (Sensor location #j) Set of acceptable models Single ‘’most-likely’’ model Used to compare measured values to the predictions of several thousand models Prediction-measure (Sensor location #i) Prediction-measure (Sensor location #j) (One per sensor location) 0 0 0 Boundary conditions? What is the prestress level? What is the expansion joint condition? What is the saddle condition? Sensors