Release Notes
1.12.0
New features
- Add
to_dBfunction to convert a signal to decibels, distinguishing amplitude (root-power) quantities (type = :amplitude, 20 log₁₀) from power/energy quantities (type = :power, 10 log₁₀). - Add
ref_dBfunction returning the usual reference values (:mobility,:accelerance,:displacement,:velocity,:acceleration,:pressure) for amplitude or power quantities. - Add
peak_value,peak2peakandrms_valuefunctions to compute the peak, peak-to-peak and RMS values of a signal. - Add
unwrap_phasefunction to compute the unwrapped phase of a signal. The unwrapping is always performed in radians before an optional conversion to degrees (deg = true), and independently for each channel of a multichannel signal.
Bug fixes
- Fix the phase unwrapping in
bode_plotwhenisdeg = true. The phase was converted to degrees before being unwrapped with a 2π range, leading to a wrong phase. - Fix the cross-spectrum convention in
tfestimate. The cross-spectrum was computed as \(X Y^*\) instead of \(X^* Y\), so that the estimated transfer functions (:h1,:h2,:h3,:hv) were the complex conjugates of the true ones (wrong sign of the phase). As a consequence, the modal parameters could not be identified from FRFs estimated withtfestimate. Breaking for code that compensated for this behaviour (e.g. withconj). - Remove the window energy correction in
tfestimate. Transfer functions and coherence are ratios of spectral densities, for which the correction cancels out when the same window is used for the input and the output. With different windows (e.g.forceandexponentialwindows for impact testing), the correction biased the estimated FRFs by a factorrms(window_input)/rms(window_output)(about -8 dB in a typical impact test). - Fix
agwnandmixed_noisefor arrays with more than two dimensions (e.g. FRF arrays of sizeno × ni × nf). The noise standard deviation was broadcast along the wrong dimension, which changed the size of the output whenni > 1. - Fix
excitation(::Hammer, t)whentstartdoes not coincide with a time sample. The signal started at the closest sample, which could lie beforetstartand led to aDomainError. The signal now starts at the first sample greater than or equal totstart, and is zero iftstartis beyond the time vector. - Fix
varestwhenbatch_size > 0, which threw aMethodError. Incomplete batches are now correctly discarded for multichannel signals. - Fix wrong analytical results:
impulse_response(Mdof) kept only the last mode instead of summing all modes.modefreqfor clamped-free bars, rods and strings returned twice the natural frequencies.modefreq(::Beam, fmax, :CF)used a wrong second root and skipped the fifth mode. Tabulated roots are more accurate and are now filtered byfmaxfor all beam boundary conditions.modeshapewas not mass-normalized for free-free bars (modes ≥ 2), clamped-free beams (first modes) and free-free beams (elastic and rotation modes).modefreq(::Membrane, fmax)used the plate dispersion relation.srs(...; alg = :RecursiveInt)was 1–4 % off. It now uses the exact Nigam-Jennings recursion.convert_csd(...; type = :vel)returned spectra with the wrong sign.comaccould exceed 1 andmpdwas non-zero for real modes with components of opposite signs.odsweighted the candidates on the left of the peak bymsf²instead ofmsf.- EFDD fallback of
FSDD: wrong time step (overestimated damping) and bell spectra accumulated across peaks. - Poles identified with
LSCE,LSCFandpLSCFon a frequency range not starting at 0 Hz are now shifted back exactly.
- Fix errors on valid inputs:
- Sdof solvers with
ω0 = 0andsolve(::SdofForcedTimeProblem; method = :interp). DirectFRFProblemwith an excitation selection matrixSeof fewer rows than DOFs, or when onlySois given (same forStateSpaceFRFProblemandStateSpaceModalFRFProblem).tfestimate,welch,csdandspectrumwith an odd block size.detrendwith breakpoints,planckwith an odd length,acnwith a complex matrix, andexcitation(::GaussianPulse, t)with a negative amplitude.SineWave,Hammer,SdofandOneDMeshnow accept arguments of different numeric types (e.g. integers).
- Sdof solvers with
- Other fixes:
solve(::SdofForcedTimeProblem)withtype_exc = :baseno longer overwrites the input vector.excitation(::ColoredNoise, t)andacnno longer skip the time windowing whenfrangestarts at 0 Hz.varest(x, DerricoEst())returns a single value for a vector.msfthrew anUndefVarErrorinstead of aDimensionMismatch.- The ESD scaling of
welchandcsduses the block durationbs/fs.xcorruses an unbiased normalization. KalmanDenoising: the search range of the process noise variance is now derived from the data, and the RTS covariance update is corrected.- Regularization-based denoising uses
idctinstead of a densen × nmatrix (memory nowO(n)).
Documentation
- Fix the
mcfdocstring: 0 for real modes, 1 for complex modes. - Add a Tutorials section, organized by user profile:
- Getting started
- From model to virtual measurement: simulation of an impact hammer test, FRF estimation, averaging and impact testing windows
- Experimental modal analysis: stabilization diagram, mode shape extraction, MAC and FRF synthesis
- Operational modal analysis: identification from ambient vibrations with Covariance-driven SSI
- Choosing a time integration scheme: stability, accuracy and numerical dissipation
- Noise estimation and denoising
- Visualizing and animating deflection shapes
- Using your own models and data
- Add a “New to the package?” entry point on the home page.
1.11.0
New features
- Add five new applications in
SVAppsExt.jlextension for educational purposes:- Free Response of a 1-DOF System
- Forced Response of a 1-DOF System
- Free Response of a 2-DOF System
- Forced Response of a 2-DOF System
- Newmark Method for a 2-DOF System
1.10.3
- Fix the coefficient of the stiffness-velocity term in the Newmark family solvers, which reduced
Newmark,HHT,WBZ,FoxGoodwin,LinearAccelerationandGeneralizedAlphawithρ < 1to first-order accuracy. The defaultGeneralizedAlpha()andMidPointare not affected. - Fix the inverted validation of
αfandαminHHT,WBZandGeneralizedAlpha, which rejected valid values and accepted invalid ones. - Out-of-range values of
ρinHHT,WBZandGeneralizedAlphanow throw anArgumentError, as does providing only one ofαfandαminGeneralizedAlpha.
Internals
- Add a second-order convergence test for the Newmark family solvers.
1.10.2
Internals
- Add Fast colormap from Paraview as the default colormap for mesh visualization and animation in
SVAnimExt.jlextension. Thecolormapargument ofviz_meshandanimate_meshfunctions now defaults tofast_cmap()instead of:jet1.
1.10.1
- Fix minor bugs in
animate_meshfunction inSVAnimExt.jlextension.
1.10.0
New features
- Add mesh manipulation functions in
SVMeshExt.jlextension:detrend_meshto detrend a mesh by removing the rigid body motionrenumber_element_connectivityto renumber the elements of a meshconstruct_meshto construct aMeshobject from the nodes and elements of a meshdeformed_gridto compute the deformed grid of a mesh from the nodal displacements
- Add
animate_meshfunction inSVAnimExt.jlextension to create an animation of the deformed mesh over time from the nodal displacements and values to be plotted on the mesh (e.g., modal shapes, ODS, etc.).
Internals
- Simplify the implementation of
odsfunction.
1.9.1
- Fix type instability issues in
gradientfunction. - Improve
compute_polesfunctions forSdofEMA.
1.9.0
New features
- Add
modal_densityfunction to compute the modal density of a structure from the knowledge of its resonance frequencies. - Add
modal_densityfunction for 1D structures (bars, rods, strings and beams) and 2D structures (plates and membranes) based on the analytical expression of the modal density for these structures.
Internals
- Replace
Interpolations.jlwithDataInterpolations.jlfor interpolation.
1.8.0
New features
- Add
wave_parametersfunction to compute the wave parameters (wavenumber, wavelength and wave speed) for continuous models.
1.7.0
Breaking changes
- Remove
AutoEMASdofProblemandAutoEMAMdofProblemstructures - Refactor
solvefunction for automatic modal extraction to take anEMAProblemand aSdofEMAorMdofEMAmethod as arguments instead of separateAutoEMAProblemstructures.
These changes have been made to solve type instability issues in the solve function for automatic modal extraction.
Internals
- Fix various type instability issues.
1.6.1
- Remove
DispatchDoctorfrom dependencies
1.6.0
Breaking changes
- Simplify
poles_extractionmethod by removingscalingargument used for peak detection. The FRF magnitude is now always converted to dB scale for peak detection.
Internals
- Remove
peak_proms!andpeak_widths!fromStructuralVibration.jl. - Fix type instability issues in
varestandpoles_extractionfunctions.
1.5.0
Breaking changes
- Refactor
poles_extractionfunction to addscaling::Functionargument for peak picking methods to specify the scaling of the FRF magnitude used for peak detection
v1.4.1
Documentation
Update documentation for FSDD method in OMA modal extraction.
Update FSDD method to take into account the fact that the number of outputs is different from the number of references.
v1.4.0
Documentation
- Add examples of time solvers with
DifferentialEquations.jlintegration.w
New features
- Add
FSDDmethod tomode_extractionfunction for modal parameter extraction using the Frequency-spatial Domain Decomposition (FSDD) method.
Breaking changes
- Refactor
FreeModalTimeProblem,HarmonicModalTimeProblemandForcedModalTimeProblemconstructors to simplify handling of modal data.
Bug fixes
- Fix bugs in
eigenmodefunction. - Fix bugs in
solvefunction forDirectTimeProblem.
v1.3.0
New features
Add
pks_indicesargument toAutoEMASdofProblemandpoles_extractionfunction to specify the indices of peaks to consider during pole extraction using Sdof methods.peaksplotis renamed topeaks_plotto avoid name conflict with thepeaksplotfunction fromPeaks.jl.Add
odsto compute Operational Deflection Shapes (ODS) using either the response spectrum or the cross-spectral density matrix.
Breaking changes
Rename
c2r_modeshapefunction toreal_normalizationfor clarity.Sdof EMA methods now take a
pks_indicesargument to specify the indices of peaks to consider during pole extraction and bypass automatic peak detection.
v1.2.0
Minor changes
Fix bugs in
eigenmodefunction for state-space models.Fix bugs in
comacfunction.
Breaking changes
Change
PLSCFalgorithm name topLSCFinpoles_extractionfunction.Remove
ismatargument fromsolvefunction ofDirectFRFProblem,ModalFRFProblem,StateSpaceFRFProblemandStateSpaceModalFRFProblem. The output FRF is now always returned as a 3D array with dimensions (no. outputs, no. inputs, no. frequencies).Remove
ismatargument fromimpulse_responsefunction for modal time solvers. The output is now always returned as a 3D array with dimensions (no. dofs, no. dofs, no. frequencies).Add
progressargument tostabilizationfunction to control the display of the progress bar during stabilization analysis.Add
width,min_promandmax_promarguments topoles_extractionfunction andAutoEMASdofProblemto specify the half-width of the peaks and the minimum and maximum peak prominence for pole extraction methods.Add
peaksplotto visualization functions to inspect peak detection in pole extraction methods.
v1.1.4
- Add
linkxaxes!function instabilization_plotto link x-axes of poles and indicator plots when zooming.
v1.1.3
- Solve some type instability issues in Modal extraction functions.
v1.1.2
- Add ToeplitzMatrices packages to compute the Hankel matrix in LSCE.
v1.1.1
- Minor bug fix in
msffunction.
v1.1.0
Breaking changes
- Window functions
- Add
flattrifunction to generate a flat-triangular window
- Add
- Modal extraction
- Structure simplification.
EMASdofProblemandEMAMdofProblemstructures are nowEMAProblem - Structure simplification.
EMASdofSolutionandEMAMdofSolutionstructures are nowEMASolution - Add
xcorrfunction to compute the cross-correlation matrix of a signal or between two signals - Add
psd_from_tffunction to compute the full Power Spectral Density (PSD) matrix of multiple signals from transfer function estimates and input PSDs - Add
half_psdfunction to compute the half Power Spectral Density (PSD) matrix from the full PSD matrix or from time-domain signals - Rename
EMAMdofStabilizationstructure toStabilizationAnalysisfor clarity and generality dpiargument ofmodeshape_extractionfunction is now passed as keyword arguments instead of positional arguments (for both Sdof and Mdof methods)typeargument argument ofmodeshape_extractionfunction (Mdof methods) has been renamed tomodetypefor clarity- Add
OMAProblemstructure to perform Operational Modal Analysis (OMA) - Add
modes_extractionfunction forOMAProblemto extract modal parameters using Covariance-based SSI (CovSSI) and Data-based SSI (DataSSI) methods - Add
halfspec_reconstructionfunction to reconstruct half-spectrum from extracted modal parameters from residues and poles for OMA and EMA-based methods - Add
compute_residualsfunction forOMAProblemto compute the lower and upper residuals of the half-spectrum - Add
display_polesargument tostabilization_plotfunction to choose which poles to display in the stabilization diagram plot
- Structure simplification.
v1.0.0
Breaking changes
- Sdof solvers
- Refactor
SdofHarmonicTimeProblem,SdofForcedTimeProblemandSdofFrequencyProblemto comply withDifferentialEquations.jlconventions
- Refactor
New features
- Signal processing
- Add
csdfunction to compute the Cross Spectral Density (CSD) between two signals or between multiple input and output signals - Add support for matrix inputs in
tfestimatefunction to compute the Frequency Response Function (FRF) between multiple input and output signals - Refactor
tfestimate,welch,csdandspectrumfunctions to accept window functions instead of pre-computed window vectors - Add
nfftparameter totfestimate,welch,csdandspectrumfunctions to specify the number of FFT points to be used
- Add
- Modal extraction
- Add
LSFit()method for modal parameter extraction using least squares fitting of the frequency response function around the resonance peak - Add
AutoEMASdofProblemandAutoEMAMdofProblemstructures andsolvefunctions to perform automatic modal extraction based on Sdof or Mdof approximations - Solutions of AutoEMA problems are stored in
EMASdofSolutionorEMAMdofSolutionstructures containing the poles and the modal parameters - Add
poles_extractionfunction to extract system poles using different Mdof methods:- Least-Squares Complex Exponential (LSCE) method
- Least-Squares Complex Frequency-domain (LSCF) method
- Polyreference Least-Squares Complex Frequency-domain (pLSCF) method
- Add
modeshape_extractionfunction to extract mode shapes from residues and poles using Mdof methods - Add
stabilizationfunction to perform stabilization diagram analysis using Mdof methods - Add
stabilization_plotfunction to visualize stabilization diagrams - Stabilization results are now stored in the
EMAMdofStabilizationstructure for plotting and further analysis - Add
frf_reconstructionfunction to reconstruct FRFs from extracted modal parameters - Add
compute_residualsfunction to compute the lower and upper residuals of the FRF - Add
mode2residuesfunction to compute residues from real modes information - Add modal extraction indicators:
- Modal Overlap Factor (MOF)
- Mode Overcomplexity Value (MOV)
- Mode Phase Collinearity (MPC)
- Mode Complexity Factor (MCF)
- Mode Phase Deviation (MPD)
- Modal Scale Factor (MSF)
- Modal Assurance Criterion (MAC)
- Coordinate Modal Assurance Criterion (COMAC)
- Enhanced Coordinate Modal Assurance Criterion (ECOMAC)
- Frequency Response Assurance Criterion (FRAC)
- Complex Mode Indicator Function (CMIF)
- Power Spectrum Indicator Function (PSIF)
- Utils
- Add
modal2polesandpoles2modalfunctions to convert between modal parameters and system poles
- Add
- Add
v0.3.0
Breaking change
v0.2.0
Breaking change
- Refactor of the visualization extension following issue #3. Theming has been removed from the plotting functions (function
theme_choiceis now exported).
v0.1.2
- Bug fix in
OneDMesh(solved issue #2)
v0.1.1
- Bug fixes in State space solvers
v0.1.0
- Initial release of the package