DamBASE - Basement
Contents
1. Residual ACF plot Sinn r 7 ACF 95 Confidence 0 8 u 0 6 lt 0 4 0 2 0 20 40 60 80 100120140160180200220240260280300320340360380400 lag days Fig 4 19 AT Residual ACF 4 9 9 AT Residual normal probability modes scientific only This analysis tool displays a residual normal probability plot VAW ETH Z rich 53 DamBASE User Manual Version 1 0 Residual normal probability ba La bd o Residuals Normal Distribution Standardized Residuals DROoOrnN w 3 2 1 0 1 2 3 Probability Standard Deviation Fig 4 20 AT Residual normal probability 4 9 10 AT Residual time series plot modes scientific only The goal of this analysis tool is to display the date related residuals Residual time series plot Pin Residuals prediction 24 95 Interval Out of limit 3 12 2 o Extrapolations a 6 6 07 03 25 10 13 06 31 01 19 09 09 05 1992 1992 1993 1994 1994 1995 date Fig 4 21 AT Residual time series plot VAW ETH Z rich 54 DamBASE User Manual Version 1 0 Chapter 5 Regression model setup for prediction 5 1 Introduction The step prediction s the last of the three main work flow steps The purpose of this step 1s to use the previously calibrated and validated regression model and to do a prediction based on new measurement data added to the data of the previous two periods in order to analyse the dam behaviour figure 1 2 on page 8 5 2 Anal2. ja DamBASE Regressor Formula Definitions Identifier _1d heat equation on depth 6 Formula _tempid 6 Assignment category Parameters convolution_time_steps 365 is startup time E depth 6 Please define a unique parameter name is startup time new_parameter diffusivity 0 1 is startup time _ startup 180 5 is startup time Definition dambase temp thermal oned date_ col_ depth arg depth thermalDiffusivity arg diffusivity convolutionTime arg convolution_time_steps Regression ER PENDULUM sin 1s T1 W_LEVEL T2 W_LEVEL T3 W_LEVEL T4 W_LEVEL Fig 4 8 Create new parameter VAW ETH Z rich 44 DamBASE User Manual Version 1 0 CH JailIDASE Regressar Formula Dein Identifier _1d heat equation on depth 6 Formula _tempid 6 Assignment category Parameters convolution_time_steps 365 is startup time E depth 6 is startup time diffusivity 0 1 is startup time startup 180 5 is startup time pa En GE m En ee ee new_parameter 5 5 is startup time Definition Regression Bar PENDULUM sin 1s T1 W_LEVEL T2 W_LEVEL T3 W_LEVEL T4 W_LEVEL Fig 4 9 Use of new parameter in field Definition 4 7 Creating new regressors A new regressor can be created by derivation of an existing regressor by click on the d button on the Regressor Formula Definitions tab of the Formula editing wizard page figure 4 10a As changing
3. 1 Navigation Working Current operation area area mode Fig 1 3 User interface concept VAW ETH Z rich 10 DamBASE User Manual Version 1 0 Chapter 2 Quick go through 2 1 Goal The objective of this chapter is to give the user an idea of how the main work flow looks like by covering all steps necessary to end up with a regression model which can be used for prediction observation The data set used within this chapter is from the Schlegeis dam in Austria as published at ICOLD benchmark 2001 and provided as an attachment to the software package 2 2 Project setup A new project can be created by clicking on the menu item Project gt Create project Figure 2 1 shows the project set up wizard for project information and the preferred operation mode the default basic user mode is used for this quick go through Finally we end up with a project overview as shown in figure 2 2 on the following page a DamBASE a DamBASE 2 mie Choose mode Please choose the mode in which you want to run the software You can switch between the modes later by using the Mode menu Project information Name Schlegeig Peonssnssussnsnussnnsunsunsnnenseen Description This is the Schlegeis observation project For more information see internal documentation 123 122 Scientific mode Dam operator Zillertal lt Back Cancel a Project information b Operation model selec
4. 105172 0 0769911 0 0812908 0 0214848 Std Coefficient Fig 4 3 AT Regressor overview with some editable regressors VAW ETH Z rich 34 DamBASE User Manual Version 1 0 Hydrostatic Seasonal cos 1s T1 W_LEVEL Regression PENDULUM cos 2s sin 2s sin 1s cos 1s T1 W_LEVEL T2 W_LEVEL T3 W_LEVEL formula T4 W_LEVEL Fig 4 4 Before editing of the model clicking on the m button allows to derive the model definition for modification VAW ETH Z rich 35 DamBASE User Manual Version 1 0 derived from Hydrostatic Seasonal cos 1s T1 W_LEVEL Custom model definition PENDULUM cos 2s sin 2s sin 1s cos 1s T1 W_LEVEL T2 W_LEVEL T3 W_LEVEL T4 W_LEVEL Fig 4 5 Derived model definition ready for editing VAW ETH Z rich 36 DamBASE User Manual Version 1 0 Regressor overview Sait _ enable sorting Regressor t stat p value VIF Coefficient StdError Std Coefficient 1 Intercept 427 502 0 nan 42 8908 0 100329 nan 2 Y cos 2s 5 10446 3 75605e 07 1 52766 0 354244 0 0693989 0 0201594 3 Y sin 2s 0 382553 0 702107 1 9323 0 0298789 0 0781038 0 0016992 4 Y sin 1s 56 4594 0 7 65255 8 77556 0 155431 0 499062 E V cos 1s 109 632 0 1 65486 7 91872 0 0722303 0 450642 6 Y TI W_LEVEL 113 495 0 6 67409 25 1838 0 221894 0 936888 7 Y T2 W_LEVEL 61 2359 0 3 29373 7 60847 0 124249 0 355112 8 V T3 W_LEVEL 14 7626 4 64539e 46 2 66394 1 55262 0 10
5. 9 Analysis tools AT supporting calibration and validation 491 ATSUmMmMaNy A a RE OS ESE RSS 4 9 2 AT Regression overall quality 6 4 9 3 AT Regression summary 1 2 22mm 4 9 4 AT Regressor overview 2 2er 4 9 5 AT Regressor time Seres VAW ETH Z rich Version 1 0 DamBASE User Manual 4 9 6 AT Physical behaviour 0 0 ee ee o 4 9 7 AT Tukey Anscombe plot residual v s fitted 498 WAV Residual ACF 23 2 re a ran Rw Bo 4 9 9 AT Residual normal probability 2 2222200 4 9 10 AT Residual time series plot 22 22mm 5 Regression model setup for prediction Sch ANMWOCUCHION y tc Dee ee A ee ee 5 2 Analysis setup and data model selection 5 3 Regression model setup 222 2 CE m m ne 5 4 Analysis tools AT supporting prediction IV Report generation and export of regression results 6 Creating reports 7 Export regression results V Save Load project 8 Introduction 9 Saving project 10 Loading project Bibliography VAW ETH Z rich Version 1 0 56 57 58 59 60 61 61 61 DamBASE User Manual Version 1 0 Acknowledgements e The development of the software is financially supported by the Swiss Federal Office of Energy SFOE The software was written at VAW ETH Zurich by Marco Gerber in collaboration with Marius B hlmann and David Vetsch The software is a reimplementation and extension of the Soft
6. T_H12_DO Behavior indicator PENDULUM v Influence factor edit Overview choose analysis tool u Time series overview a 3 ad v era Value PENDULUM z PENDULUM 72 z 64 56 3 48 E w ven a 40 24 32 40 48 56 64 72 PENDULUM 25 10 07 10 19 09 02 09 14 08 27 07 x ad Basic Fig 2 8 Data model overview with analysis tools loaded Regression model setup A new regression model can be created by right clicking on the analysis and choose Create regression model on the context menu Next a name and a description can be chosen figure 2 9 It is recommended to provide this information to distinguish between multiple regression models are set up and compared to each other Note that not only the name can be changed later time but the description as well VAW ETH Z rich 17 DamBASE User Manual Version 1 0 Regression information Name hydrostatic seasonal Description This is the first approach of a regression model with hydrostatic and seasonal regressors Fig 2 9 Regression model set up Regression model time range selection On the next wizard page a time range for the regression period can be chosen figure 2 10 As this 1s the regression model set up for calibration it is recommended to choose a time range which allows to have some validation data available after the regression period In this example 5 years are available which are split up into
7. are implemented in semi infinit domain with a convolution integral Because small depths near the surface respond to high frequency variations it is recommended that the follow ing depth values are chosen depending on the frequency of the data Tab 4 4 Frequency of the data for a given depth from the surface of the wall lt week lt month 4 MN EB Tab 4 5 Temperature regressors Identifier direct temperature direct_temp Direct use of temperature assigned influence factor 1D heat equation on depth n templd n One dimensional heat equation for a depth of n Time averaging regressors Regressors for averaging time Tab 4 6 Average regressors Identifier Average with n time steps avg n Average with n time steps for influence factor direct average direct_avg Direct use of influence factor Delay regressor Regressor for time shifting of n steps Tab 4 7 Delay regressor Delay of n time steps delay n Apply delay of n time steps to influence factor Extra regressors Regressors not related to any other category See description in table 4 8 VAW ETH Z rich 39 DamBASE User Manual Version 1 0 Tab 4 8 Extra C __ Identifier Formula 0000 Formula Description 1 x gt 0 steps 100 x steps 100 x lt x gt 0 sqrt ramp 100 sgrt ramp 100 ramp x A X lt spline spline Spline function X and Y coordinates can be pro
8. assignment dialog Data model analysis Some predefined analysis tools related to the data model have been loaded on the working area figure 2 8 As 1t s key to know how the data looks like for succeeding regression model set up three analysis tools allow to get more insight into the data Therefore the analysis tool Time series overview can be used to make sure that for instance the temperatures have been assigned correctly value range of the y axis as well as the periodic seasonal pattern over time Also outliers e g caused by incorrect working sensors may be detected with this plot see figure 3 3 in section AT Time series overview Analyse values as time series on page 26 On the other hand values depending on the operation of the dam such as the water level can be analysed better on a scatter plot as provided by the Influence factors analysis tool This graph tells a user how a dam pendulum behaves physically related to a change in water level VAW ETH Z rich 16 DamBASE User Manual Version 1 0 File Project Mode About Navigation A Data Analysis p Schlegeis 4 a calib valid Data source C DamBASE data corrected Schlegeis calibration validation_corr csv duaisi d data model Date column DATE choose analysis tool A Behavior Indicators Water Level Temperature Influence factors T1_H12 UP f T_H12_MI
9. cases where a regressor requires to work on an influence factor a can be used to highlight the influence factor within the Formula For instance if this property looks like T5 and later this regressor gets used within a regression formula and is assigned with water level column called W_LEVEL the final formula extends to TS W_LEVEL Assignment category This property is used to identify the assignment category a regressor belongs to Some assignment categories require to have an influence factor of a certain assignment of the data model water level or temperature and some do not see Data model setup and column assignment on page 25 Tab 4 9 Influence factors by assignment category Assignment category no influence factor water level temperature any influence date column is used assigned assigned factor influence factor influence factor Parameters and Definition Property Definition contains an R script being the implementation of the Formula To parametrize this formula user defined Parameters can be declared For example if the script requires a value as an argument such as the degree of a polynomial regressor this value can be passed to the script by adding a parameter called degree Within the script this parameter can be accessed by a preceding arg followed by the name of the parameter The final part of the script in this case would be arg degree This concept allows customizing a regressor without having to chang
10. measurements represent the behaviour of the dam such as pendulum whereas others have an influence on its behaviour The measurements that represent the behaviour are called behaviour indicators whereas those representing the influences are called influence factors With regard to the regression model setup this separation is key and therefore defined by the data model Further some data model related analysis tools such as AT Influence factors require to have such a separation It is important to note that only selected measurement data selected by column of assigned columns is used for further steps For example 1 e any kind of analysis done with analysis tools as well as for regression model setup and calculation 3 2 2 Best practise for data model setup After a regression model has been set up the data model cannot be changed any more but still analysed Therefore it is necessary to think about the purpose of the data model which can either be used for calibra tion and validation or for prediction observation at the time of data selection In the first case it might be VAW ETH Z rich 25 DamBASE User Manual Version 1 0 useful to assign all columns to either behaviour indicators or influence factors whereas a data model used for prediction observation may have only those columns ass gned which are used by the regression model 3 3 Analysis tools AT 3 3 1 AT Summary Overview of number of measurements AT Summary
11. of the assignment category is prevented by the software it is necessary to choose a regressor of the corresponding assignment category Note that changing the assignment category after creation of the regressor would cause incomplete regressor definition VAW ETH Z rich 45 DamBASE User Manual Version 1 0 Regressor Formula Definitions Men A Assgament category Parameters factor 1 is startup time Definition dambase seasonal sin date_ factor arg factor Regression a PENDULUM sin 1s cos 1s T1 W_LEVEL T2 W_LEVEL T3 W_LEVEL T4 W_LEVEL a Derivation d of existing regressor Regressor Formula Definitions Identifier special_kin Formula special_sin Assignment category Parameters factor 1 is startup time multiplicator 42 is startup time x lt arg multiplicator arg factor dambase seasonal sin date_ factor x Regression formula PENDULUM sin 1s cos 1s T1 W_LEVEL T2 W_LEVEL T3 W_LEVEL T4 W_LEVEL b New regressor with additional parameters and extended Definition property Fig 4 10 Introduce new regressor by derivation of an existing regressor VAW ETH Z rich 46 DamBASE User Manual Version 1 0 4 8 Algorithm selection The following algorithms are currently available Ordinary least squares OLS Prais Winsten estimation PWE Ridge regression RR and Principal components regression PCR which are selectable on the cor res
12. through it will not be used and thus is not enabled More about the water level scaling strategy can be found in section 4 2 1 on page 31 VAW ETH Z rich 12 DamBASE User Manual Version 1 0 Analysis information Name calib valid This is the analysis for calibration and validation Fig 2 3 Create analysis CH Navigation 4 p Schlegeis a calib valid General information Name calib valid Description This is the analysis for calibration and validation Water level scaling Is enabled no Full supply level 0 Minimum operating level 0 Fig 2 4 Analysis overview VAW ETH Z rich 13 DamBASE User Manual Version 1 0 a DamBASE Water level scaling strategy For the water level scaling you can define the minimum operating level as well as the full supply level of the dam Is enabled Full supply level 0 Minimum operating level 0 Fig 2 5 Water level scaling strategy Data model setup Next the data model needs to be created by right clicking on the analysis object created in the previous step In the dialog the data set file can be selected as well as the CSV delimiter figure 2 6 on the next page On the next wizard page the date column has to be chosen and a date format has to be specified standard format patterns are possible To continue the chosen format has to be applied to the dataset and the date values in the table view will be updated In the next step the columns tha
13. 0 Analysemethoden f r die Vorhersage und Kontrolle des Verhaltens von Talsperren Tatin et al 2013 Tatin M Briffaut M Dufour F Simon A Fabre J P amp Rousset B 2013 Thermal deformation of concrete dams justification clarification and improvement of statistical analysis Proc 17th biennial conference of the British Dam Society Dams Engineering in a social and environmental context A Pepper Ed Leeds Weber B 2002 Weber B 2002 Vorhersage des Verhaltens von Talsperren mit Hilfe des Soll Ist Vergleichs Statistischer Teil Retrieved on 26 05 2015 from http www dambase ethz ch Vorhersage_Verh_ Talsperren_Statistik B_Weber pdf VAW ETH Z rich 62
14. 3 2 2 Best practise for data model setup 0 0022 eee 3 3 Analysis tools AT u 2 0218 ow a RS A ee Ee ee oe dde ow de 3 3 1 AT Summary Overview of number of measurements 3 3 2 AT Time series overview Analyse values as time series 3 3 3 AT Influence factors Analyse values in relation to a behaviour indicator III Regression model set up for calibration validation and prediction 4 Regression model setup for calibration and validation AL AdU ON 2 2 0 wu a ES hee eee Eh SS eee 4 2 Analysis setup and data model selection 2 2 0 0 200048 4 2 1 Water level scaling strategy 2 2 oo onen 4 2 2 Operation mode related templates n n noa a 4 2 32 Avatlable templaleS s 2 a EH 2 243 a E 4 2 3 1 Hydrostatic Seasonal Time EDF 2 22 2222 4 2 3 2 Hydrostatic Seasonal zu 2 3 ech Oh we Ja asia ep 4 2 3 3 Hydrostatic Thermal 2 2 2 CC Co Emo 4 2 3 4 Water level polynom 3rd degree 2 2 2 22 2 m nn nn 4 2 3 5 Water level polynom 4th degree 4 2 3 6 Seasonal with scaling 2 0 4 3 Enabling disabling regressors 2 Comm 44 AvallaDlestestessofs ru me aa as A Ee ee Oe da 4 5 2SUUCIITS Of ATCOTESSON daria ae Mara Se Er ee ee J A ACUSTOMMZIN TCOTCSSONS e ui a A A A a arg oe sd Ay Creams MEW ESCTESSOLS zesce AEREA A a ee 8 ALSO selection e E e Bn ee eS he ee Se A ee Se RE Se 4
15. 3 2 on page 25 Due to the nature of the work flow many of such regression models can be set up on the same data model In order to keep this as a unit one data model and all regression models depending on it all these elements are finally covered within an analysis object Thus many of these analysis objects finally make up the project Furthermore any kind of analysis data analysis as well as regression analysis can be done by using analysis tools The aim of an analysis tool s to do one distinct analysis on either the data model or regression model As the kind of analysis a user is interested in at a certain time depends on personal preferences as well as the context these analysis tools can be chosen freely by the user more on this s explained in section User interface concept on page 9 VAW ETH Z rich 7 DamBASE User Manual Version 1 0 Analysis 1 Data model Regression model 1 Regression model n Analysis n Data model Regression model 1 Regression model n Fig 1 1 Basic concepts relation 1 4 Main work flow With the current version of DamBASE a certain period t of the data is used for the setup of the regression model Based on this model a prediction can be carried out and all related indicators are referred to as prediction Nevertheless a new approach is proposed here where the main work flow is split up into three chronologically ordered steps called calibration validation and prediction s
16. 4 years for calibration and 1 year for validation remaining data set VAW ETH Z rich 18 DamBASE User Manual Version 1 0 Time range Please choose the time range for which the regression should be applied From 1992 01 01 To 1996 01 01 v Fig 2 10 Regression model time range selection Regression model behaviour indicator selection The behaviour indicator for this regression can be selected on the successive wizard page figure 2 11 See chapter Regression model setup for calibration and validation on page 30 for more about this topic VAW ETH Z rich 19 DamBASE User Manual Version 1 0 a DamBASE Behavior indicator Behavior indicator PENDULUM Fig 2 11 Regression model behaviour indicator selection Regression model template selection Finally the regression formula can be set up on the next wizard page On the tab panel Model Selection an appropriate model template 1 e a formula can be chosen figure 2 12a In this quick go through we choose the model template Hydrostatic Seasonal to get a useful starting point for the first regression model This model contains regressors sin cos describing a seasonal effect as well as Chebyshev regressors describing the water level Further on the tab Model Edit the formula can be changed in terms of adding or removing single regressors figure 2 12b By clicking Finish the regression is carried out 1 e computed and an overview with severa
17. 5172 0 0769911 19 V T4 W_LEVEL 4 65673 3 5061e 06 2 08483 0 378549 0 0812908 0 0214848 Fig 4 6 All regressors are selectable 4 4 Available regressors This section describes all regressors available in DamBASE If any regressor not listed here is required a new one can be derived from an existing one and modified to any new requirement see Customizing regressors on page 44 and section Creating new regressors on page 43 for how to do this Because some assignment categories require to have an influence factor of a certain assignment any direct_x regressors need such an influence factor on instantiation This 1s important if you intend to use a temperature influence factor directly then a direct_temp regressor needs to be used rather than direct_avg because in this case only temperature influence factors are available when using such a regressor within the formula Also the software prevents to use a regressor with a certain influence factor required 1f no such influence factor 1s assigned on the data model All time related regressors use time steps for their calculations This 1s important to note because depending on the distribution and frequency of the data some approaches will not work appropriate if the data properties do not fulfil certain conditions VAW ETH Z rich 37 DamBASE User Manual Version 1 0 Water level regressors Tab 4 1 Water level regressors Chebyshev polynom of Ti Chebyshev polynom up to degree
18. 583 0 0935857 0 0848707 0 0816976 0 0223224 Fig 4 15 AT Regressor overview of a modifiable model definition 4 9 5 AT Regressor time series modes basic scientific Version 1 0 AT Regressor time series displays multiple graphs of either each regressor figure 4 16a or parts of the polyno mial regression formula grouped by assignment category figure 4 16b Further regressor results regardless of whether they are summarized by assignment category or split up into single values can be displayed scaled or not scaled In the first case all regressor values are multiplied by their corresponding coefficients VAW ETH Z rich 50 DamBASE User Manual Version 1 0 Regressor time series Sian Grouping 9 splitted C by assignment category Scaling 9 scaled C not scaled E sin 1s 3 cos 1s n T1 W_LEVEL 3 T2 W_LEVEL 5 T3 W_LEVEL tn E T4 W_LEVEL Uv 25 10 07 10 19 09 02 09 14 08 27 07 1992 1993 1994 1995 1996 1997 date a Split into regressors Regressor time series ern Grouping splitted 9 by assignment category Scaling 9 scaled 5 not scaled Seasonal Water Level 72 36 48 40 32 regressor values scaled 25 10 07 10 19 09 02 09 14 08 27 07 1992 1993 1994 1995 1996 1997 date b By assignment category Fig 4 16 AT Regressor time series VAW ETH Z rich 51 DamBASE User Manual Version 1 0 4 9 6 AT Phys
19. Laboratory of Hydraulics Hydrology and Glaciology ofthe Swiss Federal Institute of Technology Zurich DamBASE Dam Behaviour Analysis Software Environment User Manual Version 1 0 Commissioned by Swiss Federal Office of Energy SFOE VAW 0849 Zurich May 2015 DamBASE User Manual Version 1 0 Contents I Introduction 5 1 Introduction and concepts 6 bl Purpose of the Sollware 4 4 estaca Bea Seh Be ee eel ds Be ee Se 6 1 2 Stucco tie mad we dae ee ae Oe ee ee el 6 LS BasiciCOMCEPDES mala EE AA AAA A ee 7 EF M n Work OW a 32 ma a A AAA A 8 1 4 1 Calibration and validation 2 2 0 2 0 eee ee eee 9 a2 Predici ns are 6 6 26 hee Sk Soe Pee ee BOSS BBS SSS 9 1 3 SOPEerauon modes uti Zar eet wes es we ee ne ee os Se ae ee e E 9 LL Baste user mode 2 2 2 245 2 AAA eee eher Eier 9 1 3 2 SCIEDUNGUSERMOdE sry Lane a Br a Eee dee 9 1 6 User INterrace CONCEDE prea ee ee ee ed OSG ne een eh ws es 9 2 Quick go through 11 Ze OA ir ee Fea Be ae an O a o a ys ee een e G 11 222 Y A A A NA ne ee a 11 2 3 Set up analysis for calibration and validation 2 CC EEE En 12 2 4 Setup analysis for predicho 23 II Data model setup 24 3 Data model setup 25 3 1 oduc om a ok he Bee as OS ok ws a ae a ae es SE De der 25 3 2 Data model setup and column assignment 0 00 ee eee ee 25 3 2 1 Behaviour indicator and influence factors 222 2 2 EEE 29 VAW ETH Z rich l DamBASE User Manual
20. N full supply level normal water level Ride minimum operation level H water level of measurement J Seasonal s 2 3333 J number of day in year Notes Generally the linear term s used for the irreversible effects Also note that 1t is recommended to use only a subset of the terms describing time effects References Simon et al 2013 Tatin et al 2013 4 2 3 2 Hydrostatic Seasonal Model equation Yo aj sin s 05 cos s a3 1 H a4 D H as T3 H 46 Ta H Scaling Hydrostatic 7 Chebyshev polynomial of degree n H water level of measurement e e J Seasonal Ss 20 36575 J number of day in year 4 2 3 3 Hydrostatic Thermal Model equation Yo a1 T H a D H 43 7T3 H 44 Ty H YD aspi 0 VAW ETH Z rich 32 DamBASE User Manual Scaling Hydrostatic T Chebyshev polynomial of degree n H water level of measurement Temperature 0 temperature of measurement 1 4 2 3 4 Water level polynom 3rd degree Model equation Yo ai H a H a3 H Hydrostatic H water level of measurement 4 2 3 5 Water level polynom 4th degree Model equation Yo ai H a H az H a H Hydrostatic H water level of measurement 4 2 3 6 Seasonal with scaling 2 Model equation Yo a sin s a2 cos s az sin 2s as cos 2s Scaling e e J Seasonal S 27 365 35 J number of day in year VAW ETH Z rich Version 1 0 33 DamBASE User Manual 4 3 Enabling disabling reg
21. User interface concept on the current page for how to switch between the two modes 1 5 1 Basic user mode The basic user mode supports regression model set up with a reduced set of regression model templates These templates cover best practice models and abandon special case regressors Further the regression can be ana lysed exclusively with analysis tools related to a physical view of the model This mode is the default mode when you start up the software 1 5 2 Scientific user mode The scientific user mode provides access to the full feature set of the software Thus it covers all functionality of the basic user mode including special model templates with extended regressors as well as statistics related analysis tools 1 6 User interface concept The user interface is split up into a navigation area on the left and a working area on the right These two areas are separated by a movable handle allowing to resize the areas 1 e depending on user preferences figure 1 3 A working area is always linked to the currently selected object in the navigation area The working areas for data and regression model are split up into three parts a configuration area on the left upper side and two VAW ETH Z rich 9 DamBASE User Manual Version 1 0 analysis areas one below and a second on the right of the configuration area also separated with a movable handle On this two working areas a user can load analysis tools into any tab It is pos
22. dentifier pattern a DamBASE Water level scaling strategy For the water level scaling you can define the minimum operating level as well as the full supply level of the dam Is enabled Y Full supply level 1770 Minimum operating level 1683 mn Coma Fig 4 2 Analysis wizard page for water level scaling definition Choosing appropriate template In order to set up a regression model a user can choose from a list of predefined model templates Due to the fact that every dam has an individual behaviour a regression formula setup out of such a template is only meant to be a starting point As some templates assume that certain preconditions are fulfilled e g temperature related influence factors are assigned on the data model it 1s important to choose a template which works with the corresponding setup If any preconditions are not fulfilled DamBASE prevents a user from using that template VAW ETH Z rich 31 DamBASE User Manual Version 1 0 4 2 2 Operation mode related templates Some regressors are a special case of others e g polynomial v s Chebyshev If such special regressors are used within a model template these templates are available in scientific user mode only 4 2 3 Available templates 4 2 3 1 Hydrostatic Seasonal Time EDF Model equation Yo 0 tta tag tas t4 tace az Ztag Z2 a9 Z3 tapoz a11 cos s aj2 sin s 413 cos 2s 414 sin 2s Scaling a _ _RN H Hydrostatic Z bN Ri R
23. e Regression Primary Secondary _ Regressor time series choose analysis tool gt v A Summary Grouping splitted by assignment category id La Lu Scaling PENDULUM Prediction 95 Interval e Out of limit 9 scaled not scaled sin 1s cos 1s T1 W_LEVEL T2 W_LEVEL T3 W_LEVEL T4 W_LEVEL PENDULUM o Extrapolations y Y Y 25 10 0710 19 09 02 09 14 08 27 07 1992 1993 1994 1995 1996 1997 date regressor values scaled FO O A E LE ARTE 25 10 07 10 19 09 02 09 14 08 27 07 Basic Fig 2 13 Regression model overview Regression model analysis By setting up a data model a set of analysis tools AT become available for the analysis of the regression results figure 2 13 The area is split up into a configuration overview part including an analysis part on the left and an additional analysis part on the right Depending on the current operation mode different analysis tools are available by default In basic user mode they are related to physical analysis only The AT Summary shows the result of the regression as a time series plot split up into regression left and validation period right separated by a gray vertical line To get an idea of how well the model describes the measurements MS Res on AT Regression summary represents the mean square error in the unit of the behaviour indicator be
24. e the script which might be error prone Start up time Some regressors might require to have a start up time before they are valid e g 1D heat equation regressors In such cases a flag called is start up time can be selected next to the parameter property to define a parameter as being a start up value in time steps Note that in these cases the minimum start date of the regression will be changed automatically to a later date 1 e original date start up time Available properties in Definition As mentioned above a Definition consists of a full working R script and can be parametrized by user defined parameters Besides this other properties are available such as the date column and in cases of having an VAW ETH Z rich 42 DamBASE User Manual Version 1 0 influence factor assigned the entire column of this influence factor can be accessed as well e g Assignment category is set to Water level the water level column will be available within Definition through expression col_ Tab 4 10 Properties available in Definition property usage always available only available if an influence factor is required see table 4 9 user defined parameter xyz arg oz yes date column dae vws SS influence factor o es VAW ETH Z rich 43 DamBASE User Manual Version 1 0 4 6 Customizing regressors Regressors can be customized on the corresponding Regressor Formula Definitions tab of the model set up wizard page
25. e used to predict new measurements This can be done by creating a new analysis object with an extended data set containing measurements of the calibration and validation period as well as new measurements for prediction and a regression model setup with the formula of the before validated model The time range in this case now covers the entire time range of the validated model calibration and validation see figure 1 2 on page 8 plus the period to predict It is now possible to use the same analysis tools as above but this time for the analysis of the more recent behaviour of the dam VAW ETH Z rich 23 DamBASE User Manual Version 1 0 Part Il Data model setup VAW ETH Z rich 24 DamBASE User Manual Version 1 0 Chapter 3 Data model setup 3 1 Introduction Before the setup of the regression model it is important to get an idea of how the underlying data set looks like Besides this knowing what column represents which values facilitates the setup of a regression model Consequently model templates can directly be used provided that it is known which column contains water level measurements and which columns contain the temperatures values for instance Thus a certain degree of automation can be achieved 3 2 Data model setup and column assignment 3 2 1 Behaviour indicator and influence factors For any dam under observation some measurements are collected such as pendulum water level as well as temperatures Some of the
26. ee figure 1 2 calibration validation prediction t Fig 1 2 Separation into calibration validation and prediction VAW ETH Z rich 8 DamBASE User Manual Version 1 0 1 4 1 Calibration and validation The goal of calibrating a regression model is to define a regression formula which is able to describe the behaviour of a dam as good as possible whereas validation makes sure that the model can describe values not part of the calibration period Both steps are done with a data set for which it is assumed that the dam works correctly data set A on figure 1 2 on the facing page As finding an appropriate model may be an iterative process multiple models model 1 and 2 can be set up and compared based on statistical and physical indicators 1 4 2 Prediction In step prediction new measurements data set A are compared to values predicted by the regression model from the calibration and validation period model 2 If indicators compared to the calibration and validation period differ too much it means that either the regression model is incomplete or the dam is behaving abnormal 1 5 Operation modes DamBASE supports two operation modes for setting up regression models and doing regression analysis The core idea is to allow a user not being familiar with the details of statistical approaches to set up a regression model that can be used for dam behaviour analysis Please note that the basic user mode is the default mode See
27. error between measured and predicted values The same property for the validation period is called MS Res prediction It is important to note that in the cases where MS Res and MS Res prediction differ too much it means that there are either some influences or some process related effects within the data not considered within the defined regression formula figure 4 1 calibration validation MS Res 2 45 MS Res prediction 18 6 Fig 4 1 Differences between MS Res of calibration and validation period 4 2 Analysis setup and data model selection For calibration and validation it is recommended to use a data set which can be split up into a period for cal ibrration 1 e for the setup of the regression model and a successive period with enough values for validation VAW ETH Z rich 30 DamBASE User Manual Version 1 0 Due to the ability of DamBASE to keep multiple models within a single analysis object 1t may be useful to set up more than one regression model and compare them by switching between the corresponding working areas 4 2 1 Water level scaling strategy DamBASE allows for scaling of the water level to minimum operating and to full supply level This feature can be enabled or disabled Because these values are related to a certain dam they can be defined on the corresponding analysis figure 4 2 Note that scaling with this strategy is only used by polynomial water level regressors with scaled min max i
28. es heavy use of regression analysis it is assumed that a user is familiar with the basic concepts of this topic An introduction in German can be found in Weber B 2002 1 2 Structure of the manual Chapter Introduction and concepts is intended to give you an overview of the basic concepts as well as in troduces the main work flow As DamBASE supports two different operation modes you also learn how to distinguish them and what the core idea of this separation is Chapter Quick go through on page 11 finishes the first part with an example covering all steps necessary for setting up a regression model This part also introduces all screens necessary for understanding what features the application provides Therefore if you are not familiar with the software yet t is recommended to read this part first before going on to proceeding parts Part Data model setup on page 24 explains in more detail how a data model is set up and what analysis tools are provided to get a better understanding of the data In part Regression model set up for calibration validation and prediction on page 29 chapters Main work flow Regression model setup for calibration and validation and Regression model setup for prediction cover the main work flow and discuss the separation of calibration validation and prediction more detailed How to generate a report and export data for post processing as well as saving and loading a project are ex plained in the two fina
29. ical behaviour modes basic scientific no automatic appearance The aim of this analysis tool 1s to provide a way for getting an idea of how the model describes the behaviour of the dam more precisely the behaviour indicator of the model and therefore the displacement of the dam from a physical point of view With this graph the partial displacement and therefore the physical impact of each influence factor grouped by assignment category can be displayed For example with this analysis tool it s possible to get the values of the applied water level polynomial opposed to the measured water level displayed on a graph figure 4 17 It is important to note that the calculation is only performed on measured values no extrapolations Physical behavior sa ss ata Water Level v 1770 1755 1740 1725 1710 32 40 48 56 64 72 Partial displacement Fig 4 17 AT Physical behaviour 4 9 7 AT Tukey Anscombe plot residual v s fitted modes basic no automatic appearance scientific This analysis tool displays a plot of fitted values v s studentized residuals VAW ETH Z rich 32 DamBASE User Manual Version 1 0 Tukey Anscombe plot residual vs fitted Sian Studentized Residuals Fig 4 18 AT Tukey Anscombe plot residual v s fitted 4 9 8 AT Residual ACF modes scientific only This analysis tool displays an ACF autocorrelation function plot for auto correlation detection on residuals
30. idual time series plot 7 I r hydrostatic M y I nl r CEOE E E i I Currently Fy P Analysis tool toolbar pJi a selected I rpc PENDULUM edit I i e u a I i object a I E o Residuals i BP Regression PENDULUM sin 1s cos 1s T1 W_LEVEL T2 W_LEVEL T3 W_LEVEL i ot a i l I g Formula T4 W_LEVEL I nn g 2 i I I I nn 5 0 iy i Execute Regression r E 1 I Er 2 x z E A Hi LLE I EEE EEEE MM DE A ME A A DE MM ME ME ee ee Ww dnt E ms u a A a 4 i a I g Pri la A A 07 03 25 10 13 06 31 01 19 09 09 05 26 12 A I I pimay ION ily Tabs I II 1992 1992 1993 1994 1994 1995 1995 I 5 y u srooseanaiais tooi z 1 Analysis tool selection al NalySIS we als N io mn mn m u En EEE EEE EEE EI T tool I A 7 I a A A IE IE A A EE ME A A DM ME ME ME A MM E E ME ME A A Eee ee BEE I I Summary I Residual normal probability I I i I A i Oak pol Jan JE E Jeu my aaa fi ot rt E I PENDULUM I a o Residuals 0 i 72 3 Prediction 7 I 3 3 Normal Distribution 7 i mn 4A 95 Interval I i g I I i 3 m Out of limit I e U n 1 So I i Bo lal I I o o Extrapolations A i 1 I I 1 24 J i 5 2 y nl A Ls Le bs a I 8 n I 25 10 07 10 19 09 02 09 14 08 27 07 I n 3 0 p I 1992 1993 1994 1995 1996 1997 I 3 2 1 0 1 2 3 l 7 I date I o Probability Standard Deviation I I I I iy 1 A I I Analysis 4 Analysis 4 I I area In area Ny 1 Scientific
31. l analysis tools will be loaded onto the working area figure 2 13 VAW ETH Z rich 20 DamBASE User Manual Version 1 0 Hydrostatic Seasonal Time EDF Tatin et al 2013 Simon et al 2013 Hydrostatic seasonal time model according to EDF with minimum and full supply level scaling Hydrostatic Seasonal no specific reference Model with chebyshev polynom for water level and seasonal functions Regression PENDULUM sin 1s cos 1s T1 W_LEVEL T2 W_LEVEL T3 W_LEVEL formula T4 W_LEVEL Custom model definition Regression PENDULUM sin 1s cos 1s T1 W_LEVEL T2 W_LEVEL T3 W_LEVEL formula T4 W_LEVEL b Model Edit Fig 2 12 Regression model selection VAW ETH Z rich 21 DamBASE User Manual Version 1 0 a DamBASE v 1 0 Schlegeis Foka File Project Mode About Navigation Regression Analysis 1 p Schlegeis 4 a calib valid Primary d data model Name hydrostatic seasonal Secondary r hydrostatic se Description This is the first approach of a regression model with hydrostatic and seasonal regressors ai choose analysis tool y From date 1992 01 01 Regression summary A To date 1996 01 01 i SS a E ejoja Behavior PENDULUM cis indicator MS Res 2 35039 Regression PENDULUM sin 1s cos 1s T1 W_LEVEL T2 W_LEVEL T3 W_LEVEL R2adj 0 964797 formula T4 W_LEVEL Observ 1462 un nn Smua edit MS Res prediction 2 2581 z Execut
32. l parts Save Load project on page 59 and Report generation and export of regression results on page 56 VAW ETH Z rich 6 DamBASE User Manual Version 1 0 1 3 Basic concepts Normally the software is intended to be used over a longer time while new data for observation 1s provided frequently During this time the construction of a dam may have changed or new approaches for observation have been introduced Further the proper work flow requires to have multiple regression models available for comparison and optimization All these requirements make it necessary to have multiple regression models kept together and available at any time Within the software a project is meant to keep all this data and regression models together figure 1 1 In order to distinguish different projects a name a description as well as a dam operator can be assigned On the other hand the final regression is done by applying a regression formula a polynomial consisting of a set of regressors on a set of data All these regression related elements except for the data set are covered within a regression model The data set used by the regression 1s further covered within a data model As the software also supports assignment of single columns such as pendulum water level and temperature to their corresponding semantic meaning behaviour indicators and influence factors this assignment 1s also part of the data model more on this can be found in chapter
33. n degree n related to influence factor Polynomial of degree n h Polynomial up to degree n Scaled polynomial of h min max Polynomial up to degree n Influence degree n min oplevel and factor used will be scaled to minimum full supply level operation and full supply level see Water level scaling strategy on page 31 for more on how to define levels Seasonal regressors Note that all seasonal regressors are intended to be used for temperature approximation in cases where no temperature data is available Therefore they are synchronized to the day of the year of the regression start date If other seasonal approaches are required a new regressor needs to be defined section 4 7 on page 45 Also note that sin as well as cos regressors are intended to be used as pairs with the same scaling factor Tab 4 2 Seasonal regressors 2T 30525 sin ns sin ns scaling factor number of day n year IN 365 25 cos ns cos ns scaling factor number of day in year Drift regressors Drift regressors support modelling of long time behaviour e g creep or inelastic deformation of the founda tion Tab 4 3 Drift regressors Direct Directusageoft t In 1 t Logarithm naturalis of 1 t VAW ETH Z rich 38 DamBASE User Manual Version 1 0 Temperature regressors Temperature regressors like 1D heat equation can be used to calculate temperature on a certain depth out of surface values They
34. ponding tab of the model setup wizard page G a DamBASE 2 x a Model Selection Model Edit Regressor Formula Definitions Algorithm Selection OLS ordinary least square 9 enabled Prais Winsten disabled Max iterations 10 Ridge disabled k 0 06 Principal Component Regression disabled Regression ds PENDULUM sin 1s T1 W_LEVEL T2 W_LEVEL T3 W_LEVEL T4 W_LEVEL Fig 4 11 Algorithm Selection 4 9 Analysis tools AT supporting calibration and validation In scientific user mode analys s tools are separated into those which are for analys ng least square fit precondi tions tab second on the right sided analysis area and those for calibration and validation remaining tabs In basic user mode only analysis tools related to calibration and validation prediction are available 4 9 1 AT Summary modes basic scientific AT Summary provides an overview of the regression fit for both calibration and validation periods separated by a gray vertical line displaying out of limit data points It s important to note that the 95 interval 1s an VAW ETH Z rich 4 DamBASE User Manual Version 1 0 indicator of the goodness of fit of the model the smaller the interval the better the fit Therefore the number of Out of interval data points can increase As a consequence 1t is necessary to take MS Res from AT Regression Summary see 4 9 3 on the next page into account S
35. previous model setup are provided within brackets Regression summary em e a SS Res 13615 4 11500 4 MS Res 12 4798 10 5376 MS Res 12 4115 10 4835 R2 0 925518 0 0629121 R2adj 0 925177 0 0631791 F stat 2711 36 10579 2 signif F 0 0 DW stat 0 0184516 0 0617813 Observ 1097 0 SS Res prediction 113001 94181 7 MS Res prediction 103 103 85 9322 Fig 4 14 AT Regression summary 4 9 4 AT Regressor overview modes basic reduced scientific The goal of this analysis tool 1s to provide not only statistical values of each regressor but also to enable or disable regressors For the latter it is important that the model definition is editable see Enabling disabling regressors on page 34 After enabling sorting on the provided checkbox the table can be sorted by clicking on the corresponding table header VAW ETH Z rich 49 DamBASE User Manual Regressor overview enable sorting uma aa ata Regressor t stat p value Intercept 517 448 0 Y sin 1s 69 3824 0 Y cos 1s 122 276 0 Y TI W_LEVEL 139 538 0 K T2 W_LEVEL 62 9999 0 a Oo Bb W N Q VIF Coefficient StdError Std Coefficient nan 42 9903 4 97087 8 77456 1 31726 7 95497 4 32 25 1483 3 14813 7 7257 Y T3 W_LEVEL 18 2882 2 04628e 67 2 06963 1 71152 T4 W_LEVEL 4 81419 1 63196e 06 2 06613 0 393308 0 0830815 nan 0 126467 0 499005 0 0650577 0 452705 0 180225 0 935566 0 12263 0 360
36. provides an overview of how many measurements are contained within the data model as well as how many values are missing Summary Number of missing 0 values Number of available 28496 values Total number of 28496 values Fig 3 1 AT Summary 3 3 2 AT Time series overview Analyse values as time series AT Time series overview plots any assigned data against the date column This helps not only to detect any outliers e g misbehaving sensors as shown in figure 3 2 but also makes sure that columns with a certain value range are assigned correctly For example the graph of a temperature on the air side is expected to have a certain value range e g from 10 C to 25 C and to have a seasonal trend as well If this 1s not the case the assignment might be wrong and needs to be changed Time series overview u ea lata Value T_H12_MI v T_H12_MI 7 8 7 2 x 6 6 H12_MI E 5 4 4 8 4 2 25 10 07 10 19 09 02 09 14 08 27 07 1992 1993 1994 1995 1996 1997 date Fig 3 2 Outlier detected with AT Time series overview VAW ETH Z rich 26 DamBASE User Manual Version 1 0 In case of having any unexplained measurements it might be helpful to load more than one AT below each other see figure 3 3 Time series overview daa value T_H12 MI 7 8 s 32 6 6 i 6 E 5 4 4 8 4 2 25 10 07 10 19 09 02 09 14 08 27 07 1992 1993 1994 1995 1996 1997 date Time series over
37. reating reports DamBASE supports creating reports in formats pdf docx odf rtf and txt markdown A report can be created out of all loaded analysis tools by right clicking on the selected regression model within the navigation area All elements of the report can be sorted on the report creation dialog figure 6 1 E DamBASE File Plot scale factor Font size for PDF only Report User manual report Fig 6 1 Report creation dialog VAW ETH Z rich 57 DamBASE User Manual Version 1 0 Chapter 7 Export regression results For post processing regression results of all loaded analysis tools can be exported to a tab separated text file See figure Export regression results on the current page d data m Create report Export Regression Results Edit Common parameters Edit Time range Edit Behavior indicator Edit Formula Execute Regression Remove Regression Show Regression Fig 7 1 Export regression results VAW ETH Z rich 58 DamBASE User Manual Version 1 0 Part V Save Load project VAW ETH Z rich 59 DamBASE User Manual Version 1 0 Chapter 8 Introduction The ability of DamBASE to save and load projects allows to work continuously on projects by adding new data and regression models The data format for the project data file is XML that allows for saveing and loading different project snapshots and keeping project management flexible As the project file can be of any name pro
38. ressors Version 1 0 Regressors can be enabled or disabled on the Regressor overview analysis tool figure 4 3 After enabling or disabling a new regression can be calculated by clicking on the Execute Regression button right below the configuration area However regressors related to a model template cannot be modified and therefore neither enabled nor disabled In such cases the related model definition needs to be enabled for modification which can be done by clicking on the m button of the corresponding model definition figure 4 4 As a consequence the name of the model definition changes and marks the fact that 1t is no longer a model of a certain template with a preceding figure 4 5 As a result all regressors can now be enabled or disabled 4 6 Regressor overview enable sorting um La to t stat 427 502 Regressor 1 Intercept 2 Y cos 2s 3 v sin 2s 4 sin 1s 56 4594 5 cos 1s 109 632 6 T1 W_LEVEL 113 495 7 T2 W_LEVEL 61 2359 8 T3 W_LEVEL 14 7626 9 T4 W_LEVEL 4 65673 p value VIF 0 nan 42 8908 5 10446 3 75605e 07 1 52766 0 354244 0 382553 0 702107 1 9323 0 0298789 0 7 65255 8 77556 0 1 65486 7 91872 0 6 67409 25 1838 0 3 29373 7 60847 4 64539e 46 2 66394 1 55262 3 5061e 06 2 08483 0 378549 Coefficient StdError 0 100329 nan 0 0693989 0 0201594 0 0781038 0 0016992 0 155431 0 499062 0 0722303 0 450642 0 221894 0 936888 0 124249 0 355112 0
39. sible to load an analysis tool multiple times with different configurations Especially in cases where different aspects need to be analysed and compared it can be helpful to place two analysis tools of the same kind below or opposite each other see figure 3 3 in section AT Time series overview Analyse values as time series on page 26 for example Every analysis tool has a tool bar for moving up and down as well as for removal It s possible to enable or disable certain graphs on plots having more than one graph by right clicking on the corresponding graph within the legend After setting up either a data or regression model by using the wizard some predefined and operation mode related analysis tools are loaded automatically on the corresponding working area It is important to mention that additional analysis tools can be loaded individually Navigation i Regression Analysis I p quick go throu i u a a a a a ee ee lel ME EEE HEN TEE DEE TEE TEE TEE HE HE ME TEE ME ME ME ME ME ME ME ME ME ME ME ME ME ME ME ME ME ME TEE HEN TO i 4 a uncorr calib 5 _ ER Name hydrostatic seasonal Configuration i E Primary Secondary I I Description area MO a 5 5 0 0 0 50 0 2 50 50 2 2 A 4 a corrected C edit tere 1 ML oai nin z E Analysis tool selection f u I r hydrostatic J T From date 1992 01 01 i Sao oc a I I m ioatea i I To date 1996 01 01 a i Res
40. t represent any behaviour indicators mostly the pendulum and the columns which represent influence factors such as water level and temperature have to be assigned The influence factors are grouped into assignment categories namely Water Level and Temperature It is important to note that after the regression model has been set up the assignment done here cannot be changed any more However any kind of analysis on the data is still possible VAW ETH Z rich 14 DamBASE User Manual Version 1 0 Data model C damreg documents usermanual data Schlegeis calibration validation csv CSV delimiter PENDULUM WLLEVEL THI2UP TH12MI 1656 1764 67 52 65 02 01 1992 657 1764 62 l 65 03 01 1992 65 6 1764 56 65 NAMA ann rrr aFranan rr a Load data set dialog Column Column Format d m Y First date 1992 01 01 Last date 1997 12 31 DATE PENDULUM T_H12_UP T_H12_MI 01 01 1992 65 6 3 2 6 5 02 01 1992 65 7 1 6 5 Ss EEE A A e ad b Apply date format dialog Fig 2 6 Data model Loading data and apply format VAW ETH Z rich 15 DamBASE User Manual Version 1 0 a DamBASE EN a Data assignment Behavior Indicators Water Level Temperature Y PENDULUM IE W_LEVEL E 1 H12 UP T_H12 MI T_H12 DO T_H15_UP T_H15_MI T_H15_DO T ATD BA Data DATE PENDULUM W_LEVEL T_H12_UP T_H12_MI T 1 1992 01 01 65 6 1764 67 5 2 6 5 0 4 IM j lt Back Cancel Fig 2 7 Column
41. tion Fig 2 1 Project information and operation mode wizard page VAW ETH Z rich 11 DamBASE User Manual Version 1 0 a DamBASE v 1 0 Schlegeis File Project Mode About Navigation General information p Schlegeis name Schlegeis Description This is the Schlegeis observation project For more information see internal documentation 123 122 Dam operator Zillertal Fig 2 2 Project overview 2 3 Set up analysis for calibration and validation As mentioned in Regression model set up for calibration validation and prediction on page 29 the main goal of calibrating a regression model is to set up a model describing the behaviour of the dam as good as possible whereas validating the model guarantees that the behaviour can be assessed by the model based on subsequent measurements not part of the calibration period Analysis setup An analysis has to be set up for calibration and validation and can be created with a name and a description figure 2 3 by right clicking on the project at the navigation tree Finishing this wizard will lead to the analysis overview as shown in figure 2 4 From here on it is possible to define any Water level scaling strategy figure 2 5 by clicking on the corresponding Edit button which allows to set the minimum operating level as well as the full supply level if any scaling of the water level is later required when setting up the regression model In the case of this quick go
42. tween the measured values and the calculated ones The same is provided for values of the validation period where it is called MS Res prediction It has to be noted that if the mean error of the validation period is different from that of the calibration it means that the prediction may not be of sufficient validity If this is the case it might be because the chosen model cannot entirely describe the real physical behaviour of the dam To make sure this is not the case AT Physical behaviour displays graphs showing how regressors related to a certain assignment category such as water level describing the displacement of the dam grouped by influence factors which are related to this assignment category In the model of this quick go through only the water level is used as an influence factor and thus the applied parts of the formula describing the water level Chebyshev polynomials are plotted against the measured water level values In order to get an idea of how each regressor describes the model value plots either grouped by assignment category or split into single values are provided by AT Regressor time series VAW ETH Z rich 22 DamBASE User Manual Version 1 0 Creating a model for comparison To optimize a model additional models may be set upthat are used for comparison multiple models can be created within the same analysis 2 4 Set up analysis for prediction After a regression model has been validated the model can b
43. ummary gt bd ls PENDULUM 80 Prediction 70 5 60 95 Interval 3 50 e Out of limit u 40 o Extrapolations 30 20 07 03 17 04 27 05 06 07 14 08 23 09 1992 1993 1994 1995 1996 1997 date Fig 4 12 AT Summary 4 9 2 AT Regression overall quality modes basic scientific The aim of this analysis tool is to provide some information about regression quality in textual form Be careful because a good fit may be indicated even if there 1s autocorrelation Thus further steps are necessary to remove the autocorrelation Regression overall quality is Lay la Overall regression quality Autocorrelation Durbin Watson Positive autocorrelation value 0 041636 Overall fit R2adj Good fit value 0 984797 Fig 4 13 AT Regression overall quality Note that Overall fit cannot be used as a single expression about the quality of the model but rather as a hint for how well the model fits on the regression period see table 4 11 for quality limits Tab 4 11 Quality limits x gt 05 andx lt 095 x gt 0andx lt 0 5 Very bad fit VAW ETH Z rich 48 DamBASE User Manual Version 1 0 4 9 3 AT Regression summary modes basic reduced scientific AT Regression summary provides some statistics related information about the created regression model see section 4 1 on page 30 for more about MS Res and MS Res prediction In order to support model optimization relative changes with regard to the
44. vided by the corresponding user parameters as R vectors E g c 1 2 3 Please mind the spaces necessary between the coordinates to work properly VAW ETH Z rich 40 DamBASE User Manual Version 1 0 4 5 Structure of a regressor A regressor in DamBASE consists of the following elements Model Edit Regressor Formula Definitions Algorithm Selection Identifier _1d heat equation on depth 6 Formula _temp1d 6 Assignment category Parameters convolution_time_steps 365 is startup time E depth 6 is startup time diffusivity 0 1 is startup time startup 180 is startup time Definition dambase temp thermal oned date_ col_ depth arg depth thermalDiffusivity arg diffusivity convolutionTime arg convolution_time_steps Regression nn PENDULUM sin 1s T1 W_LEVEL T2 W_LEVEL T3 W_LEVEL T4 W_LEVEL Fig 4 7 Regressor Formula Definition Identifier Identifies a regressor with a symbolic name This identifier will appear in the drop down menu on the model definitions area of the Model Edit tab see figure 4 4 for an example Formula This property assigns a symbolic formula to a regressor However the final formula used by the regressor 1s defined by property Definition Compared to property Identifier this property is meant to be used all over the software to identify the regressor from a mathematical point of view VAW ETH Z rich 41 DamBASE User Manual Version 1 0 In
45. viding enough information on the project object level as well as on all corresponding objects description field is key to keep the overview and knowing at any time what and why something was done while modelling As the state of analysis tools is not saved anything related to changes on analysis tools will not result in a pop up dialog to save the project on closing the application Nevertheless after loading a project and double clicking on either a data or regression model on the navigation area all analysis tools are loaded with regard to the current user mode VAW ETH Z rich 60 DamBASE User Manual Version 1 0 Chapter 9 Saving project A project can be saved by using the menu entry Project Save project Chapter 10 Loading project Projects can only be loaded if no other project is currently opened This means that if you intend to open a project while working on another you first have to close the application and load the project after restart by using the menu entry Project Load project VAW ETH Z rich 61 DamBASE User Manual Version 1 0 Bibliography ICOLD benchmark 2001 ICOLD benchmark 2001 ICOLD benchmark workshop on dam safety in Graz Simon et al 2013 Simon A Royer M Mauris F amp Fabre J P 2013 Analysis and interpretation of dam measurements using artificial neural networks Proc 9th ICOLD European Club Symposium Venezia Swiss Committee on Dams 2010 Swiss Committee on Dams 201
46. view Bee value wav y W_LEVEL 1770 _ 1755 gt 1740 m 1 gt 1725 1710 1695 25 10 07 10 19 09 02 09 14 08 27 07 1992 1993 1994 1995 1996 1997 date Fig 3 3 Multiple AT Time series overview open below each other VAW ETH Z rich 27 DamBASE User Manual Version 1 0 3 3 3 AT Influence factors Analyse values in relation to a behaviour indicator Influence factors can be analysed by plotting them against the behaviour indicator e g plot the water level which depends on the dam operation against the pendulum to get an idea of how they are physically related Influence factors erm Behavior indicator PENDULUM v Influence factor 1770 1755 1740 W_LEVEL 24 32 40 48 56 64 72 PENDULUM Fig 3 4 AT Influence factors with water level plotted against pendulum displacement of the dam VAW ETH Z rich 28 DamBASE User Manual Version 1 0 Part Ill Regression model set up for calibration validation and prediction VAW ETH Z rich 29 DamBASE User Manual Version 1 0 Chapter 4 Regression model setup for calibration and validation 4 1 Introduction As described in section Main work flow on page 8 the main goal of calibrating and validating is to create a regression model describing the behaviour of a dam as good as possible However the quality of a model cannot be measured absolutely but in relation to certain properties One of these properties is MS Res representing the mean square
47. ware DamReg also commissioned by the Swiss Federal Office of Energy SFOE written by Benedikt Weber e The basic methods and ideas used in the software originate from the reference report by the former working group on numerical methods of the Swiss committee on Dams headed by Dr Georges R Darbre Swiss Committee on Dams 2010 Citation Advice e For User Manual Gerber M B hlmann M Vetsch D F 2015 DamBASE User Manual Ver sion 1 0 Laboratory of Hydraulics Hydrology and Glaciology VAW ETH Ziirich Available from http www dambase ethz ch 27 May 2015 e For Website DamBASE Dam Behaviour Analysis Software Environment 2015 http www dambase ethz ch e For Software DamBASE Dam Behaviour Analysis Software Environment Version 1 0 VAW ETH Zurich Gerber M Buehlmann M Vetsch D F 2015 VAW ETH Z rich 4 DamBASE User Manual Version 1 0 Part I Introduction VAW ETH Z rich 5 DamBASE User Manual Version 1 0 Chapter 1 Introduction and concepts 1 1 Purpose of the software The main goal of DamBASE is to perform statistical regression analysis for dam observation Besides this further concepts to simplify the modelling process are also provided such as a simplified user mode as well as a high flexibility in terms of extending and customizing the software in order to deal with problems related to the individual behaviour of a dam As the entire observation process supported by this software mak
48. ysis setup and data model selection As a new data set 1s required an analysis object needs to be created with the corresponding data loaded When assigning columns on the data model 1t 1s recommended to assign only columns of influence factors that are required by the regression model 5 3 Regression model setup An important step here 1s to choose the correct time range for the regression period covering the periods of the calibration validation model including the new measurement data The latter will be used for prediction To apply the same regression model used for calibration validation the identical regression formula has to be set up to make sure that the prediction runs with the calibrated and validated regression model 5 4 Analysis tools AT supporting prediction Besides all analysis tools available for calibration and validation the two most important ATs are AT Regression summary page 49 as well as AT Residual time series plot page 54 as they provide information about MS Res prediction However it s also recommended to check model consistency with other analysis tools especially f deviation appears 1 e unexpected behaviour which might be due to model incompleteness or inconsistency or physic ally based impacts on the dam VAW ETH Z rich 55 DamBASE User Manual Version 1 0 Part IV Report generation and export of regression results VAW ETH Z rich 56 DamBASE User Manual Version 1 0 Chapter 6 C
Download Pdf Manuals
Related Search