Auditoría de modelos numéricos para macizos rocosos
Dr. Alejo O. Sfriso Universidad de Buenos Aires SRK Consulting (Argentina) AOSA
materias.fi.uba.ar/6408 latam.srk.com www.aosa.com.ar
[email protected] [email protected] [email protected]
Auditoría de modelos numéricos para macizos rocosos
Propósito de las auditorías técnicas
2
(Hudson & Feng 2010)
• Evaluar la lógica del trabajo en función de su objetivo • Establecer si los mecanismos, variables y parámetros relevantes han sido considerados • Mostrar que el análisis de respaldo es apropiado dentro de los límites del conocimiento disponible • Considerar si las conclusiones están justificadas en términos de los objetivos del proyecto • Proporcionar rastreabilidad de la información, análisis y decisiones adoptadas La auditoría debe ser independiente, sin limitaciones técnicas ni constractuales, y agregar valor al proyecto
1
monitoring and back analysis, leading to the final design. Similarly, the auditing is applied to the seven design steps discussed in [2], these being the project purpose, key features of the site/rock mass/project, design approach strategy, choice of modelling method and appropriate code, initial design, integrated modelling and feedback information and final design and verification/validation.
The validity of the modelling and design of a rock engineering project will depend on the accuracy of the supporting information
1.3. ‘Soft’ and ‘hard’ technical audits and the audit evaluation
Auditoría de modelos numéricos para macizos rocosos
Note that the technical auditing can be ‘soft’ or ‘hard’, with the characteristics as indicated in Fig. 2. Thus, for modelling, i.e. using one or more of the eight main types of modelling methods shown in Fig. 1, either auditing type can be utilized, depending on the purpose of the auditing. The soft audit can be used initially to support the development of the modelling programme. However, the hard audit is necessary for the total audit evaluation and ability to state whether the modelling is adequate for the purpose.
'Soft' Audit
'Hard Audit'
Audit Evaluation
Obtains the basic information for establishing the essence of the problem
Obtains the detailed information on all the procedures being used
Establishes whether the modelling is adequate to meet the objectives
Ability to present what is being done
Los ocho métodos de diseño en ingeniería de rocas
Ability to state Ability to state whether the the details of modelling is what is being adequate for the done purpose (Hudson & Feng 2010)
Fig. 2. The ‘soft’ and ‘hard’ audits and the audit evaluation.
Objective
Method A
Hudson, X.-T. Feng / International Journal of
Use of pre-existing standard Rock Mechanics & Mining methods Site Investigation
Method B
Method C
Basic Analytical numerical methods, methods, FEM, Sciences 47 (2010) BEM, 877–886 stress-based DEM, hybrid
Method D Extended numerical methods, fully-coupled models
Level 1 1:1 mapping
The two demonstration auditing examples that follow are: Database Integrated (1) thePrecedent procedure for auditing a site expert investigation measurement— type Rock mass systems Level 2 systems, & analyses and classification, approaches, in situ modifications rock stress (soft and (2) auditing the modelling for Not 1:1 mapping other systems RMR, Q, audit) GSI internet-based approaches the design of hydropower caverns at the Laxiwa site on the Yellow River in China (semi-hard audit). Whether a soft or hard audit is appropriate in any given Design based on forward analysis Design based on back analysis situation will depend on the requirement and according to the descriptions in Fig. 2; both forms of audit are demonstrated in the Construction following sections. The limited audit illustrated for the second demonstration thetypeshydropower in (from China Fig. 1. Flowchart of rock audit mechanics for modelling used to support rockcaverns engineering design [2]). is termed ‘semi-hard’ because it is not possible to include the full detail given the limitations on the paper length.
modelling and design and record
he above list, an additional benefit enables highlighting of any ‘guess perience that is potentially subject ther experience and updating as 3
also presented within the context rating the eight basic modelling here are four methods A–D with eft to right, each with two subether the method involves 1:1 g of the project geometry); the sic steps in the rock engineering pects of Fig. 1 can be subject to ning the site and rock mass gation, use of the eight types of used singly, in combination or in the initial design, construction is, leading to the final design. plied to the seven design steps he project purpose, key features of sign approach strategy, choice of iate code, initial design, integrated ormation and final design and
2. Demonstration example 1: the procedure for technically auditing a site investigation measurement—in situ rock stress (soft audit)
Auditoría de modelos numéricos para macizos rocosos
Los 2.1. dosBackground niveles de auditoría técnica
(Hudson & Feng 2010)
The validity of the modelling and design of a rock engineering project will depend on the accuracy of the supporting information
udits and the audit evaluation
ting can be ‘soft’ or ‘hard’, with the ig. 2. Thus, for modelling, i.e. using ypes of modelling methods shown can be utilized, depending on the oft audit can be used initially to e modelling programme. However, or the total audit evaluation4 and delling is adequate for the purpose.
'Soft' Audit
'Hard Audit'
Audit Evaluation
Obtains the basic information for establishing the essence of the problem
Obtains the detailed information on all the procedures being used
Establishes whether the modelling is adequate to meet the objectives
Ability to present what is being done
Ability to state the details of what is being done
Ability to state whether the modelling is adequate for the purpose
Fig. 2. The ‘soft’ and ‘hard’ audits and the audit evaluation.
Objective
Method A Use of
Method B
Method C
Method D
Basic
Extended
2
Fig. 5. (a) 3D model for numerical analysis and (b) calculation model for excavation of the Laxi
Un ejemplo de auditoría: 38 preguntas garantizan la calidad del modelo
Subject Area 3 in Table 3 the numerical code us analyses and presentation mechanics and rock engine increasing and numerical mod 1. THE MODELLING OBJECTIVE The purpose of the modelling? the auditing questions them 2. CONCEPTUALIZATION OF THE PROCESSES BEING MODELLED The sub-system(s) being isolated for study. The physical processes involved. capture the critical aspects in 3. SPECIFICATION OF THE MODELLING CONTENT ensuring that there is a check What are the physical variables, connecting relations, parameters, boundary that there have been checks conditions, initial conditions, etc. that uncertainties are capture 4. MODELLING SOLUTION REQUIREMENTS reasonable, and that the mod What type of model output is required, given the stated modelling purpose? 5. MODELLING SOLUTION TECHNIQUE Subject Area 4 in Table How is the required model output to be obtained? explicitly through the identifi 6. NUMERICAL CODE UTILIZED and the possible need for cor Which numerical code is to be used? How do we know that the code is In the current absence of operating correctly? 7. SUPPORTING MODEL DATA & DATA INPUT METHOD the rock engineering design p J.A. Hudson, X.-T. Feng / International Journal of Rock Mechanics & Mining Sciences 47 (2010) 877–886 881 What are the necessary supporting data? How are they to be obtained? How ing areas such as aircraft are they to be input? illustrated in this demonstr 8. MODEL SENSITIVITY ANALYSIS Fig. 5. (a) 3D model for numerical analysis and (b) calculation model for excavation of the Laxiwa cavern group. answers to the questions lis How does the model output depend on the model input in terms of whether a that an adequate procedur sensitivity analysis is required? 9. PRESENTATION OF MODELLING Subject Area 3 inRESULTS Table 3 concerns the modelling techniquefollowed. The auditing can be Table 2 Is it possible to demonstrate that the numerical is operating correctly? itself: the numerical code used, code the supporting data, sensitivityactivity. Needless to say, con The 11 subject headings for the semi-hard technical audit of the rock mechanics clearly presented? modelling supporting the cavern excavation design for the Laxiwa project on Are the the modelling analyses results and presentation of the results. Because our rockso that any corrective action 10. SOURCES OF ERRORS Yellow River, China. Table 2
Auditoría de modelos numéricos para macizos rocosos
J.A. Hudson, X.-T. Feng / International Journal of Rock Mechanics & Mining Sciences 47 (2010) 877–886 The 11 subject headings for the semi-hard technical audit of the rock mechanics
(Hudson & Feng 2010) modelling supporting the cavern excavation design for the Laxiwa project on the Yellow River, China.
881itself:
mechanics and rock engineering knowledge is continuously
What are the main sources of errors?
Auditoría de modelos numéricos para macizos rocosos
increasing and numerical models are continuously being improved, 1. THE MODELLING OBJECTIVE The purpose of the modelling? 11. MODELLING ADEQUACY the auditing needany to problem be adjusted to3.3. Comments on the Laxiwa 2. CONCEPTUALIZATION OF THE PROCESSES BEING MODELLED Does the modelling seem questions adequate forthemselves the purpose?may Are there The sub-system(s) being the critical in this subject area. The key issues are 5 isolated for study. The physical processes involved.areas? Iscapture any corrective action aspects required? 3. SPECIFICATION OF THE MODELLING CONTENT ensuring that there is a check on the appropriateness of the model, It was noted earlier that What are the physical variables, connecting relations, parameters, boundary that there have been checks on how the model is used, ensuringexample has been termed ‘se conditions, initial conditions, etc. Fig. 5. (a) 3D model for numerical analysis and (b) calculation model for excavation of the Laxiwa cavern group. that uncertainties are captured by sensitivity studies as far as ispossible to include all the relev 4. MODELLING SOLUTION REQUIREMENTS and that the modelling results appear valid. What type of model output is required, givenArea the stated modelling purpose? Subject 3 in Table 3 concerns the modellingreasonable, technique able 2 audit, via the eleven subject a conducted within the eleven subject headings listed in Table 2. 5. MODELLING TECHNIQUE itself: the numerical code used, the supporting data, sensitivity Subject Area 4 in Table 3 covers model adequacy more he 11 subject headings for the semi-hard technical audit ofSOLUTION the rock mechanics The individual auditing questions and the answers for the designerrorstheir detailed answers in Table odelling supporting the cavern excavation design for the Laxiwa project onoutput the How is the required model to be obtained? analyses and presentation of the results. Becauseexplicitly our rock through the identification of actual or potential ellow River, China. 6. NUMERICAL CODE UTILIZED mechanics and rock engineering knowledge is continuously and how the answers reveal th approach usedthe in this case are listed in Table 3 with the eleven increasing and models are continuously beingand improved, possible need for corrective action. Which numerical How do wenumerical know that the code is 1. THE MODELLING OBJECTIVE The purpose of the modelling?code is to be used? in determining the cavern e subject headings listed under four subject areas. the auditing questions themselves may need to be adjusted to In the current absence of any internationally agreed check on 2. CONCEPTUALIZATION OF THE PROCESSES BEING MODELLED operating correctly? The sub-system(s) being isolated for study. The physical processes involved. capture the METHOD critical aspects in this subject area. Theaudit keythe issues are engineering support. To demonstrate this The subject areas in design Table process 2 are different from those in 7. SUPPORTING MODEL DATA & DATA INPUT rock (as occurs in other engineer3. SPECIFICATION OF THE MODELLING CONTENT that are there is a to check on the appropriateness of the model, Whatrelations, are the necessary supportingensuring data? How they be obtained? How 1 because What are the physical variables, connecting parameters, boundary Table now such we areasdealing a design ing areas aircraftwith design), theproblem, auditingrather procedurerequire illustration of the m that there have been checks on how the model is used, ensuring conditions, initial conditions, etc. are they to be input? that uncertainties are captured by sensitivity as far as is illustrated in measurements. this demonstration example shows how than studies site investigation However, the 2same logic is thecorrectness of the supporting d 4. MODELLING SOLUTION REQUIREMENTS 8. MODEL SENSITIVITY ANALYSIS reasonable, and that the modelling results appear valid.answers to the questions listed in Table 3 enable confirmation What type of model output is required, given the stated modelling purpose? Noting that, in this case, t followed in Table 3 beginning with the objective, Subject Area 1, How does the model output depend onSubject the model terms3 ofcovers whether a 5. MODELLING SOLUTION TECHNIQUE Areainput 4 ininTable model adequacy more How is the required model output to be sensitivity obtained? thaterrors an adequate is the being followed or analysis is required? explicitly through the identification of and actual following or potential the key stepsprocedure through to conclusions, in has this beenthe auditing questions are gi 6. NUMERICAL CODE UTILIZED 9. PRESENTATION OF MODELLING RESULTS and the possible need for corrective action. followed. The auditingincan be used during Which numerical code is to be used? How do we know that the code is case the modelling adequacy, Subject Area 4. or after the modellingaudit more efficient the form Is it possible to demonstrate that the In numerical code is operating correctly? the current absence of any internationally agreed check on Needless to say, contemporaneous auditing is preferred operating correctly? activity. more detail, e.g. whether a 7. SUPPORTING MODEL DATA & DATA INPUT METHOD Subject Area 2 in Table 3 concerns the conceptualization of the the rock engineering design process (as occurs in other engineer(Hudson & Feng 2010) Are the modelling results clearly presented? that any corrective action can be immediately implemented. What are the necessary supporting data? How are they to be obtained? How ing areas such as aircraft design), the auditing so 10. SOURCES OF ERRORS required to a particular quest processes procedure being modelled, the specification of the modelling are they to be input? illustrated in this demonstration example 2 shows how the What are the main sources of errors? 8. MODEL SENSITIVITY ANALYSIS content, modelling solution requirements and the modelling answers to the questions listed in Table 3 enable the confirmation How does the model output depend on11. the MODELLING model input in terms of whether a ADEQUACY 3.3. Comments Laxiwaarea casebecause, example as indicated by that an adequate procedure is being followed or has been sensitivity analysis is required? solution technique. This on is athe crucial Does the modelling seem adequate for the purpose? Are there any problem 9. PRESENTATION OF MODELLING RESULTS followed. The auditing can be used during or after the modelling areas?code Is any corrective action required? 4. Conclusions Is it possible to demonstrate that the numerical is operating correctly? the associated sub-questions in Table 3, there needs to be
Un ejemplo de auditoría: 38 preguntas garantizan la calidad del modelo activity. Needless to say, contemporaneous auditing is preferred
It was noted earlier that the technical auditing of this case so that ha any corrective action can be immediately justification of the el exact modelling del approach in terms of • ¿Se establecido conimplemented. claridad propósito modelo? example has been termed ‘semi-hard’ because it is has not been comprehensiveness of the physical variables represented and In order to ensure that the possible to include all the relevant details in the paper. However, the 3.3. Comments on the Laxiwa case example • ¿Cómo se sabrá cuando la modelación esté completa? modelling and rock engineer the specification of all the conditions. audit, via the eleven subject areas and thirty-eight questions with conducted within the eleven subject headings listed in Table 2. It was noted earlier that the technical auditing of this case their detailed answers in Table 3, illustrates the style of the hard audit The individual auditing•questions and the answers for the design example has been termed ‘semi-hard’ because it is has not been ¿Qué sistemas de macizo rocoso fueron considerados? and how approach used in this case are listed in Table with details the eleven possible to include all the3relevant in the paper. However, the the answers reveal the procedures used and their suitability audit, the eleven subject areas and thirty-eight questions with onducted within the eleven subject subject headings listedlisted in Tableunder 2. in determining the cavern excavation sequence and appropriate headings fourvia subject areas. • ¿Qué procesos físicos fueron considerados? their detailed answers in Table 3, illustrates the styleprincipales of the hard audit he individual auditing questions and the answers for the design support. To demonstrate this example as a truly hard audit would The audit subject areas in and Table different those how 2 theare answers reveal thefrom procedures usedin and their suitability pproach used in this case are listed in Table 3 with the eleven require illustration of the more penetrating investigation of the 1 because now we are dealing with athe design problem, rather and appropriate in determining excavation sequence ubject headings listed under Table four subject areas. • in ¿Cuál es lacavern variable independiente principal? support. To demonstrate a truly would The audit subject areas in than Table 2site are different from those correctness of the supporting data, numerical analyses, etc. investigation measurements. However,this theexample same as logic is hard audit require illustration of the more penetrating investigation of the able 1 because now we are dealing with a design problem, rather Noting that, in this case, the manner in which the answers to followed in Table 3 beginning with the objective, Subject Area 1,sistema correctness of the supporting data, numerical analyses, etc. han site investigation measurements. However, the same logic is ¿Cómo •steps se perturba elwhich para activar mecanismos? auditing questions are given has been left open, to make the following key through to in the ininthis Noting that, thisconclusions, case, the manner the the answers to ollowed in Table 3 beginningand with the objective, the Subject Area 1, the auditing questions are given has been left open, toaudit make more the nd following the key steps through to the conclusions, in this efficient the form of the answers should be specified in case the modelling adequacy, in Subject Area 4. • Lista de las variables físicas audit more efficient the form of the answers should be specified in ase the modelling adequacy, in Subject Area 4. more detail, e.g. whether a narrative or numerical answer is Subject Area 2 in Table 3 concerns thee.g. conceptualization more detail, whether a narrative of or the numerical answer is Subject Area 2 in Table 3 concerns the conceptualization of the required to a particular question. processes being modelled, the specification of the modelling required to a particular question. rocesses being modelled, the specification of the modelling • Lista de los acoplamientos termo-hidro-mecánicos ontent, the modelling solution requirements the modelling content, the and modelling solution requirements and the modelling olution technique. This is a crucial area because, as indicated by solution technique. This is a crucial area because, as indicated by 4. Conclusions he associated sub-questions in Table 3, there needs to• be ¿El modelo es 1D, 2D, 3D4. oConclusions alguna combinación? the associated sub-questions in Table 3, there needs to be ustification of the exact modelling approach in terms of omprehensiveness of the physical variables of represented and In order to ensure that the in procedures justification the exact modelling approach termsusedofin rock mechanics • ¿Se modela un continuo o un modelling and rock engineering design are appropriate for thediscontino? he specification of all the conditions. comprehensiveness of the physical variables represented and In order to ensure that the procedures used in rock mechanics modelling and de rock borde engineering design are appropriate for the the specification of all the conditions. • Especificación de las condiciones 6
Are the modelling results clearly presented? 10. SOURCES OF ERRORS What are the main sources of errors? 11. MODELLING ADEQUACY Does the modelling seem adequate for the purpose? Are there any problem areas? Is any corrective action required?
3
Auditoría de modelos numéricos para macizos rocosos
Un ejemplo de auditoría: 38 preguntas garantizan la calidad del modelo • • • • • • • • •
(Hudson & Feng 2010)
Especificación de las condiciones iniciales ¿Cómo se establece la condición final? ¿Cuál es la salida requerida del modelo? ¿Cumplen los resultados con el objetivo del modelo? ¿Cómo se obtiene el resultado? ¿Un código, una corrida? ¿O un conjunto de experimentos numéricos? ¿Hay controles de calidad? Control de datos de entrada, validación con soluciones conocidas, corridas duplicadas? ¿Qué código numérico se emplea? ¿Porqué se emplea ese código? ¿Dónde se originó el código?
Auditoría de modelos numéricos para macizos rocosos
7
Un ejemplo de auditoría: 38 preguntas garantizan la calidad del modelo • • • • • • • • •
(Hudson & Feng 2010)
¿Cómo se validó el código? Lista de tipo y justificación de condiciones de borde Lista de datos de entrada incluyendo fuente y justificación ¿Hubo que ajustar datos antes de su entrada al código? ¿Cómo dependen los resultados de los valores de los parámetros de entrada? ¿Se ha realizado un análisis de sensibilidad? ¿Que tipo? ¿Cómo resume resultados del análisis de sensibilidad? ¿Se puede demostrar que el programa funciona bien? ¿Se puede demostrar que los datos son hipótesis razonables para un macizo rocoso?
8
4
Auditoría de modelos numéricos para macizos rocosos
Un ejemplo de auditoría: 38 preguntas garantizan la calidad del modelo • • • • • • • • •
(Hudson & Feng 2010)
¿Cómo se presentan los resultados del modelo? ¿Se ajusta la presentación de resultados con el objetivo? ¿Ha corregido algún error hasta ahora? Lista de los errores potencialmente significativos ¿Puede alguno de esos errores invalidar el resultado, concepto y conclusiones? Las preguntas anteriores ¿indican que en principio el modelo es adecuado para su propósito? Si no, liste las áreas problemáticas ¿Qué acción correctiva se requiere? ¿Debe repetirse esta auditoría luego de correcciones?
9
5
