l17sys_dynamics2 - System and Change Dynamics Nathaniel...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: System and Change Dynamics Nathaniel Osgood 4/21/2004 Systems Thinking and Project Management Primary critique: Traditional methods too Fragmented Restrictive in assumptions Local in attention to implications of changes Hesitant regarding representation of "soft" factors Too dependent on people link components Too willing to ignore important "side effects" Seen as potentially major contributor in project Learning (model captures institutional knowledge) Planning (identify robust decision rules, leverage pts) Control (how to best handle deviations) Evolving More Complex Diagrams Relationship with Owner Delay + Disputes with Owner - Labor Productivity + Morale + Delay - Relationship with Owner - Disputes with Owner Schedule Pressure + Labor Productivity + Morale + Disputes with Owner Delay Relationship with Owner + Labor Productivity + Morale - Concern About Cost of Delay Relationship with Owner Delay + Disputes with Owner - System Dynamics Basics Represents system as coupled series of ordinary differential equations (ODEs) Standard state-equation formulation Continuous time formulation Stochastic components permissible (special handling) Analytic solutions not possible: Numerically integrate Graphical representation for problem focus State equations as stocks Components of differentials as follows Intermediate computations as auxiliaries, table functions, etc. How a SD Model is Created Conceptualize system using causal loop diagram Convert CLD to "stock & flow" structure State variables (accumulations) as stocks Changes to state variables as flows All change in system state occurs through flows All loops include at least one stock Intermediate calculations, outputs as auxiliaries Add to equations to capture relations among vars Calibrate to historic data Run scenarios to identify effect, robust policies Example Creation of a System Dynamics Model Step 1: Map out Causal Loops Potential - Customers Word of + Mouth Sales + + + Customers Step 2: Identify state variables of interest Potential Customers Customers Step 3: Identify flows of interest Potential Customers Customers New Customers Example Creation of a System Dynamics Model Step 4: Define Supporting Variables Fraction of Customers in Population Total Population Likelihood of Potential Customer Joining Function Giving Likelihood of Potential Customer Becoming New Customer based on Fraction of Customers Potential Customers Customers New Customers Insert equations to describe linkages E.g. Total Population = Customers+Potential Customers Fraction of Customers in Population= Customers/Total Population Example Creation of a System Dynamics Model II Step 1: Map out Causal Loops Work Pressure + + Work Remaining Productivity + Fatigue Step 2: Identify state variables of interest Work Remaining Fatigue Step 3: Identify flows of interest Work Remaining Work Being Completed Fatigue Growth in Fatigue due to Schedule Pressure Recovery from Fatigue Example Creation of a System Dynamics Model Initial Work Size Normal Productivity Work Hours as a Function of Fraction of Work Fraction of Work Completed Completed Work Hours Normal Work Hours Growth in Fatigue due to Schedule Pressure Work Remaining Work Being Completed Productivity Coefficient as a Function of Direct Work Pressure Productivity Coefficient from Fatigue Productivity coeefficient from Fatigue as a Function of Fatigue Fatigue Recovery from Fatigue Mean time to Recover From Fatigue Statistics Statistics can assist in calibrating relationships in a model Remember that there are typically many indirect effects not shown! Examples of How We Use a System Dynamics Model Typical first step: Assume some "baseline" scenario Just point of reference not particularly privileged "Policy" scenario analysis Change "policy parameters" (e.g. hiring policy, fraction change vs. rework, etc.) and look at results Policy robustness to uncertainty For different policy parameters, examine implications of major points of uncertainty Sensitivity analyses (both of above, and to focus further data collection) Examine impact of different external conditions Uncertainty in System Dynamics Often address uncertainty using sensitivity analysis Goal is to see how much our choices depend on uncertainty Thorough analysis requires monte-carlo trials Two types of uncertainty Static uncertainty (e.g. uncertainty about the value of a model parameter) Specify distribution for model variable at start of run Dynamic uncertainty (stochastic processes): Sample throughout simulation Stochastic Process Monte Carlo Essentially numerically solving stochastic differential equation TinySensitivityForIndividualTrajectories Oil Price 40 32.5 25 17.5 10 0 1457 2914 Time (Day) 4371 5828 Monte Carlo Output Empirical fractiles shown with color coding Baseline 50% 60% 70% 80% Average Variable Cost per Cubic Meter 0.6 90% 95% 98% 100% 0.45 0.3 0.15 0 0 1457 2914 Time (Day) 4371 5828 Coexistence...or Integration? Road to (awkward) coexistence System dynamics: "Dynamic" complexity Interactions, time delays, feedbacks Traditional tools: "detail" complexity Problems: How to synthesize results? Which to trust? How to link models? Most frequent: CLD for qualitative insight, traditional tools for quantitative tradeoffs Use of single model Integration: e.g. DPM DPM: Construction Oriented Model Created by Park, Pea-Mora Includes components of network-based models Used to analyze real-life projects Notable components Richer set of dependencies Refinement of internal/external linkages Dependencies between design acts. and construction acts. Specification change/rework distinction, cycles Characterization of "knock-on" upstream/downstream effects Inter-Activity Dependencies Example: Plumbing & Electrical Roughing and Drywall Example: Pouring of concrete and erection of shoring for next floor Internal Dependencies Example: Subsequent Floors Example: Site-Wide Drywall Role of Quality DPM operationalizes quality as fraction of work that is acceptable according to specifications Quality problems often not discovered until later! Decreasing time to discovery is key! Statically, we can expect higher quality to lead to Higher costs Higher time to completion Dynamically, quality serves as a driver for Rework Specification changes Specification Change vs. Rework: Big Picture Rework Common in other field (e.g. software) E.g. tendency to periodically rewrite software Limited in construction Rework perceived as exorbitantly expensive Delays in discovering quality problems may rule out Specification Changes When divergence between specifications and asbuilt work, changes specifications Superficially attractive: Seems to limit in field changes Multiple types Managerial changes (including owner requested) Unintended changes Unintended Changes Upstream Hidden Change Unintended Change Low Work Quality Poor Work Conditions External Scope Changes Specification Changes vs. Rework in Construction: Tradeoffs Rework More short-term work Especially if delayed in discovering original divergence! Limits extent of impacts Specification Changes Less short-term work Can lead to lots of "side effects" as specification changes propagate to other components of specification Example of Change Propagation Conflicting drawings about duct routing leads to changes to one set of drawings to accommodate re-routing of ductwork from another fan room Plumbing, electrical work reexamine routing Need for larger HVAC unit requires Re-working shop-drawings Reexamination of electrical system for higher load Reconsideration of piping supporting HVAC Examination of structural loading o f building Delays can affect customer relations, idle resources, interfere with other A/E activities, increase pressure... Big Picture: Role of Flexibility Flexibility forms a critical ally against risk of change One pays for flexibility but often pays more for lacking flexibility Design Overlap Feedbacks Uncertainties Assumptions in Design Owner's Requests on Changes Overlapping between Design and Construction Oversizing Potential Design Change Impact on Construction Project Costs Project Duration Construction Processes Design Changes Overlapping Construction Work R2 Done before Upstream Completed Increase in Time Pressure Workforce R1 Estimated Project Duration Productivity R3 Construction Changes Delay Construction Overlap Feedbacks Uncertainties Assumptions in Design Owner's Requests on Changes Overlapping between Design and Construction Oversizing Construction Processes Overlapping R2 Time Pressure R1 Productivity Delay R3 Construction Changes Increase in Workforce Potential Design Change Impact on Construction Design Changes Construction Work Done before Upstream Completed Project Costs Project Duration Estimated Project Duration Big Picture: Fast-Tracking and Overlapping of Activities Has Multiple Risks Risk 1: Greater risk of rework/changes Starting out with less certainty as to how preceding activity (design or construction) will proceed Risk 2: Greater vulnerability to rework/change Already in a hurry; schedule affords less slippage More near-critical activities Additional Feedbacks 1 Schedule Pressure + Delays + Change Ripple-Through Effects Preference for Managerial Changes + Focus on Short-Term Productivity + + Additional Feedbacks II Delay + Impact of Changes on Schedule + Delay until Discovery of Discrepency + + + Supervisor Work Level + Supervisor Fatigue + Changes + Likelihood of Overlooking a Discrepency between Specifications and As-Built Work Change Process Loop 1: DS Pre-check Identifies Divergence Time to Request Reprocess from Downstream Reprocess + + Time to Reprocess + Reprocess Requests From Downstream + R1 Upstream Changes Discovered in Downstream Downstream Delay + Delay Schedule Pressure + Hidden Changes Released to Downstream Quality Management Thoroughness Willingness to Adopt Managerial Change Downstream Quality + + + + Downstream Reprocess Change Process Loop 2: Divergence Discovered During DS Work Time to Request Time to Request Reprocess from Reprocess from Downstream Reprocess Downstream Reprocess + + + + + Reprocess Requests From Downstream Reprocess Requests From Downstream Time to to Time Reprocess Reprocess + + Delay Delay Schedule Schedule Pressure Pressure + + R2 R1 Upstream Changes Upstream Changes Discovered Discovered in Downstream in Downstream + + + Hidden Changes Hidden Changes Released to Released to Downstream Downstream Downstream Downstream Delay Delay + + ++ Quality Management Quality Management Thoroughness Downstream Downstream Quality Willingness toChange Managerial Adopt Managerial Change + Willingness to Adopt + + Downstream Reprocess Downstream Reprocess + Thoroughness Quality Change Process Loop 3: Specification Change Adopted Time to Request Reprocess from Downstream Reprocess + + + Time to Reprocess + Reprocess Requests From Downstream + R1 Upstream Changes Discovered in Downstream Downstream Delay Delay Schedule Pressure + + Hidden Changes Released to Downstream Quality Management Thoroughness Willingness to Adopt Managerial Change Downstream Quality + + + + Downstream Reprocess Basic DPM Work Package Stock-and-Flow Structure RP Request from Downstream RP Address After Release Rate RP External Managerial Change RP Internal Managerial Change Managerial Change Ratio Work Availability From Internal Concurrency Total External Concurrence Work In Progress Work Available Actual Work Quality Activity Scope RP Address Rate Quality Management Avg Quality Thoroughness Management Time Initial Work Introduce Rate Pending Work Release Rate Avg Hidden Change RP UP Change Accommodate Rate Applied RFI Reply Time Work to DO Work Rate Work Awaiting Quality Management Min Work Time Potential Work Rate from Resource Request for Information Rate Fraction of RFI Fraction of Total RFI Work Release Rate Work Released <Managarial <Managarial Change Ratio> Change Ratio> Work Pending UP Change RP RP Request to UP Rate Work Awaiting RFI Reply DPM Study: Take-Home Messages Quality has a pervasive impact on time, cost Indirect savings from high-quality work may well outweigh extra direct costs of this quality Changes have cascading effects in a project Despite up-front costs of rework, many times preferaable Fast-tracking experiences "double jeopardy" Greater likelihood of changes Greater schedule sensitivity to changes System Dynamics Simulation vs Process Simulation Process simulations generally Proceed "bottom up" Detailed simulation of activity processes System dynamics simulations Proceed "top down" Try to capture entire system of major factors Typically do not model particular activities and resources For emergent statistics, time behavior Investigate particular component of interaction on project (but can model entire project) Based around discrete event simulation do not include "soft factors" (Morale,fatigue...) Generally do not model decision rules For emergent time behavior Include soft factors Rough estimates if necessary Based around ODE/SDE Try to capture decision rules Strive for transparency Seek to use model to capture institutional knowledge for project learning Important Distinctions Distinction 1: Precision vs. Accuracy Distinction 2: Power of modeling. vs. ease of creating useful model Distinction 3: General modeling vs for a specific purpose Distinction 4: Modeling for prediction vs. for understanding TP4 Presentations TP4 involves both a Written report Final Presentations (20-25 minutes) The written report will be due May 10 for everybody The presentations will be held on either May 10 May 12 Grading Rework for those who wish to do extra work on PSet 3 Extra credit Assignment based on "Skycraper" Watch 5 hour-long videos Proposed time: Tuesday, April 27 5-8 Friday, April 30 5-7 Write a 5-page analysis ...
View Full Document

{[ snackBarMessage ]}

Ask a homework question - tutors are online