370 likes | 662 Views
A Systems Scientist’s Thoughts on Model-Based Systems Engineering. Wayne Wakeland , PhD Systems Science Program Portland State University. Dynamic Connected/coupled Governed by feedback Boundaries are artificial (often permeable) Complex & non-linear History dependent Edge of chaos.
E N D
A Systems Scientist’s Thoughts on Model-Based Systems Engineering Wayne Wakeland, PhD Systems Science Program Portland State University
Dynamic Connected/coupled Governed by feedback Boundaries are artificial (often permeable) Complex & non-linear History dependent Edge of chaos Self-organizing Self-replicating (living systems) Adaptive/evolutionary Characterized by trade-offs Counterintuitive Policy resistant Emergent The Nature of Systems
In complex systems, cause and effect are often distant in time and space We may act to produce short-term benefits and long-term costs; we forget about delay
The solution to one problem may cause another problem (unintended results) Example: The ”Green Revolution” agricultural technologies were introduced into Asia in the late 1960s as a solution for food insecurity. Decades later, they have proved detrimental in terms of biodiversity loss, increased use of agro-chemical based pest and weed control, water logging, salinization and land degradation. Artist Gary Larson Slide adapted from LEAD International and Sustainability Institute
The Iceberg: EVENTS – PATTERNS – STRUCTURE What happened? Events REACT Patterns of Behavior Increased leverage and opportunities for learning What has been happening? ANTICIPATE Systemic Structure Why has this been happening? How can I improve the performance of the system? DESIGN Mental Models
Systems Science Methods • Systems science methods focus on understanding the general properties and behavior of complex systems by creating models and finding patterns in data • We strive to interact effectively with collaborators from various disciplines by using clear and well-annotated graphics and diagrams to present models and data analysis results
Key Concepts and Principles • A system consists of elements & relationships, with specific purpose/goal/function • Whole > sum of the parts • Structure causes behavior • Circular causality • Outputs influence inputs; cannot separate cause and effect • Mental models (often hidden) shape our thinking • Systems Archetypes (common structures & behaviors) • E.g., Fixes that fail, Shifting the Burden, Success to the Successful, Limits to Growth, Tragedy of the Commons • And much more (far too much to cover this morning) • Complex adaptive systems, living systems, open systems, structural coupling, autopoiesis, adaptation, resilience, evolution, …
Systems Thinking • Seeing the forest and the trees • Interconnectedness • Thinking dynamically • Behavior over time • Delayed impacts/consequences • Thinking closed loop (vs. linear causality) • Endogenous thinking (system as cause) • Thinking operationally • How things actually actually work
Example: Fixes that Fail Archetype • The story: due to budget problems, spending on maintenance decreases, which balances the budget…BUT, over time, breakdowns increase, forcing more spending, which stresses the budget even worse than before!
Specific Modeling Methods • System dynamics • Focuses on modeling the underlying feedback structures with differential equations • Equation are solved to simulate behavior over time • Discrete system simulation • Uses a Monte Carlo approach to analyze how the variety/randomness impacts system performance • Often emphasizing business operations and processes, especially in manufacturing and supply chain logistics • Agent based simulation • Used to study how low-level interactions between individual agents influences overall system behavior/performance
System Dynamics Example • Project Management • Brooks’ Law: “Adding manpower to a late software project makes it later” • SD has been used to simulate complex projects and evaluate potential decisions, actions, policies
Typical Project “Disasters” • The Channel Tunnel -- original estimate, • $3 billion; final cost, $10 billion • Boston’s “Big Dig” -- original mid- • 1980’s estimate, $2.5 billion; latest • estimate, $14.5 billion (9/2001) • Aircraft development -- nearly double • initial estimate • New Car Development -- original plan, • 400 person-years of effort; final cost, 800 • person-years
Discrete System Simulation • Detailed, step-by-step emulation of the flow of entities through the system • With uncertain arrivals, processing times, and/or "routing" (branching) • The computer monitors each simulated entity as simulated time proceeds • Enter system, move thru, according to the various probability fns. governing timing and sequence of events • The computer also records pertinent data regarding the simulated entities and servers • wait times, throughput, queue lengths, process times, utilization… • Creates a synthetic "sample" of system performance data • Sample data is then analyzed statistically
Types of Problems DSS can Address • Performance issues in existing systems • Long waits, high inventory, poor utilization of resources, low throughput • Need to estimate performance of a system under design • What Might One Learn? • Where the bottlenecks are and how they might be alleviated • How to improve flow, reduce queues and wait times, and increase utilization & throughput • The optimal number of servers, queues, buffers, etc. • Effective operating rules or policies
Examples of Discrete System Simulation • Computer network • Freeway system • Business process (e.g., insurance office) • Criminal justice system • Chemical plant • Fast-food restaurant • Supermarket • Theme park • Emergency Response system • Manufacturing facilities • Bank operations • Airport operations (passengers, security, planes, crews, baggage) • Transportation/logistics/ distribution operations • Hospital facilities (emergency room, operating room, admissions)
DSS model animation (closely mimics the actual system) The model contains complex logic regarding: A) Different fault occurrences, B) Part filling requirements, and C) Realistic variations seen in complex assembly processes
Model Results • The simulation showed the behavior of the proposed new automated system • It suggested that given expected faults, operator utilization will be 45-55% • Thus, if the operator must load parts, do audits, and perform fault correction, they could not handle two machines, as would be needed to achieve the cost targets
Agent Based Simulation • More of a stretch for systems engineering than the others… • Key Features • Agents • Environment • Rules • Spatial aspects • Can reflect heterogeneity of individuals
Key ABS Concepts • Decentralized control • Bottom up as opposed to top down • Emergence • Self-organization • Evolutionary considerations • Examples • Spread of Forest Fires • Flocking • Crowd behavior • Ants (and how ants can find optima) • Network effects
Crowd Crush Model • The problem: crowd panics • Sheffield, England 1986, 96 dead • Phnom Phen, Cambodia, November 23, 2010: 347 dead • Duisburg, Germany, July 25, 2010,19 dead • This model was developed as class project • Alexandra Nielsen, Systems Science 525, Fall 2010
Crowd Crush: defined • Die of asphyxiation, not blunt force trauma • Can die standing • Warning of a crush • Surrounded on all sides • More than 4 people per square meter • Force to kill • The force of 5 people pushing on one person can break a rib, collapse a lung, smash a child's head
Simulation purpose • Discover why some crowds are lethal and others not • Can crowd deaths occur in non-aggressive crowds? • Does aggression or reactivity (jostling) have a greater impact on crowd deaths? • Is there some combination of factors that is reliably lethal? (So we can avoid it) • What interventions can prevent deaths? • Is there a critical density after which nothing will work? Netlogo model…
Model Testing • Interventions • Opening closing entrance exit • Shortening corridor (simulate smaller crowd) • Panic on seeing another dead • Validate vs. anecdotal evidence • Wal-Mart door rush • Cambodia see a “body” → panic • Opening door in a crowd → death (Barnsley Public Hall disaster)
Applicability and Limitations • Large crowds, single doorway • Love Fest • One entrance of a soccer stadium (Liverpool) • Good for understanding simple crowd dynamics • Limited by simplifying assumptions (extremely simple) • No falling • Only forward motion • No groups, altruism, variability in agents • Forces not vectors, not true physical force
Findings • Don't allow a huge build up, then open a door • Closing the gates before clearing the corridor helps, but not much • Do anything you can to prevent panic