2. Disclaimer The concepts and methods presented in this document are for illustrative purposes only, and are not intended to be exhaustive. Ontonix assumes no liability or responsibility to any person or company for direct or indirect damages resulting from the use of any information contained herein. Any reproduction or distribution of this document, in whole or in part, without the prior written consent of Ontonix is prohibited. Reverse-engineering of the concepts, methods or ideas contained in this document is strictly forbidden. The methods described in the present document are protected by US patents. OntoSpace is a trademark of Ontonix All other trademarks are the property of their respective owners. Copyright 2010, Ontonix S.r.l. All Rights Reserved.
3. CONTENTS Why Manage Complexity What is Complexity? Applications of Complexity in Engineering and Manufacturing Back-up Information
4. Why Complexity Management Complexity is rapidly increasing in all spheres of social life. This leads to multiple converging stresses and increases the levels of turbulence of the global economy as well as of the society. Doing business is increasingly difficult. Complexity must be managed before it reaches dangerous levels and threatens sustainability.
5. What is Complexity? Complexity is a function of structureanduncertainty(entropy). It quantifies the degree of sophistication and the “amount of chaos” within a system. It is a fundamental property of dynamical systems, just like energy. Structure (Topology of information flow) Uncertainty (Noise content in information) COMPLEXITY
6. Complex or Complicated? A system may be complicated, but have very low complexity. A large number of parts doesn’t generally imply high complexity. It does, in general, imply a complicated system. Complexity implies capacity to surprise, to deliver unexpected behaviour. In order to measure the amount of complexity it is necessary to take uncertainty into account, not just the number of parts.
19. Representing Structure Conventional Map: Difficult to analyze when number of nodes and links is large (“spaghetti effect”). Hubs are difficult to identify. Ontonix Complexity Map: Easy to analyze even when the number of nodes and links is very large. Hubs are easy to identify. Link Node Node Link
24. From Data to Images and Structure y Image with little or no structure: no information is exchanged between x and y. Image with evident structure: much information is exchanged between x and y. Image with evident structure: much information is exchanged between x and y. x y x y x
25. Critical Complexity Complexity cannot grow indefinitely and has a maximum. Close to this maximum, called critical complexity, a given system becomes fragile and vulnerable. In the proximity of critical complexity systems possess numerous modes of behaviour and can switch from one mode to another sponaneously. Modes represent (behavioural) attractors. Critically complex systems are very difficult to manage and can easily run out of hand. After critical complexity decay begins unless structural changes are made to the system. Mode 2 Mode 1 Mode 3 Mode 4
26. Complexity and Robustness It is not convenient to function in the proximity of critical complexity – behaviour becomes unpredictable. Thevulnerabilityof a system is proportional to its distance from critical complexity. A well-managed system is kept at a safe distance from critical complexity. Sudden changes in complexity point to traumas(endogenous or exogenous) and may be seen asearly warnings and precursors of crises. Upper complexity bound (critical complexity) Ccritical – Ccurrent Current system complexity Robustness is function of Ccritical – Ccurrent. This measure is known as topological robustness and quantifies the system’s ability to preserve its functionality. This is the key concept behind our innovative complexity-based rating scheme. Lower complexity bound
27. What Happens at Critical Complexity Critical complexity (upper complexity bound) y Entropydominates – systemis “chaotic” y y x x y x Current complexity Structure dominates – system is deterministic 0 The x-y scatter plots shown in this slide represent typical relationships between pairs of variables in a system which functions in the proximity of the lower and upper bounds of complexity. x Lower complexity bound
28. Complexity vs Time Step i Step i+1 Step i+2 In systems which evolve in time, complexity changes, as well as the corresponding lower and upper bounds. The same may be said of the Complexity Map, its structure, density, hubs, etc. In such cases data is analyzed using a moving-window approach. t1 t2 Time Time t1 Complexity = 28.5 Time t2 Complexity = 44.8
29. Characteristics of a Highly Complex Business or System Difficult to manage Highly exposed Vulnerable Unhealthy Fragile Unsustainable
30. 18 Complexity x Uncertainty = Fragility When uncertainty meets high complexity, the result is fragility. Simple systems can cope better with uncertainty than highly complex systems. Highly complex systems are more exposed to the effects of uncertainty because of the countless ways in which they process information. They can fail in many ways, often due to apparently innocent causes. Uncertainty in the environment, cannot be avoided. We must learn to live with it. Hence the need to manage complexity. Since fragility is the prelude to risk, risk management can be accomplished via complexity management.
31. 19 Nature Increases Complexity (Functionality): There is a Price to Pay! Fragility Functionality Time
40. 23 Complexity-Based CAD: Pedestrian Bridge Geometric parameters Quarter model view: Rib Spacing is the amount of holes between ribs T The dimension fraction is D/T D x The spacing factor is S/T S Height t is the flange distance t Thickness factor = x/Height If the thickness factor is increased Cut depth, width and radius determine the shape of the ribs
41. 24 Complexity-Based CAD – The Concept Starting from the initial nominal model, a sequence of randomly generated solutions is created. This is done using Monte Carlo techniques and a multi-run environment. For every solution, a CAD system is used to automatically generate an FE mesh. For every mesh a static and an eignevalue analysis is run in order to determine stresses, deflections and natural frequencies. The process is repeated a few hundred times and is fully automatic (one loop). The results are processed and feasible solutions are determined by specifying desired levels of: Stresses Deflections Natural frequencies Various solutions are found to satisfy constraints and performance objectives.
44. 27 Example of Complexity-Based Design: Turbine Disk Design Solution 1 Solution 2 Solution 2 has much lower complexity (15.8 vs. 22.6) and slightly higher robustness than Solution 1.
45. 28 Example of Complexity-Based Design: The James Webb Space Telescope Option 1 Option 2 Option 3 Option 4 Best option: lowest complexity with same performance James Webb Space Telescope payload adapter. Courtesy EADS CASA Espacio.
46. 29 Crash Test Data Processing Analysis of crash-test data shows that over the past decade, complexity has been increasing. 1980 2005 1980 2005
47. 30 Measuring Robustness in Mechanical Systems Robust design and related techniques have been object of discussion for over a decade. However, the robustness of designs conceived using such methods has never actually been measured and no global measure of robustness has ever been proposed. Recently developed complexity-based robustness measures allow engineers to quantify the global robustness of any dynamical system.
48. 31 More on Robustness: The Connectivity Histogram Loss of this hub damages greatly the Process Map Robust Fragile Additional information on robustness may be obtained examining the shape of the Connectivity Histogram. Spiky histograms (known as Zipfian) denote fragile topologies, while flatter ones point to more robust systems.
50. 33 Power Turbine Monitoring Critical complexity Alert complexity Complexity Minimum complexity Time
51. Process Plant: Vulnerability Analysis To evaluate the complexity of a process or a system it is first necessary to obtain the equivalent process map(s). These are computed automatically by OntoSpace™. In order to extract the maps, OntoSpace™ requires data from the process sampled with a certain frequency at a set of significant locations or sensor points.
55. 38 In-Flight Structural Health Monitoring Engine 1 Complexity Map Airframe Complexity Map Engine 3 has a vibration problem. This manifests itself in a complexity value which is 30% greater than that of the other 3 engines. Engine 2 Complexity Map Engine 4 Complexity Map Engine 3 Complexity Map
56. Desalinization Plant Monitoring All relationships between Prod (FL) and other parameters Process Entropy Process Complexity STAGE 1 (red nodes) HUB: key parameter STAGE 1 (blue nodes) Process Robustness Week 19 Week 20 The abovemaps, whichcorrespondtotwodifferentweeks, illustrate some basicfeatures and information which can be obtainedbyanalyzingProcessMaps. 39
57. Desalinization Plant Monitoring The evolutionsofentropy and complexity show howevenduring a single month the processishighlynon-stationary. Onenoticeshow on givenday the very low levelofCLcoincideswith a minimum valueofbothcomplexity and entropy The time-historyshowshow at a certainpoint Brine (CN) increases. The correspondingportionof the time-historyishighlightedby the redcircle (seealsoentropyevolution). The reddottedarrowshowshowentropyprogressivelygrows. Thiskindofbehaviourpointsto some kindof “accumulationofenergy” afterwhich the system maysuddenlyswitchto a differentmode ofbehavior – seebluedottedarrowpointing down.
60. Monitoring Patients During Thoracic Surgery Operation Intensive Care Large oscillations of complexity point to globally unstable patient. Patient becomes critically unstable. Courtesy Erasmus Medical Center
61. 44 Measuring The Credibility of a Computer Model Test Simulation How well does the numerical model emulate the real thing?
62.
63. Ctest < Cmodel - Model (generally) generates noiseComplexity measures the amount of structured information
72. 49 Is Optimality Convenient? In highly turbulent environments, seeking optimality is unjustified. In fact, optimal designs are inherently fragile. Robust solutions should be sought instead. This can be accomplished not by maximising (arbitrary) objective functions but by accepting compromises in terms of performance and seeking simpler solutions to problems.
73. Pre-Alarms and Crisis-Anticipation SERVICES: CRISIS ANTICIPATION Complexity may be measured and monitored versus time. Sudden fluctuations of complexity point to instabilities, traumas or imminent crises. The magnitude of the trauma may be quantified by the difference in complexity before and after the event. Based on similar information it is possible to identify thresholds of complexity beyond which one may expect a crisis and therefore take measures in order to mitigate its effects. Collapsing system. Loss of complexity is equivalent to loss of functionality. When complexity reaches a zero value, the system has no longer any structure and ceases all activity. System with step-wise increases complexity. This case corresponds to the US housing market. The time span is 5 years. The 2007 sub-prime bubble is indicated by red arrow. System with mildly increasing complexity (middle orange curve is complexity, the other curves correspond to lower and upper complexity bounds. Highly unstable system – the case is relative to a hospital patient in an ICU. Each complexity fluctuation corresponds to crisis. Time span is 8 hours.
74. Crisis Anticipation in IT Systems SERVICES: CRISIS ANTICIPATION Courtesy, Banca Popolare di Sondrio IT systems in banks are dynamical systems composed of HW, SW, human interaction (client access via web). The dynamics of large IT systems brings them on occasions to states of high vulnerability which may be anticipated via real-time complexity monitoring.
75. Evolution of Design Paradigms Analysis-based Simulation-based 20-th Century 21-st Century System Fragility Complexity Management Complexity-based Design (MCS) Uncertainty Management (MCS) Robust Design (MCS) Stochastics (MCS) MDO Optimisation Sensitivity analyses System Complexity Parametric studies Trial and error MCS = Monte Carlo Simulation
77. About Ontonix Established in 2005 byexperts in the aerospace, nuclear and civilengineeringindustries. Over 60 man-yearsexperience in advancedrisk management and Monte Carlo Simulation. Ontonix is the first company tohavedevelopedrationalmeansofmeasuringcomplexityand relatingittobusiness performance and sustainability. Initiallyestablished in the USA, Ontonix isnowheadquartered in Como, Italy and hasoffices in the USA, Poland, Braziland South Africa. In 2005 Ontonix launched OntoSpace™, the first SW system forpracticalcomplexity management. In 2007 the first on-line rating service hasbeeninaugurated. Ontonix currentlyservescustomers in suchfieldsas banking, air-traffic management, medicine, defense and engineering.
78. Publications Computational Stochastic Mechanics in a Meta-Computing Perspective. 1997 Theory of Eigenvalue Orbits.1998 Principles of Simulation-Based Computer-Aided Engineering. 1999 Application Strategies of Robust Design and Complexity Management in Engineering. 2009. Co-author Dr. H. Sippel Beyond Optimization in Computer-Aided Engineering. 2002 Practical Complexity Management. 2009 A New Theory of Risk and Rating. 2010
79. Saconsulting Madrid, Spain Our Global Presence Ontonix UK Glasgow, UK CPS Frankfurt, D Ontonix Sp z o.o. Warsaw, Poland Business Dimensions Geneva, CH Ontonix LLC Novi, USA BLUE Eng. Bursa, Turkey Soyotec Beijing, PRC VAS Hinteregg, CH OntoMed LLC Ann Arbor, USA Ontonix S.r.l. Como, Italy FlexSci Beijing, PRC Ontonix RSA Pretoria, RSA Ontonix Brasil Sao Paulo, Brazil