Optimization
CM_Opti MSE Course (lecture 1)
CM_Opti MSE Course (lecture 1)
9
0.0 (0)
Fichier Détails
Cartes-fiches | 9 |
---|---|
Langue | Deutsch |
Catégorie | Mathématiques |
Niveau | Université |
Crée / Actualisé | 06.06.2021 / 06.06.2021 |
Attribution de licence | Non précisé |
Lien de web |
https://card2brain.ch/box/20210606_optimization
|
Intégrer |
<iframe src="https://card2brain.ch/box/20210606_optimization/embed" width="780" height="150" scrolling="no" frameborder="0"></iframe>
|
Which two main areas of application exist in optimization?
- Optimization of business processes (production, logistics, services, operations management, ...) -> main focus in course:
- Optimize aircraft assignment in flight operations
- Optimize production planning in complex manufacturing plants
- Optimize machine utilizatino in shop floor scheduling
- Optimization of technical processes (engineering):
- Optimize machining conditions in metal-cutting (speed, pressure, angle, ...)
- Optimize parameters of chemical or physical experiments
What is the difference betwen qualitative vs. quntitative optimization?
- Quantitative analysis and optimization:
- Bases of quantifiable information and knowledge:
- Numerical, measurable data, mathematical models and algorithms
- Bases of quantifiable information and knowledge:
- Qualitative analysis and optimization
- Bases on non-quantifiable information and knowledge:
- "informal" facts, verbal descriptions of processes and procedures, unstructered information, experience, implicit know-how
- Bases on non-quantifiable information and knowledge:
- Example of typically quantitative optimization problems:
- Finding "the best" equipment (machines, tools, etc.) for a cerain task
- Finding "optimal" locations for facilities (plants, warehouses, ...) in a supply network
- Improving business processes
What do we typically have in Business Consulting projects?
- Phase 1: Qualitative analysis (very important, often up to 80%)
- Often unclear problem descriptions, mess of information, contradictory opinions
- Phse 2:
- Either qualitative improvements: e.g. definition and implementation of new processes
- Or quantitative, e.g.:
- Identification of certain subsystems crucial for performance
- Having quantitative characteristics, not solvable by "common sense" (intuition, ...)
- Need for "decision support" fromo methematical models
What should you alway remember when working on optimizatin problems?
- In general, before applying quantitative methos:
- Huge amount of challenging and crucial prelininary qualitative work is necessary:
- Finding out, what the real problem is, defining scope and project boundaries, ...
- Understanding all necessary details of the business areas involved, ...
- Often this is the most difficult part of the project
- Solving the wrong problem or solving the right problem inadequately
- Qualitative analysis and initial, conceptual design crucial for project success
- Often more challenging than development of quantitative methods
- Matter of experience, intuition, dialogue with business partners
- Huge amount of challenging and crucial prelininary qualitative work is necessary:
What are the two main types of optimization?
- Continuous optimization:
- Infinite number of solutions represented by continuous variables
- Graph of the objective function is arbitrary "landscape"
- Main part of the theory: local optimization
- Methods mainly bases on differentiability information (1st and 2nd derivatives)
- Reminder: gradient (vector of 1st partial derivatieves) pointing the direction of the steepest increase of the objective function
- Very diffucult in case of non-differentiable of even non-continuous functions
- Constrained optimization (with constraints) are more difficult than optimization without constraints
- Global optimization: mostly "stochastic search"
- Discrete optimization
- Number of solutions is finite (or countable): represented by integer variables
- Not interesting for mathematics theory until ca. 1930
- There exists a trivial and finite algorithm: Enumeration
- Invention of computers -> possibility to solve real life problems
- Major part can be fomulated as discrete problems
- Goal: to solve (exactly or approximately) a real life problem efficiently
- Question: What is a good (efficient) algorithm for a given problem?
What means linearity in discrete optimization?
- Discrete Optimization: finite number of solutions
- Every solution represented by a set of variables
- Solution set is a finite set of point in n-dimensional space
- "Contour" of these points has a "linear" shape (with corners): caused by the finiteness
- Mathematically: "Convex hull (Hülle) of the points is a cnves polyhedron"
Where is discrete optimization used mostly in and what are its main topics?
- Central methodology of Operations Research (OR)
- Great importance for Operations Management:
- Applications: Production, Logistics, Services, ...
- Many large and complex problems from industry solved successfully
- Main Topics
- Linear Programming (LP)
- Integer Linear Programming (ILP), Mixed Integer Programming (MIP)
When was the Simplex algorithm for Linear Programming (LP) invented and what was the biggest OR success of it?
- 1947 by Dantzig
- 1992: American Airlines saved 1'400 Mio. $ over 3 years by yield management