Optimization Techniques 1. Optimization concepts Abe Alkaff Ab. Alkaff Course objectives Create and solve Mathematical Models for understanding and solving the optimization problems Learning Outcomes: Student can analyze the optimization problem, building mathematical model, and evaluating the proper optimization methods, Requirements: Engineering Mathematics II Main References: 1. Alkaff, A. Operation Research Course Handbook, ITS 2. Hillier and Lieberman., “Introduction to Operation Research”, 10th Ed, Mc Graw Hill International Edition, 2015 3. Taha, H.A., “Operation Research: an Introduction”, 8th Ed, Prentice Hall, 2006 Abe Alkaff Work Methods Real Problems abstracting Model implementing Problem Solving Mathematical interpreting Sensitivity Analysis solving Model Solving - variable - parameter Variable - objective and parameter - constraint function Value from Variable Ab. Alkaff Optimization problems illustration Electric power systems generator: • Three generator with supply capacity of 100, 80, 150 • Three region that require electric power as follows 140, 90, 70 • Cost distribution per unit from each generator to every region is known: D P 1 2 3 1 4 6 2 2 7 3 5 3 1 4 6 Find the distribution pattern that minimizes the distribution cost! Electric Power Distribution Scheme Region Generator 1 100 80 ? 1 4 6 ? 2 ? ? ? 2 140 ? 3 3 ? Abe Alkaff Ab. Alkaff 1. Mathematical Models Building a. Variable : - Distributed power to each generator for every region (πππ ) : distributed power from generator i to region j - Total distribution cost (π) b. Parameter : - Generator Load : πΎ1 = 100, πΎ2 = 80, πΎ3 = 150 - Power Demand : π1 = 140, π2 = 90, π3 = 70 4 6 2 Where πΆππ is the cost per unit - Distribution Cost : C = 7 3 5 from generator π to region π 1 4 6 Ab. Alkaff c. Objective (variable and parameter function): Minimize the distribution cost Min π = 4π11 + 6π12 + ..... +6π33 d. Constraints (variable and parameter function): - Total distributed power that distributed from each generator should not greater or equal than the distribution power limit π11 + π12 + π13 ≤ πΎ1 π21 + π22 + π23 ≤ πΎ2 π31 + π32 + π33 ≤ πΎ3 Ab. Alkaff - Total distributed power to every region should not less than every region demand π11 + π21 + π31 ≥ π1 π12 + π22 + π32 ≥ π2 π13 + π23 + π33 ≥ π3 - Distributed power to every region should not be negative πππ ≥ 0 Ab. Alkaff e. Common mathematical models that fit: 3 3 Min Z = ΰ· ΰ· πΆππ πππ π=1 π=1 With constraints: 3 ΰ· πππ ≤ πΎπ π = 1,2,3 π=1 3 ΰ· πππ ≥ ππ π = 1,2,3 π=1 πππ ≥ 0 Model (Mathematics) Transportation Ab. Alkaff 2. Model Solving Solved by using transportation method, assume the obtained values: π13 = 70 π31 = 140 π22 = 80 π32 = 10 And the other πππ = 0 The total distribution cost π is equal to 560 (more minimum than the other alternatives) Electric Power Distribution Solution Scheme Generator Region 1 100 80 ? 140 4 1 2 3 6 ? ? ? ? 2 ? 70 3 ? Abe Alkaff Ab. Alkaff 3. Solution Interpretation • Generator 1 Distribute 70 MW Power to Region 3 • Generator 2 Distribute 80 MW Power to Region 2 • Generator 3 Distribute 140 MW Power to Region 1 and 10 MW to Region 2 • Power Demand Fulfilled in Every Region • There are 30MW Generator Capacity Remains in Generator 1 Ab. Alkaff 4. Sensitivity Analysis • What if generator 1 capacity is reduced to 90 MW? • What if generator 3 capacity is reduced to 130 MW? • What happens if the distribution cost from generator 1 to region 2 is increased from 6 to 7? • Idem, if it is reduced from 6 to 5? • Idem, if it is reduced from 6 to 2? • What if demand of region 3 is reduced to 60 MW? • What if demand of region 3 is increased to 80 MW? • What if demand of region 1 is increased from 140 MW to 150 MW? Exercise 1 Abe Alkaff A company produces three types of products using three types of raw materials. To make the first product, it takes 1 unit of raw material 1. To make the second product, it takes 2 units of raw materials 1, 1 unit of raw materials 2, and 1 unit of the first product. To make the third product, it takes 4 units of 1 raw material, 1 unit of 2 raw materials, 1 unit of 3 raw materials, and 1 unit of the second product. There are 50 raw material 1, 20 units of raw material 2, and 5 units of raw material 3. Find the number of each product that must be made so that the company gets the maximum profit. The first product's profit is 1 per unit, the second product is 7 per unit, and the third product is 15 per unit. Problem Scheme Abe Alkaff Ab. Alkaff Variable: Parameter: Constraint: Criteria: Ab. Alkaff Mathematical Model: Sensitivity Analysis: Exercise 2 Abe Alkaff The company manufactures 2 types of products and draws up a production plan for the next 3 months to meet all monthly demands. Companies can work regular working hours (RT) or overtime (OT). The company can produce more than demand and save the results for next month's use. Production capacity, demand, production costs, monthly storage costs are given in the table. Create a mathematical model that can be used to develop a production plan that minimizes the total cost of production and storage cost! Problem Scheme Abe Alkaff Ab. Alkaff Variable: Parameter: Constraint: Criteria: Ab. Alkaff Mathematical Model: Sensitivity Analysis: Exercise 3 Abe Alkaff Plane manufacturing companies want to fulfill orders for jet planes that meet specific specifications by their customers. The company received orders from three customers. However, because the factory is still busy fulfilling previous orders, not all orders could be done. Thus, a decision had to be made on how many planes to work for each customer. Parameter Initial production costs Profits per plane Factory production capacity Maximum number of planes order Customer 1 2 3 $ 3 Million $ 2 Million 0 $ 2 Million $ 3 Million $ 0.8 Million 20% 40% 20% 3 2 5 The table above shows the initial costs of manufacturing the plane, the benefits obtained from each customer, the factory capacity used in the manufacture of the plane, and the maximum orders from each customer. Find the number of planes that can be produced for each customer for maximum profit. Abe Alkaff Variable: 1. Number of plane made for a particular customer (π₯π ; ππ πππ‘ππππ) 2. Processing certain customer requests, for start-up costs (π¦π ; ππ ππππππ¦: 0 → not processed, 1→processed) Parameter: 1. Initial production costs: • Customer 1 = $3 Million • Customer 2 = $2 Million • Customer 3 = 0 3. Factory capacity for each plane: • Customer 1 = 20% • Customer 2 = 40% • Customer 3 = 20% 2. Profits from each plane: • Customer 1 = $2 Million • Customer 2 = $3 Million • Customer 3 = $0.8 Million 4. Maximum order: • Customer 1 = 3 • Customer 2 = 2 • Customer 3 = 5 Abe Alkaff Objective: Maximizing the profits that can be obtained from plane production (profit per unit minus initial production costs) πππ₯ π = 2π₯1 + 3π₯2 + 0,8π₯3 − 3π¦1 − 2π¦2 Constraint: 20π₯1 + 40π₯2 + 20π₯3 ≤ 100 (Factory production capacity) π₯1 ≤ 3 (Maximum order for Customer 1) π₯2 ≤ 2 (Maximum order for Customer 2) π₯3 ≤ 5 (Maximum order for Customer 3) π₯1 ≤ ππ¦1 (Indicating that the work for Customer 1, if the Customer 1 order is done, then the value π¦1 = 1, where the value of M is a very large number) π₯2 ≤ ππ¦2 (Indicating that the work for Customer 2, if the Customer 2 order is done, then the value π¦2 = 1, where the value of M is a very large number) Abe Alkaff Mathematical model formulation: πππ₯ ππ 2π₯1 + 3π₯2 + 0,8π₯3 − 3π¦1 − 2π¦2 20π₯1 + 40π₯2 + 20π₯3 ≤ 100 π₯1 ≤ 3 π₯2 ≤ 2 π₯3 ≤ 5 π₯1 ≤ ππ¦1 π₯2 ≤ ππ¦2 Solution: π₯1 = 0; π₯2 = 2; π₯3 = 1; π¦1 = 0; π¦2 = 1 Z = 4.8 Interpretation: Producing 0 plane for customer 1, 2 planes for customer 2, 1 plane for customer 3. With maximum profit of $4.8 million Sensitivity Analysis Abe Alkaff Exercise 4 Abe Alkaff A bus company must provide drivers for its buses. Needs every hour varies. There are 3 types of drivers, namely a part-time driver who works for 4 hours with salary of $ 8 / hour, a full-time driver who works for 8 hours continuously with salary of $ 12 / hour, then a split-time driver who works for 4 hours, takes a break. 4 hours, and back to work 4 hours with salary of $ 14 / hour. The number of drivers available is limited. Full-time and split-time drivers come from the same bus driver service provider, namely as many as 20 people. The number of people available can be added because the salary is increased by 25% from the initial salary. Meanwhile, the availability of part-time drivers is 10 people. Each driver can only work at most 1 shift per day. However, shifts that started at the end of the day can resume the next day. How many drivers are needed of each type to meet the daily needs of the least cost (salary)? Exercise 4 Abe Alkaff Abe Alkaff Variable: 1. Total demanded driver for each of types in specific period of time. a. ful-time → ππ b. part-time → ππ c. split-time → π π d. Additional full-time → π π e. Additional split-time → π π Parameter: 1. Driver salary (part-time = $8/jam, full-time = $12/jam, split-time = $14/jam, additional fulltime = $15/jam, additional split-time = $17.5/jam) π1 = 4 β’ π5 = 12 2. β’Driver demand for each time period. (π π ) β’ π2 = 8 =4 β’β’ ππ 3 1= 10 =8 β’β’ ππ 4 2= 7 β’ π3 = 10 β’ π4 = 7 3. 4. β’ π6 = 4 β’ π5 = 12 β’ π6 = 4 Part-time driver availability = 10 Driver availability full-time f dan split-time s = 20 Abe Alkaff Objective: The minimum total cost (salary) to be spent by the bus transportation company. π = 4 8 π1 + 8 12 π1 + 8 14 π 1 + 8 15 π1 + 8 17.5 π1 + β― + 4 8 π6 + 8 12 π6 + 8 14 π 6 + 8 15 π6 + 8 17.5 π6 Constraint: 1. The number of drivers working in a certain period of time must not exceed the number of drivers available. π1 + π 1 + π2 + π 2 + π3 + π 3 + π4 + π 4 + π5 + π 5 + π6 + π 6 ≤ 20 π1 + π2 + π3 + π4 + π5 + π6 ≤ 10 2. The number of drivers working in a certain period of time must not be less than the driver's demand for that time period. π1 + π6 + π1 + π 1 + π 5 + π1 + π6 + π1 + π5 ≥ 4 π2 + π1 + π2 + π 2 + π 6 + π2 + π1 + π2 + π6 ≥ 8 π3 + π2 + π3 + π 3 + π 1 + π3 + π2 + π3 + π1 ≥ 10 π4 + π3 + π4 + π 4 + π 2 + π4 + π3 + π4 + π2 ≥ 7 π5 + π4 + π5 + π 5 + π 3 + π5 + π4 + π5 + π3 ≥ 12 π6 + π5 + π6 + π 6 + π 4 + π6 + π5 + π6 + π4 ≥ 4 Abe Alkaff Mathematical Model: 6 πππ₯ π = 4 ΰ·(8ππ‘ + 2(12ππ‘ ) + 2(14π π‘ ) + 2(15ππ‘ ) + 2(17.5ππ‘ )) π‘=1 ππ ππ‘ + ππ‘−1 + ππ‘ + π π‘ + π π‘−2 + ππ‘ + ππ‘−1 + ππ‘ + ππ‘−2 ≥ π π for π‘ = 1,2,3,4,5,6 6 ΰ· ππ‘ + π π‘ ≤ 20 π‘=1 6 ΰ· ππ‘ ≤ 10 π‘=1 Solution: π1 = 1; π2 = 8; π3 = 2; π4 = 7; π5 = 5; π6 = 3; π6 = 1 With value of π = 2016 Interpretation: 1. 2. 3. Provides 8 people full-time driver for period 2, 7 people for period 4, 3 people for period 6. Provides 1 person part-time driver for period 1 and 6, 2 people for period 3, and 5 people for period 5. Spends cost in a total of $2,016 for paying the driver salary in 1 day. Course Plan 1. 2. 3. 4. 5. 6. 7. 8. Abe Alkaff Optimization concepts Basics of optimization mathematics Analytical solution of optimization problem Numerical Solution of optimization problem: bisection, bisection with derivative, golden section, steepest descent, conjugate gradient, Newton Linear programming: standard and non-standard linear programming Linear programming variations: integer, mixed, quadratic, transportation, assignment, and transhipment Dynamic programming: standard, continue, stochastic, optimal control, and long-term Heuristics Methods Grading Abe Alkaff • Weekly Assignments (+/- 8 Assignments): 35-40% • Report • Presentation • Mid-term Exam, Final Exam/Project: 50-60% • Mid-term, Final Exam: Open Book • Class Participation: 10% • Weekly Assignments Or Final Project Is Worked in Group of 2 People • Assignments is graded based on: • Substance(Complexity, Realistic, Correctness) • Systematic (Completeness, Order) • Language aspect Assignment 1 Abe Alkaff 1. Do the exercise 2 2. Change this slide into article in word format 3. Find an optimization problem from anywhere • Write the problem description • Define and symbolize: • variable value that should be known, • Parameter that should be used, • Constraints that should be fulfilled, and • Criteria that should be optimized • Create the mathematical model • Find the solution, which is the variable value that should be known (random or could be taken from the source that is used) • Give an interpretation to the solution, which is interpretation from the obtained variable • Think about some sensitivity analysis that could occurs on that problem • Submit it in form of softcopy via email to abealkaff@gmail.com with cc to drahmadi41@gmail.com and to share-its in Saturday no later than 23:59