A university within a business school 1 Integrating Resource Planning with Job Scheduling for Service Optimization Gang Li Bentley University Waltham, MA Joint work with Anant Balakrishnan University of Texas Austin, TX Brian Roth BNSF Railway Fort Worth, TX
28
Embed
Integrating Resource Planning with Job Scheduling for Service Optimization
Integrating Resource Planning with Job Scheduling for Service Optimization. Gang Li Bentley University Waltham, MA Joint work with. Brian Roth BNSF Railway Fort Worth, TX. Anant Balakrishnan University of Texas Austin, TX. Service Optimization. Outline Motivation - PowerPoint PPT Presentation
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
A university within a business school
1
Integrating Resource Planning with Job Scheduling for Service
OptimizationGang Li
Bentley UniversityWaltham, MA
Joint work with
Anant BalakrishnanUniversity of Texas
Austin, TX
Brian RothBNSF Railway Fort Worth, TX
Service Optimization
Outline
1. Motivation2. Modeling the problem3. Solving the problem effectively4. Application5. Conclusion
A general solution procedure (for minimization prob.) Upper Bound:
Apply a heuristic method to find a feasible solution, which provides a upper bound to the decision problem.
Lower Bound: If relaxing integer requirements, the relaxed Linear Programming (LP)
model can be efficiently solved, whose solution provides a lower bound to the decision problem.
Gap: Keep improving both the lower bound and the upper bounds until the percentage difference between the two bounds, (defined as the gap), reaches to zero. We then ensure the optimal solution.
This framework has been implemented in many commercial optimization software, such as CPLEX.
Speeding up the Solution Process Preprocessing stage: Reduce size of the model
Combine (aggregate) jobs Reduce size of repositioning network Reduce job time windows, eliminate variables and constraints
Progressive solution strategy: Solve a series of simpler problems Each problem is an extension of previous problem Optimal solution of previous problem provides a feasible initial
solution and strong lower bound for the succeeding problem Cutting plane method: Dynamically add strong inequalities to
the model Customized branch-and-bound rule Tuned computational parameters of CPLEX
The New Maintenance Plan• chooses dozens of best-fitted teams from hundreds of candidate teams• assigns selected teams to jobs according to teams’ skills and costs• determines a detailed work plan that satisfies all service requirements
25
Manual Planning Process (Before 2005): Cumbersome and time-intensive o Find a feasible plan that satisfies all timing coordination requirements
and time window requirements.Results: Lots of timing-coordination violations
o Balance the three major cost components.Results: Only able to focus on a single cost component, e.g., the routing cost.
Large size of the model: 200,000 variables & 30,000 constraints Resistance to change
Solution: A slow but step-by-step implementation processo S1: Based on the manually selected no. of resources and job-resource
assignment, determine the optimal job scheduling and resource routing.
o S2: Based on the manually selected no. of resources, determine the optimal resource assignment, job scheduling and resource routing.
o S3: Provide an integrated solution, which minimizes the total resource selection, assignment and routing cost.
Change from Manual Planning to Model-based Planning
Conclusion: Main Contributions Addressed a complex resource planning and job
scheduling problem that is a combined capacity planning, resource assignment, and job scheduling problem.
Modeled the problem on a general framework and developed strong inequalities and effective solution strategies to improve the computational performance.
Successfully applied the proposed model and methods to annual railway track maintenance planning in a major railway company.