Optimization-Based Production Scheduling for Complex Manufacturing Plants Delivered as a Service using High Performance Computing Architecture &amp; Algorithms
Advanced manufacturing companies require significant decision-making support to schedule their day-to-day operations. However, there is a significant weakness in todays Optimization-based approaches to scheduling, in that they often suffer from the curse of dimensionality in cases where the problem solution cannot be scaled up to solve practical problems. This project addresses DOEs interest in turn-key solutions advancing the use of HPC technology in manufacturing and is intended to result in increased manufacturing efficiency and job creation in Phase III. We propose to develop an optimization-based scheduling application that uses High Performance Computing technology to address the curse of dimensionality issues. In Phase I, we will develop the generic scheduling framework and prove the concept on a test-bed and in Phase II, we will address the deeper technical commercial aspects of handling data confidentiality and security and we will also harden the model and the solver code for success in the commercial environment. The broad long-tem objectives and specific aims of this phase I project can be summarized as follows: 1. Develop a general framework for the formulation of optimization-based scheduling models, a. Develop a unified representation of production scheduling, b. Formulate a mixed-integer programming (MIP) model, c. Extend the model to handle restrictions from a wide range of sectors; 2. Design and develop a HPC/HTC solution algorithm for solving the scheduling model, a. Implement and test known MIP solutions techniques in SPC resources, b. Develop new problem-specific decomposition methods, c. Develop a suite of solution methods for different classes of problems; 3. Design and develop a prototype, a. Develop user interface (UI1) where users can define their problem, b. Develop interface (UI2) to interface UI1 with the code-behind, c. Automate the generation of the optimization model from UI1 and UI2, d. Demonstrate the prototype with a real-life example data set. Commercial Applications and Other Benefits: The envisioned commercial production scheduling software application will be used to obtain optimal production schedules at complex multi-product multi-line manufacturing sites. This has the potential to result in significant efficiencies and reductions in operating costs including reduced utilities usage, reduced waste, reduced downtime, etc. It is intended to increase the global competitiveness of the manufacturing sector and lead to job creation. The goal of this SBIR Phase I effort is to demonstrate the feasibility of delivering a next generation optimization-based production scheduling application using high-performance computing (HPC) or High Throughput Computing (HTC) architecture and algorithms. The problem of scheduling a set of jobs on a set of machines is well-known. Various algorithms are available to choose from. Optimization-based algorithms attempt to optimize an objective function (e.g., minimize production cost) and meet all constraints simultaneouslyas opposed to methods that make decisions sequentially. Optimization-based approaches are essentialand probably the only effective approachin cases where the scheduling problem is severely constrained or when conflicting objectives exist. Optimzation-based approaches suffer from the curse of dimensionality making the approach difficult to scale up to address commercial problems. However, enhancements in computing power and the sophistication of solvers and modeling languages have given us hope to overcome the problem of dimensionality and produce practical industrial scale scheduling solutions. Based on our experience, we believe that the right algorithm design and the right computing architecture can produce a breakthrough industrial strength optimization-based scheduling application. The application envisioned here also aims at lowering the barrier for manufacturing (non-science) businesses to leverage the HPC/HTC resources to obtain optimal production schedules. We propose to do this by encapsulating the mathematical model in the most generic waythat addresses a broad spectrum of constraints, parameters, objectives and structural elements. The barrier is also lowered by moving the problem formulation and solution to behind the scenes (away from the operator) so that the end-user need not have an advanced math degree to operate the application to obtain an optimal schedule. The proposed approach is radically different from existing scheduling solutions used by most manufacturing facilitieswhich are based on rules of thumb.
Small Business Information at Submission:
Optimal Solutions, Inc.
17 Kershaw Ct. Bridgewater, NJ 08807-2595
Number of Employees: