Integrated Production & Maintenance Optimization Engine

A conflict observed in industrial practice

Before beginning this research, I designed and implemented computerized maintenance management software for multiple companies. Across those settings, I repeatedly observed preventive-maintenance activities being deferred or omitted when they competed with immediate production priorities.

The problem was not simply a lack of maintenance plans or information systems. Production and maintenance were being planned as separate activities, so conflicts were often resolved only after the production schedule had been established.

I developed an integrated optimization approach in which production sequencing and maintenance timing are determined together.

The decision problem

The research focused on permutation flow shops: environments in which multiple jobs pass through the same sequence of machines. Job sequence affects machine idle time, work waiting between operations, completion times, and delivery tardiness.

Conventional scheduling models commonly assume that machines remain continuously available and that planning information is complete and certain. Operationally, however, machines must stop for maintenance, maintenance requirements depend on accumulated operating time, and processing or maintenance durations may be uncertain.

The engine therefore represents production jobs and meter- or age-based maintenance activities within a common scheduling problem. Its primary objectives are to reduce job tardiness and maintenance cost while satisfying production and equipment requirements.

Three connected formulations

I developed three progressively richer formulations:

1. Deterministic mixed-integer model: Integrates multiple age-based maintenance activities with permutation flow-shop scheduling under known inputs.
2. Fuzzy bi-objective model: Represents the trade-off between production and maintenance objectives and accounts for the effect of maintenance on subsequent processing times.
3. Stochastic mixed-integer model: Introduces uncertainty in processing and maintenance times and supports the combination of compatible maintenance activities.

Because flow-shop scheduling is computationally difficult as the number of jobs increases, I also studied the solution space and developed Genetic Algorithm approaches for the deterministic and fuzzy models. The stochastic formulation was solved through simulation–optimization.

These approaches reduced dependence on commercial mathematical-programming solvers and made the methods more suitable for incorporation into existing operational information systems.

What the engine produces

Given a set of jobs, their machine sequence and processing requirements, due dates, maintenance activities, operating-age thresholds, and maintenance durations, the engine produces a coordinated plan for job processing and maintenance execution.

The result is a single schedule in which maintenance windows are selected as part of the production decision. Planners can therefore evaluate delivery performance and maintenance cost together instead of resolving conflicts between two independently developed plans.

The formulations also capture practical maintenance details that simpler scheduling models omit, including multiple maintenance types, the combination of maintenance activities, maintenance effects on later processing, and uncertainty in task durations.

Validation and evidence

I evaluated the three formulations through extensive computational experiments and three variations of an operations-and-maintenance case study involving construction machinery. Each dissertation chapter applied its respective model and solution method to a version of this case.

This validation demonstrated the computational behavior and applicability of the proposed methods under the studied conditions. It was computational and case-based validation—not evidence of production deployment or measured plant-floor improvement.

Current status and significance

The research was completed in 2018. It provides an implementation-oriented foundation for planning tools that coordinate production commitments with asset-maintenance requirements.

Its practical significance lies in treating maintenance as part of the production decision rather than as an interruption to an already finalized schedule. The solver-independent methods were designed to support automation and incorporation into manufacturing or service information systems. Any later software integration, organizational adoption, or operational benefit should be documented separately if evidence becomes available.