DecisioWorksThe business question
How can changeover preferences be expressed when different products run consecutively? This example compares sequences after adjusting color- and container-capacity-related weights. It shows how to inspect trade-offs, rather than assuming that one adjustment will improve every measure.
Input: how product attributes enter scheduling
| Object | Count or attributes | What it tells us |
|---|---|---|
| Demand | 37 records | Products and quantities to schedule |
| Work centers | 2 | Resource structure |
| Materials | 115 records | Material objects and relationships |
| Process-adaptor relationships | 418 records | Compatibility between operations and resources |
| Product attributes | Color, cap type, container capacity | Changeover differences; container capacity is a product attribute |
| Run configuration | Work center 2; 40 positions × 3 cycles | 120 sequence positions in total |
Configuration: change preferences, retain comparison conditions
| Setting | Baseline | Adjusted |
|---|---|---|
| Capacity, demand and inventory constraints | Enabled | Unchanged |
| Sequence and circular changeover constraints | Enabled | Unchanged |
| Color-related weight | 0.1 | 0.8 |
| Container-capacity-related weight | 1 | 1.5 |
| Solver resources | 16 threads / 60-second limit | Unchanged |
The system reads attributes and compatibility relationships, builds constraints and objectives, and returns a sequence with analysis results. Keep both configurations and outputs. Compare physical changeover counts first, then examine weighted evaluation terms.
Output: what did this re-run show?
| Measure | Baseline | Adjusted |
|---|---|---|
| Valid candidates returned | 1 | 1 |
| New result rows | 120 | 120 |
| Sequence changeovers | 41 | 46 |
| Circular changeovers | 6 | 14 |
| Weighted changeover cost | 14.9 | 66.4 |
| Total API request time | 66.57 seconds | 66.50 seconds |
| Action events / errors | 12 / 0 | 12 / 0 |
In this recorded comparison, changeover counts increased after the weights were raised. A feasible plan found within the available resources is not necessarily a proven global optimum. Scores calculated under different weights are also not an improvement percentage. Review sequences, business measures and constraint satisfaction together.
What should the next comparison change?
- Identify the changeover type you want to reduce and confirm attribute meanings and units.
- Hold data and runtime resources constant; compare one preference change at a time.
- Inspect adjacent jobs and cycle transitions to locate the changes.
- Consider delivery, output and shop-floor conditions before keeping the result, adjusting again or returning to the baseline.