Applies to DecisioWorks v1.4.0

Stamping planning case: from 1,920 order lines to a controlled planning chain

This case uses the reference StampingWorkshop/data.db and the production_planning orchestration recipe to show how orders, capacity, and material readiness become planning inputs. The sample inspected on 2026-08-31 contained 1,920 order lines, 50 capacity records, 38 operation-level material links, 500 kitting records, 25 time units, and 3 work centers. Its database integrity check passed.

1. Input data

Order lines in order_item provide product, due time, and quantity. Routes and operations connect products to work centers. ingredient binds material demand to operations, kitting_information provides material availability by time, and capacity.used is the fraction already occupied or unavailable, from 0 to 1. It is an input, not resource utilization calculated from this run.

2. Constraints and objectives

The supplied recipe enables capacity, demand, and kitting constraints, uses soft capacity, and leaves inventory constraints disabled. Its objective uses job_bias=-0.001 and waittime=1.0, mapping due-date and completion-time pressure to the workshop plan's cumulative-waiting objective material. The configured solver request is 16 threads and 120 seconds, subject to edition and license limits.

3. Processing flow

The action chain is acquisition → constraint parsing → generation → selection → evaluation → optional dispatch. Dispatch leaves the standard database unchanged; only an explicitly configured standalone JSON result file can be written.

4. Outputs and interpretation

Baseline and rerun comparisons should keep the data version fixed, change only a few objectives, rules, or resource conditions, and retain the configuration difference.

5. Observed runs and comparison

The following records come from DecisioWorks v1.4.0 Commercial Edition runs under an Enterprise license on 2026-08-31. The API baseline used 16 threads with a 120-second limit and returned ok in 14.41 seconds. The adjusted run kept the same data, constraints, and compute settings, changing only job_bias from -0.001 to -0.003 and waittime from 1.0 to 2.0; it returned ok in 14.29 seconds. Both runs produced one valid candidate, 66 result records, 12 action events, and no errors. Success here means a usable solution was found, not that global optimality was proved.

For the baseline, total demand and output were both 680,500, final shortage was zero, and quantity coverage was 100%. On-time delivery still requires checking order and time-unit details. Available capacity was 1,188,000; processing load was 371,588.48; planned load including setup and waiting was 377,080.42; utilization was about 31.74%. These capacity and load values use the sample analysis output's units and should not be read directly as minutes.

The adjusted run kept these aggregate metrics unchanged, while the weighted job_bias contribution changed from -7,854.678 to -23,564.034. This confirms that the weight entered the calculation. Equal aggregates do not establish identical task assignments; compare individual planning records as well. Because both weights changed together, the comparison does not isolate the effect of either weight.

The screenshots come from a separate Commercial Edition Scenario Lab run that completed with two threads and 60 seconds. The API records above use 16 threads and 120 seconds. Review each set against its own data version, configuration, and output rather than treating them as one experiment. For a business deployment, apply the same comparison method to your data, equipment conditions, and acceptance criteria.

Reading the original Chinese screenshots

Start with the left-side baseline configuration, then the result status and aggregate metrics on the right. Finally inspect the work-center and time-unit breakdowns and the objective-material audit. The original interface labels are retained to preserve the run evidence.

An em dash in the PlanBias or PlanSignal count means no numeric count is shown there; it should not be read as a confirmed zero.