Tutorial: Optimizing Injection Molding Parameters

The Problem

A plastics injection molding line is producing parts that fail tensile strength specifications. The engineering team suspects three process parameters:

• Temperature: resin temperature
• Pressure: injection pressure
• CycleTime: cooling time

Changing one parameter at a time only reveals its effect under one fixed combination of the other two, and misses interactions between parameters. A designed experiment that varies all three simultaneously isolates each parameter's contribution with fewer total runs.

Design the Experiment

Select Data > New DoE Design... from the menu bar. The DoE Analysis tab opens with the design wizard.

Enter Factors and Levels

Enter the three factors with the two conditions you plan to test:

Factor name	Level 1	Level 2
Temperature	High	Low
Pressure	High	Low
CycleTime	Short	Long

Choose an Array Type

An orthogonal array is an experiment plan table in which factor level combinations are balanced. Every pair of factors has all level combinations appearing equally often, which allows each factor's effect to be estimated independently. The array type determines how many experimental runs you need. The number after L is the run count, and each type has a limit on the number of factors it can handle.

Type	Runs	Max factors
L4	4	3
L8	8	7
L16	16	15

You can use any array type whose maximum is at least your number of factors. A larger array than the minimum gives more data to separate effects from noise.

With 3 factors, all three types are available. L4 needs only 4 runs, but with 3 factors the main effects alone consume all degrees of freedom, leaving zero residual degrees of freedom and making it impossible to test whether effects are real. L8 uses 8 runs, which leaves residual degrees of freedom for error estimation and allows proper hypothesis testing.

Plan Replication

An orthogonal array specifies one run per condition. Replicating each condition provides an estimate of within-condition variability (error). In this tutorial, each condition is replicated once, giving 16 total runs.

The wizard generates one set of conditions. Add replicate rows after running the experiment.

Generate the Dataset

Confirm that Randomize run order is enabled. Randomization prevents systematic bias from factors like equipment warm-up or material batch variation that correlate with run order.

Click Preview to inspect the experiment plan. You will see 8 rows combining Temperature, Pressure, and CycleTime at High/Low settings, with the Response column empty. The idea is to run the experiment following this table and fill in the measured strength for each row.

Click Generate to add this experiment plan as a dataset in your project.

Run the Experiment and Enter Data

Open the generated dataset in the Data Table tab. Factor columns contain the experimental conditions; the Response column is empty. Run the experiment and double-click cells in the Response column to enter measured strength values.

For this tutorial, use the pre-filled sample data instead. Click Injection Molding in the Sample Data section of the launcher. This loads 16 rows: 3 factors x 2 levels x 2 replicates.

Column	Description
Temperature	Resin temperature. High or Low
Pressure	Injection pressure. High or Low
CycleTime	Cooling time. Short or Long
Strength	Tensile strength in MPa

Run the Analysis

Select Analysis > DoE Analysis... from the menu bar.

1. Select Injection Molding from Dataset
2. Select Strength for Response Variable
3. Check Temperature, Pressure, and CycleTime under Factors
4. Leave Model set to Main effects only. Start with main effects to understand each factor's contribution, then add interactions if needed
5. Leave Significance Level at the default 0.05

If you select the wrong variable, change the dropdown or checkbox and click Run Analysis again. There is no need to start over.

Click Run Analysis.

Read the Main Effects Plot

Click the Main Effects sub-tab and check Show 95% confidence intervals.

Each factor's mean strength at each level is shown with confidence intervals. The difference between levels and the width of the intervals show how much each factor affects strength.

• Pressure: Mean 41.63 MPa at High, 34.95 MPa at Low. The difference is about 6.7 MPa, the largest of the three factors
• Temperature: Mean 40.50 MPa at High, 36.07 MPa at Low. The difference is about 4.4 MPa
• CycleTime: Mean 39.27 MPa at Long, 37.30 MPa at Short. The difference is about 2.0 MPa

Narrow confidence intervals indicate precise estimation. The horizontal dashed line is the grand mean.

Check for Interactions

Change Model to Main effects + all 2-factor interactions in the settings panel and click Run Analysis.

An interaction means that one factor's effect changes depending on the level of another factor. For example, if injection pressure only improves strength when resin temperature is High, then Temperature and Pressure have an interaction. When factors interact, changing one factor alone may not produce the expected result, so you need to find the best combination.

The Interaction sub-tab shows interaction plots. If the two lines in each subplot are roughly parallel, interactions between that factor pair are small. The more the lines cross, the larger the interaction.

In this dataset, lines are nearly parallel across all three factor pairs. The p-values for interaction terms in the ANOVA table confirm this. If an interaction is significant, examine the cell means in the interaction plot and choose the factor combination that yields the best response.

Check Assumptions

The analysis above relies on certain statistical assumptions, such as similar measurement variability across conditions. Diagnostics for these assumptions appear at the bottom of the ANOVA Table sub-tab.

With balanced orthogonal arrays, these assumptions rarely break down. When they do, the data itself usually looks obviously wrong. You can safely skip this section at first. For details on reading the diagnostics, see DoE Analysis assumptions.

Make Decisions from the Results

The analysis shows:

• Pressure has the largest effect on Strength: about 6.7 MPa higher at High than Low
• Temperature increases strength by about 4.4 MPa at High, but the effect is smaller than Pressure
• CycleTime has a relatively small effect of about 2.0 MPa
• Interactions between the three factors are small. Each factor's effect is consistent regardless of the other factors' levels

Small interactions mean the three parameters can be optimized independently. Changing Pressure to High is the most effective single action, followed by Temperature. Whether extending CycleTime is worthwhile depends on whether the ~2 MPa gain justifies the longer cycle.

This experiment used randomized assignment of conditions, so the estimated effects support a causal interpretation within these experimental conditions. The effects are specific to the materials, equipment, and factor levels used in this experiment. A typical next step is a confirmation run at Pressure = High, Temperature = High to verify that the same improvement holds under actual production conditions.

Related Pages

• DoE Analysis -- Detailed usage and statistical model documentation
• ANOVA -- One-way and two-way analysis of variance