Statistical Process Control (SPC)

Overview

Statistical Process Control (SPC) is a method for continuous monitoring and control of manufacturing processes using statistical techniques. The objective is to identify deviations from the stable process state early and take corrective action before defective parts are produced.

SPC is based on the fundamental idea that every process is subject to natural variation. As long as only random causes (Common Causes) are responsible for the variation, the process is considered statistically controlled. When special causes (Special Causes) occur, the variation behavior changes and the control chart shows anomalies.

Fundamental Principle of SPC

Benefits of SPC

• Prevention rather than reaction: Problems are identified before scrap occurs
• Data-based decisions: Interventions are based on statistical signals, not intuition
• Continuous improvement: Trends and patterns become visible
• Documentation: Complete record of process behavior
• Standards compliance: Fulfillment of requirements from IATF 16949, VDA, ISO 9001, and others

Info: SPC is not a one-time analysis, but a continuous process. Control charts are updated in real-time or at regular intervals during production.

Input and Configuration

Setting Up Subgroups

The formation of subgroups is a central aspect of SPC. Each subgroup consists of a small number of parts that were produced under as identical conditions as possible.

Configuration Parameters

Tip: Choose the subgroup size so that the variation within a subgroup reflects only random variation. Typically, 5 consecutive parts are used. For automated processes with low variation, 3 parts may also be sufficient.

Data Input

Enter measurement values subgroup by subgroup in my8data. You have the following options:

1. Direct input: Enter the measurement values line by line (per subgroup)
2. Import: Upload a CSV or Excel file with pre-structured data
3. Clipboard: Paste copied data from other programs

Calculate Control Limits

my8data automatically calculates the control limits from the entered data. The calculation is based on the first 20-25 subgroups (startup phase). These limits can then be fixed for ongoing monitoring.

The factors A₂, D₃, and D₄ are tabulated constants that depend on the subgroup size.

Warning: Calculate control limits only from data of a stable process. First remove all points attributable to special causes and recalculate the limits. Otherwise, the limits will be set too wide and special causes will not be detected.

Control Charts

Types of Control Charts

my8data provides various control chart types that are selected depending on the data type and subgroup size:

Control Charts for Continuous Characteristics (Variable Data)

Structure of a Control Chart

Each control chart consists of two parts:

1. Location Chart (upper chart): Monitors the position of the process (average)

   • Shows x̄ (subgroup average) or individual values
   • Detects shifts and trends in process level

2. Variation Chart (lower chart): Monitors the variation of the process

   • Shows R (range) or s (standard deviation)
   • Detects changes in process variability

Zones of the Control Chart

The area between UCL and LCL is divided into three zones that are relevant for applying violation rules:

Info: In a stable, normally distributed process, approximately 99.73% of all points lie within the 3-sigma limits (UCL/LCL). A point outside these limits is therefore likely attributable to a special cause.

Interpretation of the Control Chart

Violation Rules (Alarm Rules)

Western Electric Rules

The Western Electric Rules (also WECO Rules) are a set of decision rules based on the zone division of the control chart. They detect not only individual extreme values but also systematic patterns that indicate a process change.

my8data applies the following rules by default:

Nelson Rules

The Nelson Rules extend the Western Electric Rules with additional patterns. my8data supports the following Nelson Rules:

Tip: Not all rules need to be active simultaneously. The more rules that are active, the more sensitive the control chart reacts, but the higher the rate of false alarms. For beginners, Rules 1-4 (Western Electric Rules) are recommended.

Configuration of Violation Rules

In my8data you can individually determine which rules should be active:

• Enable/disable individual rules via checkbox
• Adjustment of rule parameters (e.g., number of consecutive points)
• Color-coded identification of violations in the control chart (red = violation)

Response to Violations

When a violation rule is triggered, proceed as follows:

1. Verify: Check whether the violation is due to a measurement error
2. Identify: Search for the special cause (5-Why, Ishikawa diagram)
3. Correct: Eliminate the cause and document the action
4. Monitor: Observe whether the process remains stable after correction

Warning: Do not respond to every single alarm with a process adjustment without identifying the cause. Unjustified interventions in a stable process (so-called overadjustment) lead to an increase in variation and worsen process quality.

Example of a Rule Violation

The following table shows an example of how violation rules work in practice:

Info: In this example, Rule 3 would have given a warning signal at subgroup 21 before the point at subgroup 22 exceeded the control limits. Early detection is the great advantage of combined rule application.