SPC Practical Guide
1. Overview of SPC
SPC (Statistical Process Control, statistical process control) is a technical method that applies statistical techniques to evaluate and monitor various stages of a process, establishing and maintaining the process at an acceptable and stable level, thereby ensuring that products and services meet specified requirements.
Core Value: SPC emphasizes prevention, monitoring process variation to promptly identify anomalies and take action, preventing the production of nonconforming products and reducing quality costs.
2. Types and Selection of Control Charts
Variable Control Charts
Xbar-R Chart: Mean - Range Chart, most commonly used, sample size 2-9
Xbar-S Chart: Mean - Standard Deviation Chart, sample size ≥10
I-MR Chart: Individual - Moving Range Chart, sample size = 1
Attribute Control Charts
P Chart: Nonconforming Product Rate Chart, variable sample size
NP Chart: Nonconforming Product Number Chart, fixed sample size
U Chart: Defects per Unit Chart, variable sample size
C Chart: Defects Number Chart, fixed sample size
| Data Type | Sample Size | Recommended Control Chart | Application Scenario |
|---|---|---|---|
| Variable | 2-9 | Xbar-R Chart | Continuous data such as dimensions, weight, time |
| Variable | ≥10 | Xbar-S Chart | Continuous data with large sample sizes |
| Variable | =1 | I-MR Chart | Single-piece production, high inspection costs |
| Attribute | Variable | P Chart | Monitoring nonconforming product rates |
| Attribute | Fixed | NP Chart | Monitoring the number of nonconforming products |
3. Steps to Implement Control Charts
1. Preparation Stage
- Identify critical quality characteristics (CTQ);
- Select the appropriate type of control chart;
- Determine sampling frequency and sample size;
- Train operators.
2. Collect Initial Data
- Collect at least 25 data sets (for Xbar-R Chart);
- Ensure the process is stable with no anomalies;
- Record data and time.
3. Calculate Control Limits
- Calculate the mean (Xbar);
- Calculate the range (R);
- Calculate UCL, CL, LCL.
4. Plot the Control Chart
- Plot the center line (CL);
- Plot the upper and lower control limits (UCL/LCL);
- Plot and connect the data points.
5. Monitoring and Improvement
- Continuously monitor the process;
- Identify anomalies and take action;
- Continually improve process capability.
4. Eight Out-of-Control Criteria
| Criterion | Description | Possible Causes |
|---|---|---|
| Criterion 1 | 1 point beyond 3σ control limits | Process anomaly, measurement error |
| Criterion 2 | 9 consecutive points on one side of the center line | Process shift, tool wear |
| Criterion 3 | 6 consecutive points increasing or decreasing | Tool wear, temperature changes |
| Criterion 4 | 14 consecutive points alternating up and down | Two processes alternating, data stratification |
| Criterion 5 | 2 out of 3 consecutive points beyond 2σ | Increased process variation |
| Criterion 6 | 4 out of 5 consecutive points beyond 1σ | Process shift |
| Criterion 7 | 15 consecutive points within 1σ | Data stratification, insufficient measurement resolution |
| Criterion 8 | 8 consecutive points beyond 1σ | Mixing of two processes |
5. Process Capability Analysis
CPK vs. PPK
| Index | Definition | Requirement | Applicable Stage |
|---|---|---|---|
| CPK | Process capability index (stable process) | ≥1.33 | Mass production stage |
| PPK | Process performance index (initial process) | ≥1.67 | Pilot production stage |
CPK Calculation Formula
CPK = min(CPU, CPL)
CPU = (USL - μ) / 3σ
CPL = (μ - LSL) / 3σ
Where: USL = Upper Specification Limit, LSL = Lower Specification Limit, μ = Mean, σ = Standard Deviation
CPK Rating
- CPK ≥ 1.67: Process capability is excessive, consider cost reduction
- 1.33 ≤ CPK < 1.67: Process capability is adequate, maintain the status quo
- 1.0 ≤ CPK < 1.33: Process capability is insufficient, improvements are needed
- CPK < 1.0: Process capability is severely insufficient, must be improved
6. Classic Case Study in the Automotive Industry
Case Study: Dimensional Control of an Automotive Component
Background: The outer diameter requirement for a shaft component is Φ50±0.05mm, and the customer has complained about significant dimensional variation.
Implementation of SPC:
- Select an Xbar-R control chart with a sample size of n=5, sampling once per hour;
- Collect 25 initial data sets and calculate the control limits;
- Identify that the 18th data set exceeds the control limits;
- Investigate the cause: Tool wear leads to dimensional shift;
- Take action: Replace the tool and adjust process parameters;
- Recalculate CPK, improving it from 0.95 to 1.52.
Effect: The customer complaint rate decreased by 80%, and process stability was significantly improved.
7. Frequently Asked Questions
Q1: How often should control charts and control limits be updated?
A1: Initial control limits are calculated based on 25 data sets. Recalculate in the following situations: 1) Process change; 2) Equipment change; 3) Material change; 4) At least once a year for review.
Q2: What should be done if an anomaly is detected?
A2: 1) Immediately stop the machine; 2) Identify and isolate suspect products; 3) Investigate the cause; 4) Implement corrective actions; 5) Verify the effectiveness of the actions; 6) Resume production.
Q3: How can insufficient CPK be improved?
A3: 1) Reduce process variation (optimize processes, maintain equipment); 2) Adjust the process center (adjust equipment parameters); 3) Relax tolerances (requires customer approval).
Q4: How to choose between variable and attribute control charts?
A4: Prioritize variable control charts (more information, higher sensitivity). Use attribute control charts when measurement is not possible (e.g., visual inspection).
Q5: How to determine the sample size?
A5: For Xbar-R Chart, recommend n=4-6; for Xbar-S Chart, n≥10; for I-MR Chart, n=1. Larger sample sizes increase detection capability but also increase costs.
Q6: Is SPC software necessary?
A6: For small batch production, Excel can be used. For large batches and multiple characteristics, it is recommended to use professional SPC software (such as Minitab, JMP) for higher efficiency and automatic anomaly detection.