What is industrial statistics?
Industrial statistics refers to the use of statistical procedures in industrial manufacturing and quality management. It helps to systematically analyse production data, monitor processes, detect errors at an early stage and continuously improve product quality. At the same time, it provides statistics
The beginnings of industrial statistics
The application of statistical procedures in industrial manufacturing can be broadly divided into three phases: in the 1930s, Walter Shewhart developed the quality control chart. In the early 1980s, Ford introduced systematic SPC into the supply chain with the Q101 programme. Today, artificial intelligence is finding its way into statistical process control.
Walter Shewhart and the birth of the quality control chart
As quality standards rose and mass production became established, industry moved towards sample inspection. In the 1930s, Walter Shewhart developed the quality control charts named after him at Bell Telephone Laboratories, thereby laying the foundations for Statistical Process Control (SPC). He is still regarded today as the ‘father of SPC’.
Statistical quality assurance in Germany: From the ASQ to the DGQ
In 1952, a group was formed in Germany to promote the introduction of American statistical procedures into domestic industry. This gave rise to the Working Group on Statistical Quality Control (ASQ), which was renamed the German Society for Quality (DGQ) in 1968. Industrial statistics were disseminated throughout the German-speaking world largely through targeted training programmes organised by the DGQ.
The breakthrough: Ford and the widespread introduction of SPC
The mere existence of statistical procedures does not in itself guarantee their widespread use. Only a certain level of pressure to ensure quality – for example, through audits – ensures broad implementation. This is precisely the approach Ford took.
Q101: How a car manufacturer revolutionised quality assurance
In the early 1980s, Ford introduced the Q101 quality management system, making the use of SPC methods mandatory both internally and amongst its suppliers. Shewhart’s quality control charts and capability indices for assessing machines, processes and products became compulsory, and their implementation was continuously monitored as part of customer audits. This significantly accelerated the spread of industrial statistics throughout entire supply chains.
SPC software put to the test: the Ford test examples
Ford recognised early on that faulty software programmes led to incorrect capability indices and intervention limits. To verify SPC systems, test examples with defined data sets, results and graphics were developed. Today, comparable reference data sets are included in ISO/TR 11462-3:2020 and are used to validate SPC software against the requirements of the ISO 7870 and ISO 22514 series.
International standards and guidelines for industrial statistics
QS-9000, AIAG and the development of ISO/TS 16949
In 1994, the AIAG (Chrysler, Ford, General Motors) published QS-9000, a uniform standard for SPC and measurement system analysis (MSA). As suppliers had to comply with the national requirements of various associations at the same time, the International Automotive Task Force (IATF), in collaboration with the ISO, developed ISO/TS 16949, a globally applicable specification which was published in 1999. QS-9000 was withdrawn in 2006.
ISO series of standards for quality control charts and capability analysis
ISO Technical Committee TC 69 is responsible for the relevant statistical standards. Of particular relevance to industrial statistics are:
- ISO 7870 (Parts 1–9): Control Charts
- ISO 22514 (Parts 1–9): Capability and Performance
- ISO 11462 (Parts 1–5): Guidelines for implementation of statistical process control (SPC)
ISO 22514 Part 2 describes time-dependent process models (A1, A2, B, C1–C4, D) and forms the basis for the correct calculation of capability indices, depending on the process behaviour in terms of variation and location over time.
The new harmonised SPC manual from AIAG and VDA
AIAG and VDA are planning to jointly publish an SPC manual based on the AIAG Reference Manual and VDA Volume 4. It will define a standardised approach to quality control chart techniques and capability indices, and will in future form part of IATF 16949 certification.
What SPC really achieves and where its limitations lie
SPC addresses two key areas: the analysis of short-term and long-term process capability, and ongoing process monitoring using quality control charts. As long as the control chart shows stability, the process state has not changed significantly.
Digital twins, sensor technology and predictive quality assurance
By correlating material, production and environmental data, a digital twin of the process can be created, which is continuously compared with the actual status. This enables deviations to be identified and corrected at an early stage. Advantages:
- Reduced measurement effort without compromising on quality
- Transparency regarding relevant influencing quantities
- Automated parameter adjustment following adequate system validation
Industrial statistics encompasses further AI applications, including the automated determination of process models and capability indices, AI-assisted visual inspections using neural networks, and time-series analyses for trend forecasting.
Implementing AI projects in a structured way: the CRISP model
The CRISP model (CRoss-Industry Standard Process for Data Mining) has established itself as a tried-and-tested process model for AI projects. Key factors for the success of such projects include a sufficient and high-quality data basis, interdisciplinary expertise in process management and data analysis, and robust validation of the solution developed.
The basis for SPC: inspection planning and validated measuring systems
A fundamental element of industrial statistics is inspection planning. VDA Volume 5 sets out the requirements for the capability, planning and management of measurement and inspection processes. In future, AI-supported selection systems will be able to automatically suggest to test planners only those measuring instruments and systems that are classified as suitable and qualified, thereby building confidence in the results of statistical process control (SPC).
This full-length article was written by Dr.-Ing. Edgar Dietrich and is available for download here.