Skip to content
Rektron

Search

    Solution

    Process control and SPC

    Tell normal variation from real signals and act while you can still make a difference.

    Talk to a specialistHow it works
    Illustration of a control chartMeasured values vary between the control limits. Towards the end six values rise in a row, a trend that is flagged as a signal before the tolerance limit is reached.USLUCLX̄LCLLSLSignal: 6 rising in a row →Measurement 134
    Illustration: control limits (UCL/LCL) from the process variation, tolerance limits (USL/LSL) from the drawing.

    The problem

    Every process varies. The challenge is to tell the normal variation from the changes that mean something is heading the wrong way, and to see it early enough to act.

    Relevant for: Operator, Quality, Production, Engineering and planning

    The context in production

    In machining, injection moulding or an assembly line, a sample of parts is measured against the tolerances on the drawing. As long as the values are within tolerance everything looks fine, but a tool that wears, a temperature that drifts or a new material batch often shows up in the measurement data long before the parts go out of tolerance.

    Statistical process control (SPC) is about using that information: plotting measured values in control charts, calculating how well the process holds the tolerances, and giving operators and engineers a clear signal when the process changes.

    Why it matters

    When a deviation is found only at final inspection or at the customer, the cost has already been incurred: scrap, rework, sorting, complaints and at worst a stopped delivery.

    Customers in, for example, the automotive and medical device industries also require the supplier to show that critical characteristics are under statistical control and that the process is capable.

    In principle

    Signal before defect

    The control limits describe what the process normally does. A pattern is a signal long before anything goes outside the tolerance, and Cp and Cpk show whether the process fits within the requirements.

    How does SPC tell normal variation from a signal?

    Illustration · Example data

    The control limits describe what the process normally does. A pattern, such as six rising values in a row, signals that something has changed, even when all values are within the limits.

    Step by step

    • On the left the values vary randomly around the centre line.
    • On the right six values rise in a row.
    • That is a signal to investigate, although no value is outside the control limits.

    Why can Cp look good when Cpk is poor?

    Illustration · Example data

    Cp only compares the spread with the width of the tolerance. Cpk also takes the position of the mean into account, and drops when the process shifts towards a limit.

    Step by step

    • Two processes with the same spread and the same Cp, 1.58.
    • The centred process has Cpk 1.58.
    • The off-centre process has Cpk 0.79, because the mean is close to the upper limit.

    From today to better

    Common challenges, and what a better way of working looks like

    What it often looks like

    • Measured values are written down on paper or in Excel and analysed afterwards, if at all.
    • Control charts exist, but are not updated at the pace production runs.
    • Operators do not know when to intervene and when to leave the process alone, which leads to both missed signals and over-adjustment.
    • Capability is calculated before a customer audit instead of being followed continuously.
    • Knowledge of how the process behaves sits with a few people.

    A better way of working

    • Measured values are recorded once, directly from the gauge or measuring machine, and linked to part, characteristic, machine and order.
    • Control charts and key figures update when the measurement is made and are shown where the decision is made, at the machine.
    • Clear rules for when the operator should act, and alarms when values go outside the limits.
    • Capability is followed over time, so that improvements and deteriorations show.
    • The history is kept, so that causes can be analysed and actions followed up.

    Rektron's approach

    How we help

    Rektron has worked with measurement data and statistical process control in Swedish manufacturing since the 1980s. In Eyesight, measured values from gauges, measuring machines and production equipment are collected in one shared platform. Trend charts and Xbar-R with control limits set per characteristic, run rules according to Nelson or Western Electric and alarms make the state of the process visible to operators, engineers and management. The run rules apply to manually entered values. Cp and Cpk are followed per characteristic, with the distribution chosen per characteristic.

    Measurement system analysis with Gage R&R is available in Eyesight. In-depth capability studies with more indices, such as Pp/Ppk, CpXR, CpXS and Q according to ISO 3951, are available in Rektron SQC. We help you choose the right set-up for your process.

    The methods are the same ones we teach in Rektron Academy, so the introduction is as much about ways of working as about systems.

    About EyesightKnowledge: Normal variation or signal?

    Capabilities used

    • SPC, run rules and alarms

      Eyesight

      Trend charts and Xbar-R with control limits set per characteristic, an alarm when a value goes outside the limit, and run rules (Nelson or Western Electric) with the rule number at the point. The run rules apply to manually entered values. Changes of operator, machine or batch can be marked in the trend.

      Read more →

    • Cp and Cpk with choice of distribution

      Eyesight

      Cp, Cpk, Pp and Ppk per characteristic, with the distribution chosen per characteristic and a histogram with the fitted distribution next to the trend.

      Read more →

    • MSA / Gage R&R

      Eyesight

      Measurement system analysis with Gage R&R in Eyesight: study, calculation according to AIAG, diagnostic charts and report.

    • Quality control

      Eyesight

      Incoming, first/last-off, in-process and final inspection with inspection instructions right in the interface.

      Read more →

    • Automatic import from CMM and vision systems

      DAQ

      Results from coordinate measuring machines and vision systems are imported without manual handling.

      Read more →

    • Directly connected gauges

      Eyesight

      Digital gauges over USB or Bluetooth, so the value lands in the right place straight away.

      Read more →

    • Analysis

      Eyesight

      Trends, histograms, Cp/Cpk over time, correlation between characteristics and analysis of stops and cycles per machine, part and order.

      Read more →

    • Capability studies and indices

      SQC

      Cm/Cmk, Cp/Cpk, Pp/Ppk, CpXR/CpkXR, CpXS and Q (ISO 3951), with centring, confidence intervals and choice of distribution, for all or filtered parts of the measurement results.

      Read more →

    • MSA with Gage R&R and attribute study

      SQC

      Measurement system analysis in Rektron SQC.

      Read more →

    The flow

    How it works

    1. 01

      Measure

      The value is recorded from a gauge, CMM or machine, or entered at the station.

    2. 02

      Secure

      The value is linked to part, characteristic, tolerances, operator and order.

    3. 03

      Visualise

      Control charts and Cp/Cpk update and are shown to the person who needs to act.

    4. 04

      Control

      Alarms and clear rules show when the process should be adjusted, and when it should be left alone.

    5. 05

      Improve

      The history is used to find causes and follow up the effect of actions.

    What changes

    • Deviations and trends show while you can still affect the outcome.
    • Operators make decisions on data, with the same rules whatever the shift.
    • Capability is followed continuously instead of being produced for an audit.
    • A shared basis for quality, production and the customer.

    Read more

    Technical information

    • Measurement data can come from digital gauges, CMMs and vision systems, PLCs and sensors, or manual entry.
    • Control limits and tolerances are set per characteristic when parts are set up.
    • Import of standard formats such as AQDEF and DMIS.
    Integrations and formats →

    Situations

    Where this becomes relevant

    The situations are described without the product: what happens, what should apply and who needs to know what.

    Knowledge

    Concepts on this page

    Questions

    Frequently asked questions

    What is the difference between tolerance limits and control limits?

    The tolerance limits come from the drawing and say what the customer accepts. The control limits are calculated from the process's own variation and say when the process has changed. A process can be within tolerance and still be out of statistical control, and that is exactly the signal SPC catches early.

    What Cpk value should we aim for?

    A common guideline is Cpk ≥ 1.33, and often higher for critical characteristics. What applies to you is usually set by the customer's requirements. Cpk assumes a stable process; for a process that is not stable, Ppk is used.

    Can we start with a few characteristics?

    Yes, and that is often the best way. Start with the characteristics that are critical to the customer or cause the most problems, and build on that once the way of working is in place.

    Related

    Solutions that fit together

    Next step

    Want to see how process control and SPC can work for you?

    Tell us about your process and your measurements. We will suggest a first step that suits your business.

    Talk to a specialistBook a demo