5.2.7 Xbar-S Chart

Xbar-S Chart Introduction The Xbar-S Chart is a pair of control charts used to monitor a continuous (measurement) characteristic when subgroup sizes are relatively large (typically n ≥ 10). It simultaneously tracks: - Xbar Chart – the stability of the process average. - S Chart – the stability of the process variability (standard deviation). Both charts must be in control before interpreting process performance or capability. --- When to Use an Xbar-S Chart Data and Sampling Requirements Use an Xbar-S Chart when: - The data are continuous (length, time, weight, temperature, etc.). - Data are collected in rational subgroups (short time spans where conditions are similar). - Subgroup size is typically 10 to 25 observations. - The process distribution is approximately normal within subgroups. - The goal is to monitor both mean and variation over time. In contrast: - For small subgroups (2 to 9), an Xbar-R Chart is usually preferred. - For individual measurements with no rational subgroups, an Individuals-Moving Range Chart is used instead. --- Structure of the Xbar-S Chart Components The Xbar-S Chart consists of two coordinated charts: - Top chart: Xbar - Plots subgroup means over time. - Detects shifts in the process center. - Bottom chart: S - Plots subgroup standard deviations over time. - Detects changes in process spread. Both charts share: - Center line (CL) – target or average value. - Upper control limit (UCL) – statistical upper boundary. - Lower control limit (LCL) – statistical lower boundary. Rational Subgrouping Rational subgrouping aims to: - Capture within-subgroup variation due to common causes. - Allow between-subgroup variation to reveal special causes over time. Typical approaches to forming subgroups: - Same operator, machine, and material within each subgroup. - Consecutive pieces taken close in time. - Subgroups taken at regular intervals (hourly, per batch, per shift). --- Computations for the S Chart Step 1: Collect and Organize Data Suppose there are k subgroups, each with n observations: - For subgroup i (i = 1, 2, …, k), data are: xᵢ₁, xᵢ₂, …, xᵢₙ. For each subgroup compute: - Subgroup mean: x̄ᵢ = (Σ xᵢⱼ) / n - Subgroup standard deviation: sᵢ = √[ Σ(xᵢⱼ − x̄ᵢ)² / (n − 1) ] Step 2: Average Standard Deviation Compute the average standard deviation across all subgroups: - S̄ = (Σ sᵢ) / k S̄ is the center line of the S Chart. Step 3: Control Limits for S Chart Control limits for the S Chart use constants B3 and B4, which depend on subgroup size n. - CLₛ = S̄ - UCLₛ = B4 × S̄ - LCLₛ = B3 × S̄ Where: - B4 > 1 and expands the upper limit. - B3 is often ≥ 0; for smaller n it may be 0, preventing negative limits. B3 and B4 values are taken from standard control chart constant tables for the given subgroup size. --- Computations for the Xbar Chart Step 1: Overall Average Compute the grand mean of all subgroup means: - x̄̄ = (Σ x̄ᵢ) / k x̄̄ is the center line of the Xbar Chart. Step 2: Control Limits for Xbar Chart For an Xbar-S Chart, the control limits for the means use constant A3: - CLₓ̄ = x̄̄ - UCLₓ̄ = x̄̄ + A3 × S̄ - LCLₓ̄ = x̄̄ − A3 × S̄ Where: - A3 depends on subgroup size n and is taken from standard control chart tables. The constant A3 incorporates: - The relationship between subgroup standard deviation and the standard error of the mean. - The desired control limit width (typically ±3 standard errors). --- Assumptions and Preconditions Statistical Assumptions The Xbar-S Chart assumes: - Independence of subgroups over time (no hidden autocorrelation). - Approximate normality of the within-subgroup distribution. - Stable measurement system (no major gauge issues). If these assumptions are violated,: - Control limits may be misleading. - Signals may be either missed or falsely triggered. Subgroup Size Considerations Subgroup size n affects: - The sensitivity of the Xbar Chart to small shifts. - The reliability of standard deviation estimates. - Values of B3, B4, and A3. Guidelines: - Keep n consistent across subgroups whenever possible. - If n varies, specialized formulas and software are needed to adjust constants and limits. --- Constructing the Xbar-S Chart Step by Step Step 1: Plan the Chart Define: - Characteristic to be measured (e.g., diameter, cycle time). - Subgroup strategy (e.g., 5 parts every 10 minutes, 15 parts per batch). - Subgroup size n (usually ≥ 10 for Xbar-S). - Number of subgroups k for baseline (commonly 20–25 or more). Step 2: Collect Baseline Data For each subgroup: - Measure n units. - Calculate x̄ᵢ and sᵢ. - Record time or sequence number. Use only data from a period when the process is believed to be relatively stable (no known major upsets). Step 3: Compute S Chart Parameters - Compute S̄. - Select B3 and B4 for the chosen n. - Calculate: - CLₛ = S̄ - UCLₛ = B4 × S̄ - LCLₛ = B3 × S̄ Step 4: Compute Xbar Chart Parameters - Compute x̄̄. - Select A3 for subgroup size n. - Calculate: - CLₓ̄ = x̄̄ - UCLₓ̄ = x̄̄ + A3 × S̄ - LCLₓ̄ = x̄̄ − A3 × S̄ Step 5: Plot the Charts - On the S Chart: - Horizontal axis: subgroup sequence (time). - Vertical axis: sᵢ values. - Draw CLₛ, UCLₛ, LCLₛ. - On the Xbar Chart: - Horizontal axis: same subgroup sequence. - Vertical axis: x̄ᵢ values. - Draw CLₓ̄, UCLₓ̄, LCLₓ̄. Ensure the two charts share the same horizontal scale for correct visual interpretation. --- Interpreting the S Chart Primary Purpose The S Chart is interpreted first because: - It shows whether process variation is stable. - Unstable variation invalidates many conclusions about the mean or capability. Basic Interpretation Rules Look for: - Points beyond UCLₛ - Indicates unusually high variability. - Possible causes: worn tools, inconsistent materials, operator changes, environment changes. - Points below LCLₛ - Indicates unusually low variability. - May signal: - A positive improvement. - A change in measurement procedure. - Data handling errors. - Trends and patterns in sᵢ - Upward trend: variability is increasing over time. - Downward trend: variability is decreasing (check if change is intentional). - Cycles: periodic changes in variation (e.g., by shift, batch, or time of day). If the S Chart is out of control, do not interpret the Xbar Chart as representing a stable mean. --- Interpreting the Xbar Chart Prerequisite Interpret the Xbar Chart only after the S Chart appears stable. Stable variation is required to correctly interpret mean behavior. Basic Interpretation Rules Check for: - Points beyond UCLₓ̄ or LCLₓ̄ - Large special-cause shifts in the process mean. - Examples: parameter changes, new materials, major setup changes. - Runs and trends - Long run of points on one side of CLₓ̄ suggests a shift in the mean. - Consistent upward or downward trend suggests drift (e.g., tool wear, gradual temperature change). - Patterns around limits - Many points close to UCLₓ̄ or LCLₓ̄ without crossing may signal increased risk of going out of control. - Sudden step changes (new level of Xbar) with continued control on S may indicate an intentional setpoint change. --- Common Out-of-Control Patterns On the S Chart Typical special-cause patterns include: - Single high point above UCLₛ - One-time spike in variability, often linked to a specific event. - Several consecutive high sᵢ values (but within limits) - Possibly emerging problem in equipment, environment, or raw materials. - Alternating high–low sᵢ values - Different conditions across subgroups (e.g., alternating operators, machines, or shifts). On the Xbar Chart Typical patterns include: - Single point beyond UCLₓ̄ or LCLₓ̄ - Sudden special cause in the mean. - Shift in center - Sequence of consecutive points above (or below) the center line. - Suggests sustained change in process setting or environment. - Trend - Consistently increasing or decreasing means. - Indicates drift rather than sudden shift. Standard run rules (e.g., about runs, trends, and zone tests) can be applied, provided the user respects the multiple-testing trade-off and context of the process. --- Xbar-S Chart vs. Xbar-R Chart When S Is Preferred An Xbar-S Chart is preferred to an Xbar-R Chart when: - Subgroup size n ≥ 10. - A more accurate estimate of standard deviation is desired. - The range may be less robust or less informative for larger n. Conceptual Connection For a given subgroup: - R (range) = max − min. - S (standard deviation) incorporates all data points and is a more comprehensive measure of spread. As n increases: - R becomes less efficient and less consistent. - S becomes a better estimator of the underlying process standard deviation. --- Using the Xbar-S Chart for Process Decisions Establishing Control Once an initial Xbar-S Chart is built: - Confirm both Xbar and S charts show no special-cause signals. - If signals exist, investigate and remove special causes. - Recompute control limits using only data from stable periods. Ongoing Monitoring After control is established: - Continue plotting new subgroups in time order. - Respond to new out-of-control signals by: - Identifying specific events associated with the affected subgroups. - Determining whether they are: - Bad special causes (to be removed). - Good special causes (intentional improvements that may justify new limits). Link to Capability (Conceptual) Once a process is stable on the Xbar-S Chart: - Use estimates based on S̄ (and possibly pooled sᵢ) to represent process standard deviation. - Use x̄̄ as the estimate of the process mean. - These estimates support process capability analyses, provided the process remains in control. --- Practical Tips and Common Pitfalls Practical Tips - Keep subgroup size consistent to use standard control chart constants directly. - Label subgroups by time (date, shift, batch) to speed root cause analysis. - Verify measurement system before acting on chart signals. Common Pitfalls - Ignoring the S Chart - Interpreting Xbar while S is unstable misleads process decisions. - Combining different conditions in one subgroup - Mixing machines, materials, or settings within a subgroup weakens rational subgrouping. - Overreacting to random variation - Treating every small Xbar movement as a signal leads to unnecessary adjustments (tampering). - Insufficient baseline data - Very few subgroups produce unreliable control limits; aim for an adequate baseline. --- Summary The Xbar-S Chart is a two-part control chart designed for continuous data with relatively large subgroup sizes. It monitors: - S Chart – stability of process variation via subgroup standard deviations. - Xbar Chart – stability of process location via subgroup means. Key points: - Build rational subgroups and calculate x̄ᵢ and sᵢ for each. - Compute S̄ and x̄̄, then apply constants B3, B4, and A3 to set control limits. - Interpret the S Chart first; only then interpret the Xbar Chart. - Use patterns and out-of-control signals to distinguish common-cause from special-cause variation. - Once stable, the Xbar-S Chart supports sound process monitoring and provides a basis for capability assessment.