5.2.2 I-MR Chart

I-MR Chart Introduction to I-MR Chart An I-MR Chart (Individuals and Moving Range Chart) is a pair of control charts used for monitoring a process when data are collected one observation at a time and subgroups of size greater than 1 are not available or not appropriate. - I Chart monitors individual values over time. - MR Chart monitors the variability between consecutive observations. I-MR Charts are applied to continuous (variable) data, such as time, length, weight, temperature, or pressure, and are especially useful when production volume is low or measurements are naturally individual. --- When to Use an I-MR Chart Appropriate Data and Conditions Use an I-MR Chart when: - Data type is continuous (measurement scale). - Subgroup size is 1 (only one observation per time point). - Frequency is sequential over time or sequence (time-ordered data). - Process is expected to be stable enough that each observation estimates the same underlying distribution. Do not use an I-MR Chart when: - Data are attribute (counts, defects) rather than measurements. - Rational subgroups of larger size (n ≥ 2) can be formed; in that case, X-bar/R or X-bar/S charts are usually preferred. - Data are severely autocorrelated beyond simple short-term dependence. Typical Applications Common applications include: - Monitoring cycle time or lead time measured case by case. - Tracking a critical dimension when only one part is measured at each interval. - Monitoring lab results when tests are run one at a time. - Observing service performance data (handling time, wait time) when batched subgroups are not practical. --- Structure of the I-MR Chart I Chart Components The I Chart displays each individual observation over time with control limits based on overall process variability estimated from the moving ranges. Key elements: - Center line (CL): overall mean of individual observations. - Upper control limit (UCL): estimated upper natural process boundary. - Lower control limit (LCL): estimated lower natural process boundary. - Data points: individual values plotted in time order. The I Chart answers the question: “Is the process location (mean) stable over time?” MR Chart Components The MR Chart displays the absolute differences between consecutive observations, called moving ranges. For subgroup size 1, the moving range is typically calculated using pairs of consecutive data points (order 2 moving range). Key elements: - Moving range (MR): MRᵢ = |Xᵢ − Xᵢ₋₁|. - Average moving range (MR̄): mean of all MR values. - Center line (CL): MR̄. - Upper control limit (UCL): estimated upper natural boundary for moving range. - Lower control limit (LCL): usually 0 for MR charts, because a range cannot be negative. The MR Chart answers the question: “Is the process short-term variability stable over time?” --- Calculations for I-MR Chart Notation Let: - Xᵢ = the i-th individual observation, i = 1, 2, …, n. - MRᵢ = moving range for pair (Xᵢ₋₁, Xᵢ), defined for i = 2, …, n. Step 1: Compute the Individual Mean Calculate the mean of individual observations: - X̄ = (Σ Xᵢ) / n This value is the center line for the I Chart. Step 2: Compute the Moving Ranges For each consecutive pair: - MRᵢ = |Xᵢ − Xᵢ₋₁|, for i = 2, …, n There will be (n − 1) MR values. Step 3: Average Moving Range Compute the average moving range: - MR̄ = (Σ MRᵢ) / (n − 1) This is the center line for the MR Chart. Step 4: Estimate Process Standard Deviation For moving ranges of size 2, the I-MR Chart uses the constant d₂ ≈ 1.128 to estimate the process standard deviation. - σ̂ = MR̄ / d₂, where d₂ = 1.128 for range based on 2 observations. This σ̂ is used to construct the control limits for the I Chart. Step 5: Control Limits for the I Chart Using σ̂: - UCLᵢ = X̄ + 3σ̂ - LCLᵢ = X̄ − 3σ̂ If the calculated LCL is below the physically possible minimum for the process (for example, negative when only positive values are meaningful), interpret accordingly, but keep the formula as is. Step 6: Control Limits for the MR Chart For moving ranges of size 2, the UCL uses the constant D₄, and the LCL uses D₃. For n = 2: - D₄ = 3.267 - D₃ = 0 Thus: - CLᴹᴿ = MR̄ - UCLᴹᴿ = D₄ × MR̄ = 3.267 × MR̄ - LCLᴹᴿ = D₃ × MR̄ = 0 Some software omits the LCL line, but conceptually it is 0. --- Assumptions and Data Considerations Statistical Assumptions I-MR Charts rely on several assumptions: - Independence: consecutive observations are independent apart from short-range dependence captured by MR. - Stationarity: process mean and variance are constant when the process is in control. - Approximate normality: individual measurements are approximately normally distributed, particularly important for accurate interpretation of control limits and rules. For moderate sample sizes and practical use, I-MR Charts are usually robust to mild deviations from normality, but extreme skewness or heavy tails can reduce effectiveness. Rational Subgrouping with n = 1 Rational subgrouping means that each data point represents a reasonable snapshot of the process at a point in time. For n = 1: - Each single observation is treated as a subgroup. - Time order is essential; do not sort data by size. - Data collection should reflect natural production or service sequence. --- Interpretation of the I Chart Basic In-Control Pattern When the process is in control: - Most points fall within the control limits. - Points are randomly scattered around the center line. - No systematic trends, cycles, or non-random patterns appear. Special Cause Indicators on I Chart Key signals of special causes include: - Point beyond control limits - Any point above UCL or below LCL. - Run of points on one side of center line - A long sequence of consecutive points all above or all below X̄ suggests a shift in the process mean. - Common practical rule: 8 or more consecutive points on one side of the center line. - Trend of steadily increasing or decreasing values - Several consecutive points consistently increasing or decreasing suggests a drift. - Common practical rule: 6 or more points in a continuous upward or downward trend. - Cycles or systematic patterns - Regular, repeating up-down patterns indicate systematic variation (e.g., shifts between day and night operation). - Hugging center line - Many points extremely close to the center line may signal overcontrol or data smoothing (e.g., incorrect rounding, tampering). - Hugging the control limits - Clustering near UCL or LCL suggests that the process is being pushed to extremes or that the limits are underestimated. These rules operationalize “non-random” behavior, signaling that the process may not be stable. --- Interpretation of the MR Chart Function of the MR Chart The MR Chart focuses on short-term variation between consecutive observations. It helps determine whether the estimated σ̂ is valid and whether the process variability is consistent over time. In-Control Pattern on MR Chart When variability is stable: - Most moving ranges fall within MR UCL. - MR points are randomly scattered around MR̄. - There is no systematic pattern (trends, cycles) in the MR Chart. Special Cause Indicators on MR Chart Important signals include: - Point beyond MR UCL - Indicates an unusually large change between two consecutive observations. - May suggest a sudden disturbance, measurement error, or process change. - Runs or trends in MR values - Several consecutive high or low MR points indicate changing variability. - For example, sustained higher MR values may signal increased noise or instability. - Zero or near-zero MR values - Repeated very small MR suggests possible data rounding, duplicate readings, or lack of actual changes between observations. If the MR Chart is out of control, the estimate of σ̂ (and therefore I Chart limits) may not be reliable. --- Diagnosing and Responding to Signals Priority of MR Chart vs I Chart A disciplined interpretation sequence often uses: - First, check MR Chart: - If MR is out of control, address variability issues first. - Fixing instability in variance improves the reliability of I Chart interpretation. - Then, check I Chart: - Examine shifts, trends, and out-of-limit points in individual values. Investigating Special Causes When a special cause signal appears: - Identify time and context: - When did the signal occur? - What was happening in the process at that time? - Check for data errors: - Incorrect recording, calibration issues, or transcription mistakes. - Examine operational changes: - Different operator, material batch, equipment setup, environment, or method. - Segment if needed: - If a clear, permanent process change occurred, analyze data before and after separately. The purpose of investigation is to distinguish controllable causes from inherent variability and to understand whether process changes are beneficial or harmful. --- Capability and Performance Using I-MR Data Although I-MR Charts primarily monitor stability, they also support capability and performance assessment once the process is stable. Precondition: Stable Process Only analyze capability after: - MR Chart indicates stable variability. - I Chart shows no special causes or non-random patterns within the in-control region. Estimating Sigma from I-MR The same σ̂ used for control limits can be used for capability metrics: - σ̂ = MR̄ / d₂ This sigma estimate reflects short-term within-process variation based on moving ranges. Relation to Specification Limits Once stability is confirmed: - Compare process spread (±3σ̂ or ±6σ̂) with customer specification limits. - Assess how much of the natural variation fits within the specifications. While formal capability indices are beyond the focus here, the I-MR Chart supports evaluating whether a stable process is also capable. --- Common Pitfalls and Good Practices Common Pitfalls Avoid these issues when using I-MR Charts: - Ignoring time order: - Sorting data before plotting destroys the control chart concept. - Mixing different conditions without distinction: - Combining data from different products, setups, or environmental conditions into one chart can mask special causes. - Too few data points: - Very small sample sizes result in unstable estimates of MR̄ and σ̂. - Practical guidance: aim for at least 20–25 points to establish baseline limits. - Using I-MR for attribute data: - Counts and proportions require different chart types. - Overreacting to single points near limits: - Control limits represent natural variation; points near the limit are not automatically special causes. Good Practices Adopt these practices for effective use: - Maintain strict time sequence for all observations. - Record context information (shift, operator, material lot) alongside each measurement to aid root cause analysis. - Regularly review MR Chart to confirm valid variability estimation. - Recalculate limits only after a documented, intentional, and sustained process change, not after every small adjustment. - Use consistent measurement methods to avoid artificial shifts or changes in variation. --- Practical Steps to Construct an I-MR Chart Data Preparation - Collect individual observations in time order. - Ensure consistent units and measurement methods. - Check for obvious data entry errors. Chart Construction Steps - Calculate X̄ from the individual data. - Compute moving ranges MRᵢ for consecutive points. - Calculate MR̄ from MRᵢ values. - Estimate σ̂ = MR̄ / 1.128. - Determine I Chart limits: - CL = X̄ - UCLᵢ = X̄ + 3σ̂ - LCLᵢ = X̄ − 3σ̂ - Determine MR Chart limits: - CLᴹᴿ = MR̄ - UCLᴹᴿ = 3.267 × MR̄ - LCLᴹᴿ = 0 - Plot: - I Chart: Xᵢ against time, with CL, UCLᵢ, LCLᵢ. - MR Chart: MRᵢ against time, with CLᴹᴿ, UCLᴹᴿ, LCLᴹᴿ. Interpretation Cycle - Examine MR Chart: - Identify out-of-control points or patterns in variability. - Examine I Chart: - Identify any special cause signals in the level (mean). - Investigate causes, make appropriate process changes. - Once stable, optionally assess capability using σ̂. --- Summary An I-MR Chart is a paired control chart for monitoring individual measurements and their short-term variability when only one observation per time point is available. - The I Chart tracks process location (mean) using individual values and control limits based on an estimated standard deviation from moving ranges. - The MR Chart tracks process variability using the absolute differences between consecutive observations. Effective use requires: - Continuous, time-ordered data with subgroup size 1. - Correct calculation of X̄, moving ranges MRᵢ, MR̄, σ̂, and corresponding control limits. - Careful interpretation of special cause signals in both I and MR charts. - Attention to assumptions of independence, approximate normality, and rational subgrouping. When applied correctly, the I-MR Chart provides a powerful method to detect special causes, stabilize processes, and support capability analysis using individual observation data.