2.4.2 Concept of Stability

Concept of Stability Understanding Process Stability Process stability describes whether a process behaves consistently over time, within predictable limits, when only random (common cause) variation is present. A stable process is said to be in statistical control. - Stable process: only common cause variation, no pattern of unusual change. - Unstable process: special cause variation present; behavior is unpredictable. Stability is evaluated using data over time, typically with control charts. It is not about whether results meet specifications; it is about whether the process behaves predictably. A process must be stable before capability (how well it meets specifications) can be interpreted correctly. Without stability, capability indices are unreliable and misleading. --- Common Cause vs Special Cause Variation Common Cause Variation Common cause variation is the natural, inherent fluctuation in a process. - Source: routine factors built into the process (machines, materials, methods, environment). - Behavior: small, random, and consistent over time. - Control chart pattern: points randomly scattered within control limits, showing no non-random patterns. Changing a stable process affected only by common causes requires changes to the process system (e.g., redesign, new standards, new technology), not local tweaks. Special Cause Variation Special cause variation arises from specific, identifiable, and often intermittent factors that are not part of normal process behavior. - Source: unusual events (equipment failure, wrong material, misadjustment, data entry error). - Behavior: sudden shifts, spikes, trends, or outliers. - Control chart pattern: points outside control limits or non-random patterns within limits. When special causes are present, the process is unstable. The focus is on: - Detecting the special cause quickly. - Investigating the root cause. - Removing or controlling the special cause. - Verifying that the process returns to a stable pattern. --- Statistical Control and Stability Definition of Statistical Control A process is in statistical control when: - Only common cause variation is present. - The distribution of data over time is predictable. - Control chart rules indicate no special causes. Being in control does not guarantee good performance; it guarantees consistent performance. A process can be: - Stable but incapable (predictably produces many defects). - Unstable and sometimes accidentally capable (good results at times, bad at others). The goal is first to achieve stability, then to improve performance as needed. Time Order and Stability Stability is inherently a time-based property. Treating data as a simple unordered list (e.g., histogram only) hides instability. To assess stability: - Maintain the time order of data. - Use control charts or run charts. - Look for changes, trends, and patterns across time. Any interpretation of variation that ignores time order risks missing instability. --- Control Limits vs Specification Limits Control Limits Control limits are statistical boundaries on a control chart that represent the natural behavior of the process. - Calculated from process data, not from customer requirements. - Typically set at ±3 standard deviations from the process mean. - Reflect what the process is producing when it is stable. If a process is stable, almost all points should lie within the control limits, and patterns should be random. Specification Limits Specification limits represent what should be produced. - Defined by customers, contracts, or design. - Indicate the acceptable range of output (e.g., USL and LSL). - Do not describe process behavior; they describe requirements. A common error is to confuse control limits with specification limits. Key distinctions: - Control limits: about stability and predictability (voice of the process). - Specification limits: about acceptability and requirements (voice of the customer). Stability is judged using control limits, not specification limits. --- Types of Control Charts and Stability Purpose of Control Charts for Stability Control charts are tools to distinguish between common and special cause variation and to judge process stability. They: - Display data in time order. - Provide a central line (mean, median, or proportion). - Include upper and lower control limits based on the data. - Reveal non-random patterns that signal instability. Choosing a Control Chart (Conceptual) The type of control chart depends on: - Data type: - Variable data (continuous, measured). - Attribute data (counted, discrete). - Sample structure: - Individual measurements. - Subgroups of multiple observations. Common conceptual categories: - Variable charts (for measurements like time, weight, temperature): - Charts for individual values and moving ranges. - Charts for subgroup means and ranges or standard deviations. - Attribute charts (for counts like defects, defectives, events): - Charts for proportions or counts of nonconforming units. - Charts for number of defects. The detailed formulas are not required to understand stability; what matters is how to interpret the patterns they show. --- Identifying Special Causes: Control Chart Rules Basic Out-of-Control Signals A control chart signals instability when patterns exceed what would be expected by chance. Important signals include: - Point(s) beyond control limits: - Any point above the upper control limit. - Any point below the lower control limit. - Runs of points on one side of the center line: - A sequence of consecutive points all above or all below the center line, longer than expected by chance. - Trends: - Several points in a row steadily increasing or steadily decreasing. - Cycles or systematic patterns: - Repeating waves or periodic patterns. - Hugging the center line or the limits: - Too little variation or too much variation compared to what is expected. These patterns indicate special causes and therefore instability. Interpreting Patterns When an out-of-control signal appears: - Treat it as evidence of a special cause. - Investigate what was different in the process at that time. - Verify whether the cause can be eliminated or controlled. - Remove the affected portion of data if it clearly reflects a special cause that will not be allowed to recur, then recalculate control limits. The goal is to return the process to a state where only common cause variation remains. --- Stability, Capability, and Improvement Sequence: Stability Before Capability Capability indices (such as measures of how well a process fits within specification limits) assume a stable process. If the process is unstable: - The distribution of outcomes changes over time. - Past performance is not a reliable predictor of future performance. - Any capability calculation has limited meaning. Correct sequence: - First confirm stability using appropriate charts and rules. - Then estimate capability using data from the stable period. - Re-evaluate stability after any major change in the process. Stable but Not Acceptable A process can be stable yet perform poorly against specifications. In that case: - Performance is consistently off-target or too variable. - Improvement requires changing the underlying process: - Adjusting process settings. - Redesigning steps or methods. - Reducing common cause variation through systematic changes. Stability ensures that improvements can be evaluated reliably because the process behaves predictably. --- Process Stabilization Strategies Responding to Special Causes When instability is detected: - Short-term actions: - Contain immediate issues (segregate suspect output, stop obvious errors). - Identify and remove the special cause if possible. - Long-term actions: - Implement changes to prevent recurrence. - Document standard responses when similar signals appear. - Train process participants on recognizing signals. Removing special causes moves the process toward stability. Maintaining Stability Once stable, a process must be monitored to remain stable. Key practices: - Use control charts routinely for key characteristics. - Investigate new signals promptly. - Avoid unnecessary tampering (reacting to normal variation as if it were special). - Establish clear procedures for: - When to adjust the process. - When not to adjust the process. Over-adjustment based on random variation can introduce new instability. --- Interpreting Stable and Unstable Patterns Characteristics of a Stable Process A stable pattern shows: - Points randomly dispersed around the center line. - No points beyond control limits. - No long runs on one side of the center line. - No sustained upward or downward trends. - No recurring cycles. Such a process is predictable within its current level of variation. Characteristics of an Unstable Process An unstable pattern may show: - One or more points beyond control limits. - Several consecutive points all above or all below the center line. - Several consecutive points increasing or decreasing. - Periodic ups and downs at a regular interval. - Sudden shifts in the center of the data. Each of these signals implies that some special cause is acting on the process. --- Practical Implications of Stability Decision-Making Understanding whether a process is stable affects decisions such as: - Predicting future performance. - Setting realistic targets. - Choosing improvement strategies. - Evaluating the impact of changes. Decisions based on unstable data are risky because the underlying behavior is changing. Data Collection and Sampling Accurate assessment of stability requires: - Data collected in correct time sequence. - Sampling plans that reflect how the process actually runs. - Enough data points to reveal patterns over time. Poor or inconsistent data collection can hide instability or falsely suggest instability where none exists. --- Summary A process is stable when it is in statistical control, displaying only common cause variation and no evidence of special causes. Stability is determined by examining data over time, most effectively with control charts that use control limits derived from the process itself. Control limits describe the natural behavior of the process, while specification limits describe customer requirements; stability is always judged against control limits, not specifications. Special cause signals on control charts indicate instability and must be investigated, understood, and addressed before reliable predictions or capability assessments can be made. Stability provides a foundation for meaningful capability analysis, trustworthy decision-making, and effective improvement efforts. When a process is both stable and aligned with specifications, it can be considered both predictable and acceptable; when it is stable but not capable, systematic changes are required to raise performance.