Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Risk-Calibrated Process Capability Approval with Finite Samples

Fei Jiang, Lei Yang

Abstract

Process capability indices such as $C_{pk}$ are widely used in manufacturing to support supplier qualification, pilot-build release, and production approval. In practice, approval decisions are often based on deterministic threshold rules of the form $\widehat{C}_{pk} \ge C_0$. Because $\widehat{C}_{pk}$ is estimated from finite samples, however, such decisions are inherently stochastic, especially when the true capability lies near the approval threshold. This paper develops a risk-calibrated decision framework for process capability approval that explicitly accounts for estimation uncertainty and asymmetric operational loss. Capability approval is formulated as a binary statistical decision problem, leading to a rule of the form $\widehat{C}_{pk} \ge C_0 + k\,SE(\widehat{C}_{pk})$, where the calibration constant $k$ is determined either by a tolerable failure probability or by a false-accept/false-reject cost ratio. The resulting formulation unifies several commonly used procedures, including deterministic thresholding, lower confidence bound rules, and probability-based approval rules, and naturally extends them to cost-sensitive decision rules derived from asymmetric operational loss. Simulation experiments and an industrial case study show that risk calibration primarily affects near-threshold decisions, improves approval stability, and can substantially reduce expected operational loss when false acceptance is more costly than false rejection.

Risk-Calibrated Process Capability Approval with Finite Samples

Abstract

Process capability indices such as are widely used in manufacturing to support supplier qualification, pilot-build release, and production approval. In practice, approval decisions are often based on deterministic threshold rules of the form . Because is estimated from finite samples, however, such decisions are inherently stochastic, especially when the true capability lies near the approval threshold. This paper develops a risk-calibrated decision framework for process capability approval that explicitly accounts for estimation uncertainty and asymmetric operational loss. Capability approval is formulated as a binary statistical decision problem, leading to a rule of the form , where the calibration constant is determined either by a tolerable failure probability or by a false-accept/false-reject cost ratio. The resulting formulation unifies several commonly used procedures, including deterministic thresholding, lower confidence bound rules, and probability-based approval rules, and naturally extends them to cost-sensitive decision rules derived from asymmetric operational loss. Simulation experiments and an industrial case study show that risk calibration primarily affects near-threshold decisions, improves approval stability, and can substantially reduce expected operational loss when false acceptance is more costly than false rejection.
Paper Structure (40 sections, 2 theorems, 78 equations, 4 figures, 4 tables)

This paper contains 40 sections, 2 theorems, 78 equations, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Capability Estimation and Boundary Instability
  3. Classical capability estimator under normal sampling
  4. Local asymptotic approximation
  5. Boundary-instability phenomenon
  6. Decision-Theoretic Formulation of Capability Approval
  7. Decision space and states of nature
  8. Asymmetric loss structure
  9. Posterior expected loss
  10. Misclassification probabilities and expected operational loss
  11. Unified Risk-Calibrated Approval Rules
  12. Asymptotic posterior approximation
  13. Sample-size-dependent approval margin
  14. A unified margin family
  15. Local asymptotic interpretation of the calibration constant
  16. ...and 25 more sections

Key Result

Theorem 1

Suppose the capability estimator $\widehat{C}_{pk}$ satisfies the asymptotic normal approximation in eq:cpk_asymptotic_normal, and the failure probability is approximated by eq:posterior_failure_prob. Then the probability-based approval rule eq:probability_rule is equivalent to the following margin Moreover, under asymmetric loss ratio $\lambda=c_{FA}/c_{FR}$, the Bayes-optimal decision rule is o

Figures (4)

  • Figure 1: Acceptance probability surface under deterministic threshold approval (background heatmap). The deterministic threshold rule accepts when $\widehat{C}_{pk} \ge C_0$. The color scale represents the probability of approval under repeated sampling. Overlaid contours show the effective approval boundaries defined by $P(\mathrm{Accept}) = 0.5$ for three rules: the deterministic threshold rule (solid), the $95\%$ lower confidence bound rule (dashed), and the cost-sensitive rule with $\lambda = 9$ (dotted). Results are based on Monte Carlo simulation under normal sampling with symmetric specification limits and centered processes.
  • Figure 2: Error probabilities under the cost-sensitive approval rule as the cost ratio $\lambda = c_{FA}/c_{FR}$ varies. The horizontal axis is shown on a logarithmic scale. Results are based on Monte Carlo simulation under normal sampling with threshold $C_0 = 1.33$, sample size $n = 32$, and replication size $B = 10{,}000$. (a) False-accept probability for selected sub-threshold capability levels ($C_{pk}^{true} < C_0$). (b) False-reject probability for selected supra-threshold capability levels ($C_{pk}^{true} \ge C_0$). As $\lambda$ increases, the acceptance region contracts, reducing false acceptance at the cost of more frequent false rejection.
  • Figure 3: Expected operational loss under deterministic threshold and risk-calibrated approval rules. The deterministic threshold rule accepts when $\widehat{C}_{pk} \ge C_0$, whereas the risk-calibrated rule corresponds to the probability rule with $\alpha = 0.05$. Under this calibration, the probability rule, the $95\%$ lower confidence bound rule, and the cost-sensitive rule with $\lambda = 19$ are equivalent. Results are based on Monte Carlo simulation with sample size $n = 32$ and $B = 12{,}000$ replications under normal sampling. The vertical dashed line indicates the approval threshold $C_0 = 1.33$.
  • Figure 4: Empirical characteristics of capability decisions in the industrial dataset. Panel (a) shows the distribution of estimated capability indices across all dimensions, separately for approximately normal and non-normal subsets. The vertical line indicates the approval threshold $C_0=1.33$. Panel (b) compares the aggregate empirical expected loss of deterministic threshold and risk-calibrated approval rules across cost ratios $\lambda$. The results illustrate how risk-calibrated decision rules can substantially reduce expected operational loss when false acceptance carries higher cost.

Theorems & Definitions (3)

  • Theorem 1: Unified Margin Representation
  • proof
  • Theorem 2: Boundary Risk Calibration