Summary
Highlights
Roger Wolf from SafetyChain introduces Statistical Process Control (SPC) as a method for quality control in the food industry. SafetyChain provides a plant management platform for food producers, where SPC fits into the processing stage. The main benefits of SPC include efficient plant operation, reduced waste and scrap, ensuring product safety, cost reduction, demonstrating quality, and satisfying regulatory requirements like FDA and USDA audits.
Manufacturing processes naturally involve variations. SPC helps determine if these random variations will meet specifications. SPC is based on the statistics of randomization, often visualized by the bell curve (normal distribution). This curve illustrates how most data points cluster around the mean, with fewer points towards the tails. Standard deviations (Sigma) are used to define these regions, with specific percentages of data falling within each sigma level.
The effectiveness of a process is determined by how well its bell curve fits within the lower and upper specification limits. A good process has a bell curve that fits entirely between these limits, indicating high compliance with specifications. If the bell curve drifts outside the limits, it signifies excessive waste, non-compliant products, or the need for downgrading. SPC aims to identify and correct such situations to improve margins.
The PPK (Process Performance Index) is a key metric that measures how well the bell curve fits within specification limits. It's calculated based on the mean value, standard deviation, and specification limits. A PPK below 1 indicates that the process does not consistently meet specifications, leading to significant yield loss. A PPK above 1 is desirable, and a target of 1.33 for food manufacturing means 99.99% of products conform to specifications. PPK allows for reliable predictions about meeting specifications from small sample sizes.
In food operations, testing everything is often impractical. SPC relies on periodically pulling representative sample sets. The number of samples depends on factors like cost, time, and product volume. Data from these samples are collected and plotted on two main charts: the X-bar chart (mean) for tracking averages and the range chart (or Sigma chart) for tracking variation within sample sets. The range chart is used for sample sets between 2 and 9, while the Sigma chart is used for 10 or more samples to calculate the true standard deviation.
Estimated standard deviation is used for sample sets under ten, calculated by dividing the range by a d2 lookup value. This is then used to calculate control limits for both the mean and variation charts. Control limits define the expected range of values for a stable process based on its historical performance. They are distinct from specification limits, which are defined by product requirements. Control limits can vary between different lines or shifts due to factors like equipment, temperature, or operators, reflecting the historical track record of the process rather than desired product specs.
When a process stays within control limits, it is considered stable. However, processes can drift over time due to factors such as machine warm-up, worn parts, or natural variability in raw materials (especially in the food industry). The Westinghouse rules provide early warning indicators to operators when a process deviates from expected statistical behavior, signaling that action may be needed to maintain stability. These rules apply to both the mean and variation charts.
Key run rules include: a single point outside the control limit (Rule 1), two out of three points in the second Sigma zone (Rule 2), four out of five points in the first Sigma zone (Rule 3), eight consecutive points above or below the mean (Rule 4), and thirteen consecutive points within the middle Sigma band (Rule 9). The last rule indicates too much consistency, which can suggest incorrect data collection or a lack of expected natural variation. These rules are visually flagged on charts, prompting operators to investigate and take corrective action.
Beyond tracking current performance, SPC uses the Process Capability Index (CPK), which differs from PPK. While PPK assesses past performance against specifications, CPK is forward-looking, asking: "Am I capable of meeting my specification?" It uses the estimated standard deviation to predict whether a process, given its inherent variations, can consistently meet specifications with an acceptable level of waste or rework. This is crucial before committing to large-scale production.
In summary, CPK determines process capability, X-bar and Range charts track progress, and run rules alert operators to potential issues, facilitating timely interventions. While this video covered the fundamentals, other SPC chart types exist for different scenarios, such as single-sample data or attribute data (counts of defects). SPC tools automate calculations, allowing operators to focus on inputs and actions rather than complex math. For more information, visit safetychain.com.