Creating a Golden Profile with Seeq

Introduction

This tutorial will guide you through the process of creating golden profiles, also known as reference profiles or golden batches, using Seeq's analytics approach. This method is essential for any recurring process where monitoring signal trajectories is critical for performance. By following these steps, you can develop a statistical framework to identify optimal operations and detect anomalies in real time.

Chapter 1: Identifying Operational Time Periods

1. Select Critical Signals

   • Identify the key signal you want to monitor. In this example, we will monitor a power signal during batch operations.
   • Add the mode signal to track production status.

2. Value Search for Production Runs

   • Use the mode signal to identify all production runs by filtering for instances where the mode indicates production.
   • If a production status signal isn't available, consider using alternative signals or advanced run logic.

3. Expand Time Range

   • Adjust the time range for analysis, such as extending it to 20 days to capture sufficient data.

4. Visualize Data

   • Utilize different views in Seeq (like chain view or capsule view) to analyze the identified run capsules.

Chapter 2: Selecting Best Historical Operation Periods

1. Identify Best Runs

   • Use capsule view to visually assess the best production runs based on historical performance.
   • Utilize the manual condition tool to mark runs that appear normal and expected.

2. Quantitative Selection Method

   • To enhance your selection, add the average power statistic for each production run in the capsules pane.
   • Sort by average power and select runs above a certain threshold (e.g., 10,700 KW).
   • Alternatively, define best runs based on duration, e.g., between 13.5 and 16 hours.

3. Finalizing Best Runs

   • After selection, confirm the identified best runs, which will be essential for developing statistical profiles.

Chapter 3: Creating Statistical Profiles

1. Develop Statistical Boundaries

   • With the best runs selected, launch the reference profile tool.
   • Model the power signal using only data from these selected runs:
     • Set a calculation granularity (e.g., 2 minutes).
     • Define the upper limit as the average power plus three standard deviations.

2. Generate Lower Limit Signal

   • Duplicate the upper limit signal and adjust it to calculate the lower limit by changing the multiplier to minus three standard deviations.

3. Visualize Combined Signals

   • Overlay the upper and lower limit signals on a single lane to observe how they represent the best run values.

4. Identify Abnormal Runs

   • Utilize the statistical limits to highlight runs that fall significantly outside the established boundaries.

Chapter 4: Monitoring Profiles

1. Create Outside Limits Condition

   • Use value search to establish a condition for operations that fall outside the calculated limits.
   • Set parameters to ignore deviations lasting less than 60 minutes.

2. Monitor Abnormal Operations

   • Implement various monitoring techniques:
     • Near real-time comparisons of current runs to historical runs in chain view.
     • Utilize scorecard metrics to summarize outside limits events in a condition table.

3. Analyze Performance Over Time

   • Track the number and duration of excursions outside limits to assess performance improvements or degradations.

Conclusion

In this tutorial, you've learned how to build golden profiles using a systematic approach that involves identifying operational periods, selecting the best runs, creating statistical profiles, and monitoring for anomalies. By applying these techniques, you can enhance process performance monitoring across a variety of operations. As a next step, consider setting up email alerts for outside limit conditions to facilitate proactive management of process performance.