Data Analytics in Batch Operations

When Batch Processing Is Critical to Your Operations, Imagine the Worth of Knowing the Predicted Value of Quality Parameters

3 of 4 1 | 2 | 3 | 4 View on one page

Analytic Tool Selection

The successful application of multivariate statistical techniques such as PCA and PLS depends in part on the toolset selected. Various techniques for PCA and PLS model development are used in commercial products. In some cases, a product may be designed to support the analysis of continuous processes. In such cases, data analysis and model development often assume that a process is maintained at just one operating condition.

To address the requirements of batch processes successfully, multivariate tools used for model development should be designed to address varying process conditions over a wide range of operation. Multiway PCA and PLS algorithms are commonly used in multivariate tools to address batch applications. Tools that support these algorithms are designed to allow a normal batch trajectory to be established for each process input and measurement automatically. In many cases, the toolset may only support off-line, post-production analysis. There is potentially much greater value in applying analytics online so that changes can be made in the batch to correct for detected faults or deviations in the predicted value of key quality parameters. The tools selected to implement online analytics should provide effective solutions to address the varying operating conditions and process holdups.

The PCA/PLS techniques applied in batch processing should account for the physical differences between interchangeable units, the product being produced and the operations associated with a batch. To provide this capacity, the PCA/PLS model should be stored and organized by process unit. For each unit, models may be organized by product and operation (See Figure 3).


Figure 3. PCA/PLS models should be organized by process unit account for the physical differences between units, the product and the operations associated with a batch.

As the batch progresses through its operations to produce a product, the model that is associated with that product and operation should be used automatically in the online system. To facilitate the application of analytics to a batch process, the tools used for model development should allow historic data for batch operations to be accessed and screened by product and operation.

In some cases, collecting process data in a format that can be used by a given analytic tool is one of the greatest challenges in PCA and PLS model development. However, when those tools are integrated into control systems, it becomes possible for a manufacturer to provide information for each batch automatically. Three techniques have traditionally been used to unfold batch data for use in model development: time-wise unfolding, variable-wise unfolding and batch-wise unfolding. However, for online PCA analysis, a relatively new approach known as hybrid unfolding offers some significant technical advantages. (See T. Blevins & J. Beall. “Monitoring and Control Tools for Implementing PAT,” Pharmaceutical Technology, March 2007, supplement.)

The time required to complete one of more operations associated with a batch may vary because of process holdups or processing conditions. However, the batch data used in model development must be of the same time duration. There are many ways to achieve this. For example, to achieve uniform batch length, the data past a certain time in the batch could be simply be chopped off or compressed or expanded in some fashion to achieve the same number of time increments. However, a more effective approach is to apply dynamic time warping (DTW), which allows such variations to be addressed by synchronizing batch data automatically using key characteristics of a reference trajectory, as illustrated in Figure 4. (See M. Boudreau & G. McMillan,  New Directions in Bioprocess Modeling and Control, ISA, 2006.)


Figure 4. Dynamic time warping allows variations to be addressed by synchronizing batch data automatically using key characteristics of a reference trajectory.

Once PCA and PLS models have been developed using data from normal batches, their performance in detecting faults and predicting variations in end-of-batch quality parameters may be tested by replaying data collected from abnormal batches. Most commercial modeling programs provide some facility to test a model in this manner. More important is their performance in detecting faults and predicting variations in end-of-batch quality parameters in real time as current batches are evolving.

The Rouen Application

At Lubrizol’s Rouen, France, facility, we’re putting these ideas to the test. We are applying online analytics to batch processes for fault detection and prediction of quality parameters. This application in the specialty chemical industry contains many of the batch components commonly found in industry. We are using an analytic toolset collaboratively developed by Emerson and Lubrizol for this installation. It is specifically designed for batch applications and incorporates many of the latest technologies, such as DTW and hybrid unfolding. However, as with any engineering endeavor, the success of the project depends greatly on the steps taken in applying this analytic technology.

To address this application, we have formed a multi-discipline team including the toolset provider and experts  from Lubrizol’s plant operations, statistics, MIS/IT and engineering staffs. In the future, the approach we are using at Rouen will be further refined and followed for other applications. Therefore, we are giving considerable thought and effort to project planning to achieve an installation success. The project steps are summarized below.

3 of 4 1 | 2 | 3 | 4 View on one page
Show Comments
Hide Comments

Join the discussion

We welcome your thoughtful comments.
All comments will display your user name.

Want to participate in the discussion?

Register for free

Log in for complete access.


No one has commented on this page yet.

RSS feed for comments on this page | RSS feed for all comments