Data Analytics in Batch Operations

By Robert Wojewodka and Terry Blevins

The trend in process plants design is to use continuous process operations, since they can often achieve higher production rates with lower capital expenditures. However, in many industries, one or more batch operations may be used in their manufacturing processes. Batch processing  often is required when small lots of material are to be produced, or when the product is produced through chemical or biological reactions that take time to complete. For example, the specialty chemical and life-sciences industries depend heavily on batch processing to produce low-volume, high-value products.

The control logic used in batch processing is often complicated by the fact that multiple products may be produced using the same equipment. The wide range of operating conditions and changes in the process that effect dynamic behavior are often a challenge from a measurement and control standpoint. Often there are added challenges in assembling, cleansing and sequencing the necessary and sufficient data to conduct a thorough data analysis. Thus historically, the application of online analytics for fault detection and prediction in real time of the end-point value of quality parameters have not been addressed adequately. Yet these remain priorities for companies that use batch processing.

By design, each input to a batch process can potentially effect some or all of the measurements used with that piece of equipment. This interactive nature of the process, combined with the slow integrating response characteristic of many batch processes, makes it difficult for an operator to identify abnormal conditions in the batch and evaluate their impact on the final product quality. The fact that quality parameters are often not available as online measurements also complicates the operator’s job. Furthermore, the historical approach engineers have taken to understanding batch and continuous operations has been through simple reports, batch end-summary statistics and simple overlays of process trends. Although useful and informative, these approaches fall short of characterizing process variation and unraveling the multivariate nature of process relationships.

Over the last 10 years, significant progress has been made in the development of technology and understanding needed to apply online statistical analytics to batch processes successfully. Integrating such tools into a process control system can provide many benefits. As one moves forward in selecting tools for online analytics, it is helpful to understand many of the challenges of applying data analytics and the changes in technology that enable success.

Batch Production Challenges

Regardless of the industry, a great many plants use some combination of batch, continuous and semi-continuous processes, as illustrated in Figure 1.

The batch unit-processing may be broken into one or more operations, depending on the product being manufactured. In the initial portion of a batch operation, the unit may be charged using discrete addition and/or by continuous feed, or some combination of discrete and continuous feed throughout the operation. Some of the inputs to the batch unit may be shared resources; others are dedicated to the equipment.

Feed rates and operating targets are determined by the target product and the operation of the batch. In some cases, multiple pieces of equipment may be available for processing a batch. Even though the equipment may appear to be the same physically, experience has shown that the performance of interchangeable units used for batch processing will often vary because of physical differences; e.g., heat exchanger area, capacity, valve characteristics, measurement location.

Within each batch operation, the material charged to the unit is retained within the unit and processed by a chemical reaction and/or mechanical means, such as heating, cooling and/or agitation. Once all material associated with a batch has been processed, the final product may be discharged to another unit for further processing or to a tank for storage. Lab analysis of some input feed streams may be available and may affect the charge rate and total charge to a unit. Quality parameters associated with the product may be available at the end of the batch or after a batch operation. The operator or the control system may use this data to correct the charge or the operating conditions used in the next phase of the batch operation.

There are significant differences in operating a batch process or a semi-continuous process and a continuous process. From a statistical analysis perspective, all continuous processes have batch aspects to them, and all batch processes have continuous aspects to them. In many cases, tools designed for continuous process analysis do not have features that are needed to address batch requirements effectively. In particular, some key areas that must be addressed to apply data analytics to a batch process successfully include:

• Process holdups. Operator- and event-initiated processing halts and restarts. Sometimes these halts and restarts are part of the batch process design, such as adding a special ingredient. Other times progression of a batch may be delayed by limitations imposed by the need to wait for common equipment to become available. Regardless of the initiating cause, the time to complete a batch operation may vary, which affects the way data must be processed during analytic model development and in the online application of analytics.
• Access to lab data. Due to the nature of the product produced by batch processing, online measurement of quality parameters may not be technically feasible or economically justified. Thus, it is common practice to take a grab sample and analyze it in the lab during various points in the batch process. In many facilities today, the lab data from grab samples may only reside within the lab system. The lab results may be communicated to the operator by phone or through a lab terminal that may not be tied to the control system. To implement online analytics, it is necessary for lab results to be available to the online analytics toolset.
• Variations in feedstock. The charge to a batch may come from storage tanks that are periodically refilled by upstream processes or by truck or rail shipments from outside suppliers. Changes in the incoming raw material properties may directly affect batch operation and quality parameters. Although the supplier may provide the properties of each material shipment, this data may be available only to purchasing or to the QA lab. If it is not available for use in online analytic tools, then this lack of information can affect the accuracy of predictions provided by the analytics.
• Varying operating conditions. A batch may be broken into multiple operations. The processing conditions may vary significantly with each batch operation and with the product that is produced by the unit. Thus, the analytic model that is applied to a batch should take into account the product that is produced and the operation that is active on the processing unit.
• Concurrent batches. Typically, most companies run concurrent batches in the same process cell. As soon as a piece of equipment is available, another batch is started. Companies do this to maximize throughput. Thus, multiple batches of the same material may be found within the unit at various stages of completion. The data collection and analysis toolset must work efficiently in that environment so each batch is analyzed and may be reviewed by the operator.
• Assembly and organization of the data. One of the limits that often prevents detailed analysis of batch processes is the inability to access, correctly sequence and organize a data set of all of the necessary data. This requirement must be fulfilled to analyze the process and move the results of the analysis online.