Managing Your Data

Real-Time Equipment, Instrument, and Lab Data Integration

Batch Expert provides a platform to interface real-time, offline, and laboratory data. This data becomes available for data warehousing and reporting, process control and monitoring, event detection, alarming, and auditing. Interfaces are available to real-time instrument such controllers and distributed control systems. Off-line measurements, such as glucose and off-gas analyzers can be brought into Batch Expert through a variety of interfaces. In addition, laboratory assays directly enter through simple web browser or collected from laboratory information systems (LIMS). In combination with batch header information that describes the initial culture provides a means to streamline and optimize the process development or manufacturing process.

Data Warehousing, Retrieval, and Reporting

In many laboratories, it is difficult to quickly collect batch header, real-time, and laboratory information regarding a particular batch run. Generating reports based upon this data is often an arduous and time-consuming process. Based upon affordable off-the-shelf databases, it is possible to centralize this data so that it can be archived and quickly retrieved in the future. This includes batch-header information, real-time, and laboratory measurements. Batch Expert and its database are configured specially for batch processes. This ensures data is efficiently stored and easily retrieved. End users have a seamless and simple interface to configure batch runs to collect data and retrieve historical data and their desktop. In a matter of moments, users can query past data based upon such criteria as batch-id, operator or project name, culture type, and more.

Event Detection and Auditing

Event detection is a key component in improving a process through real-time process characterization and identification and is ultimately crucial for control of critical process parameters. Events may be categorized as internal physical events occurring within the process or external control events taken by either an automatic control system or manual human intervention. Examples of physical process events include: the pH is increasing, a sharp peak in DO, a process sensor has failed. Examples of control events include: the operator has inoculated the fermenter, the operator has increased the agitation set point, the control computer is ramping up the feed flow. Many of the physical process events can be inferred by looking at the raw process and laboratory data. Although the human eye can detect some events by looking at trend or Shewhart charts, it is important to be able to detect and record these events automatically on-line using mathematical algorithms provided by Batch Expert. The element of time is also an important aspect of all events. Certain events occur at a specific time. Some events are of a discrete nature such as “valve A has been opened at 10:05 AM”. Other events occur over a range of time such as “the temperature has been increasing for the past 6 hours”.

Once events are detected and logged to a database, they provide a unique audit trail of the process and are used for process diagnosis and triggering control actions. Events can also be used to create more abstract events. For example, by looking at a detected respiration peak, a microbiologist (and hence the Batch Expert software) may infer that induction has begun. Induction is now a new event that can be used for taking further control actions.

Managing the Control of Your Batch Process

Recipe Based Control

Batch Expert includes a flexible environment to develop, implement, maintain, and transfer supervisory control across multiple equipment units, data interfaces, and controller types. Control strategies are implemented through batch recipes. Recipes are a means to manage and implement different strategies based upon product type or experiment i.e. medium type, strain, culture, etc. Recipe are based upon procedural elements that can include event-based control, PID control, alarm configuration, semi-automatic operator interaction, etc. The recipes facilitate transitions differing strategies are implemented based upon events. These events are as simple as the current phase of the batch or as advanced as detecting and reacting to the current metabolic phase in combination to batch age, feed rates, etc. The transparency of recipes is the best means to develop, scale-up and transfer new processes from the bench-top to pilot or production scale.

PID and Supervisory Control

PID (proportional integral derivative) control is the foundation of most control systems for the control of single feedback control loops. In day-to-day operation, special considerations are often needed to improve PID control. This is due to the transient nature of the process and desired set points, the recipe and phases of a batch, and the heuristic events that define a batch such as induction. Batch Expert’s PID controllers accommodate recipe-based batch phase transitions, flexible set point profiles, event-based detection and reaction, and advanced applications such as supervisory control, cascade control, and gain scheduling. Batch Expert is designed with both the operator/technician in mind. Operators and technicians have flexible but limited capability to make changes to set points across batch phases while developers can design advanced control methodologies that can then be transitioned to operations.

Event-Based Control

Once events are detected, they can be used to automatically invoke control actions. For example, detecting that induction has begun may invoke a different feeding strategy. Detecting that a pH sensor has failed can cause the Batch Expert software to begin using soft sensor pH value from a neural network for feedback control of caustic to the process instead of using the actual pH sensor.

User Interaction

Recipe Interaction

Scientists and engineers can work with Batch Expert to create, edit, and modify various recipe objects in the recipe library. These recipes specify how event detection, process control, and alarms will be handled throughout the course of the batch process. End users such as operators can choose and initiate these recipes when a particular batch is ready to run. Recipes include features for semi-automatic transitions. These transitions require operator interaction. For example, a phase change may require operator interaction or a change to equipment that requires manual intervention.

Operator Alarms, Recommendations and Checklists

When running a new batch or beginning a new phase or step within a batch, a list of checklist items is often useful to display to the operator. Batch Expert provides facilities for entering checklist reminders to the operator before starting the batch or beginning a new batch phase or step. While a batch is running, certain events may trigger alarms for the operators, which may require some manual intervention to the process. Batch Expert provides a rich alarm display that allows users to view, sort, acknowledge, and obtain further details on alarms. Events or alarms may trigger more detailed recommendations for the operator, which are displayed in a spreadsheet format.

The ability to view live and historical data in a trend chart is crucial for scientists, engineers, and operators to visualize and analyze a batch process. Batch Expert provides configurable multi-pen trend charts, which can display the real-time data or even allow the use to overlay historical data with the real-time data. Different batches can also be displayed on the same trend chart for quick visual comparisons.

Improving Process Development and Research

Advanced Control and Process Monitoring

Batch Expert is a unique platform that can accommodate research and development in process control, monitoring and optimization. The platform is built upon an expert system that can combine rules and models. With minimal training and a basic understanding of computer programming advanced users can develop either procedural based programs or utilize the built in block programming language. Advanced control and monitoring application can combine models such as neural networks with rules and events.

Neural Network Modeling and Data Mining

Batch Expert includes a tool to develop neural network models off-line based upon either experimental or historical data. By following good modeling practices it is possible to mine data to optimize the process. This can include optimization of steady state operating conditions or models that predict the optimal feed profile. Additional applications include developing software-sensors for unreliable or off-line measurements.

Technology Transfer

Batch Expert improves technology transfer and the effort required to develop process by combining several features. The key features are system and data integration, recipe design and implementation, data warehousing and retrieval, improved control, monitoring, and alarming, and consistency of operation from the bench-top to the pilot scale.