How can you use bigengine?
There are many unique applications of the platform to real life Pharma needs that can not be solved with the tools the market is currently offering. Here are some examples.
Emissions level compliance & Energy Savings
Bigengine was used to eradicate an Environmental Regulatory Problem while adding relevant energy savings in the process.
Volatile Organic Compounds (VOC) emissions levels (EL) are highly regulated in the Pharma industry. The combination of solvents and reactors in parallel processes can potentially cause alert episodes of emissions out of specifications (OOS).
After 8 weeks Bigengine avoided OOS:
- Predicting which combinations of raw materials and reactors would cause such episodes by learning from OPC gathered data (root cause effect analysis)
- Recommending optimal combinations of reactors for a given sequence of solvents.
Additionally relevant energy where achieved from lower needs of cooling power to refrigerate the VOC Burn Furnaces.
Anomaly detection in fermentation processes
Bigengine was put to search root causes in fermentation processes with a low efficiency previously non recorded.
A vaccine batch performance required a higher standard fermentation time. The batch record was not able to explain the root cause.
bigengine finds relationships between fermentations and transversal processes that can potentially impact (cleaning, maintenance, shifts, BOM, …).
Aggregating data (operators, cleaning agents, cleaning actions, speed, temperature, etc.) and normalizing processes, Bigengine detects anomalies and determines the root cause: The simultaneous use of two authorized solvents in the same cleaning process was later slowing down the fermentation process.
Bigengine recommends optimal cleaning processes.
A cleaning process follows stablished phases and cycles. In flexible sites (eg. CMO) the possible combinations of manufacturing entails a complexity that is mitigated exceeding the needed number of iteractions per phase and the overuse of cleaning agents.
Bigengine finds the relationships between cleaning processes and all the processes and context that interact with the equipment (previous WO, shift, operators, next WO, holding time, solvents, etc.) and generates a model of continuous learning that is able to recommend, for every cycle, the parameters and values for the optimal cleaning at the given Quality Specifications.
✓AI (NN + RandomForest)
Enhanced Golden Batch
Current real-time batch data is correlated with historical batch information, verifying that the multiple variables are moving inside a “tunnel” defined by the “Golden Batch” and the limits imposed by the Design Space of the given set of variables.
The OOS and the OOT are part of the checks that are performed in real time. The predictive models and deep learning algorithms that include additional variables not considered in the initial Design Space (operators, maintenance operations, log-book or cleaning operations) are able to detect anomalies that could drive to an early corrective action. The causality detection component proposes improvements to the Design Space and enhances the Golden Batch definition.
✓AI (NN + RandomForest)
Predictive Overall Equipment Effectiveness (OEE)
Bigengine increases the value of traditional OEE by finding hidden relationships between the Availability, Performance and Quality based on contextual information gathered from SCADA/DCS, ERP, LIMS, MES/BMS.
It also provides performance proposals through Bigengine predictive features.
✓AI (NN + Clustering)
PAT & ICH8 Driver
The Continuos Process Verification is considered the status of the art regarding the product manufacturing. Many factors are needed to achieve this maturity level but the use of Process Analytical Technologies integrated with the deep knowledge in the process, is a main driver to succeed.
When massive information is ingested in real time as an input to Deep Learning Models, built with the science and the experience of previous batches, the output of such models allow to:
- Identify and explain all critical sources of variability.
- Manage the variability inside the process itself.
- Predict the quality of the product and its attributes in the design space established by the materials, the environment and other conditions involved in its manufacture.
Quality can not be verified only in products; should be checked starting from the design and considering the entire process variables.
Scenarios: Visibility across sites and CMOs
To ensure the expected quality during manufacturing processes, some steps must be considered:
- Identify and measure the critical attributes of materials and related product quality processes.
- Design a measurement system that allows processes in real time (e.g. at-line, on-line or in-line) monitoring – these critical attributes.
- Design processes that provide control adjustments to ensure the correct values for these critical attributes.
- Develop mathematical relationships between product quality attributes and measures during the manufacturing of a product.
When the manufacturing processes are split across different sites or even distributed through CMOs, the definitions of the mentioned steps must also be ensured.
Bigengine provides the ability to create scenarios where the data model that has been previously designed on site is used off-site on the CMOs generated data and it can be visualized and analyzed on-site by the product owner.
Thus the manufacturing process data flows transparently and safely across sites and also from the CMOs activity.
With Bigengine, our customers have the ability to ensure the quality of their products analyzing in real time the off-site production, applying advanced analytics and deep learning on the data generated outside and with compliance with the Data Integrity guide lines.
✓Massive regulated data capture & aggregation
✓Real Time Clustering (Used to identify specific behaviors produced in the CMO)