Soft-Sensor Bioprocess Monitoring for Continuous Manufacturing

Infers hard-to-measure process variables in near real time for tighter process control Evidence basis: Recent bioprocess studies including AutoML soft sensors report feasibility for real-time nutrient and metabolite estimation; review evidence emphasizes lifecycle monitoring needs and alignment with continuous manufacturing guidance

The Problem

“Near-real-time soft-sensor monitoring for continuous bioprocess manufacturing”

Organizations face these key challenges:

1

Critical variables are hard to measure online or have long assay turnaround times

2

Fragmented process data across historians, MES, LIMS, and equipment systems

3

Nonlinear process behavior makes rule-based estimation unreliable

4

Model drift from raw material variability, fouling, scale changes, and equipment aging

5

Regulated environments require explainability, validation, auditability, and change control

6

Smaller firms often lack in-house expertise for continuous manufacturing design and control strategy development

7

Post-approval change assessments are document-heavy and slow

Impact When Solved

Infer nutrient, metabolite, biomass, and moisture-related variables in near real time without waiting for offline assaysEnable tighter closed-loop control in continuous operations such as spray drying and upstream/downstream bioprocessingReduce process variability, off-spec material, and operator reaction delaysSupport post-approval stability impact assessments with AI-assisted classification and recommendation workflowsProvide expert decision support for small and mid-size manufacturers adopting continuous manufacturingImprove lifecycle monitoring, model-based release readiness, and continued process verification

The Shift

Before AI~85% Manual

Human Does

  • •Review process data manually across batches and unit operations
  • •Coordinate quality and process checks in spreadsheets and reports
  • •Investigate deviations after results show out-of-range conditions
  • •Decide corrective actions based on retrospective trend review

Automation

  • •No AI-driven monitoring or prediction in the current workflow
  • •No automated prioritization of process risks or opportunities
  • •No real-time inference of hard-to-measure process variables
With AI~75% Automated

Human Does

  • •Review AI-flagged process risks and inferred variable trends
  • •Approve process adjustments and quality-related interventions
  • •Handle exceptions when model outputs conflict with operating context

AI Handles

  • •Infer hard-to-measure process variables in near real time
  • •Monitor process conditions continuously for emerging deviations
  • •Prioritize high-impact risks and opportunities for operator review
  • •Surface actionable alerts and trend summaries for tighter process control

Operating Intelligence

How it works

AI watches every signal continuously.

Humans investigate what it flags.

False positives train the next watch cycle.

Confidence89%
ArchetypeMonitor & Flag
Shape6-step linear
Human gates1
Autonomy
67%AI controls 4 of 6 steps

Who is in control at each step

Each column marks the operating owner for that step. AI-led actions sit above the divider, human decisions and feedback loops sit below it.

Loop shapelinear

Step 1

Observe

Step 2

Classify

Step 3

Route

Step 4

Exception Review

Step 5

Record

Step 6

Feedback

AI lead

Autonomous execution

1AI
2AI
3AI
5AI
gate

Human lead

Approval, override, feedback

4Human
6↺ Loop
AI-led step
Human-controlled step
Feedback loop
TL;DR

AI observes and classifies continuously. Humans only engage on flagged exceptions. Corrections sharpen future detection.

The Loop

6 steps

1 operating angles mapped

Operational Depth

Technologies

Technologies commonly used in Soft-Sensor Bioprocess Monitoring for Continuous Manufacturing implementations:

Continuous solution preparationOther
3 mentions
Continuous spray drying lineOther
3 mentions
OEE improvement workflowOther
2 mentions
Secondary drying integrationOther
2 mentions

Key Players

Companies actively working on Soft-Sensor Bioprocess Monitoring for Continuous Manufacturing solutions:

IQVIAMasterControlVeeva

Real-World Use Cases

AI decision support for post-approval stability impact assessment

When a company changes something about a medicine after approval, AI can help estimate whether the change could affect stability and what extra testing is needed.

classification + recommendationproposed; the source supports the workflow context but does not mention deployed ai systems.
10.0

Integrated continuous process control for pharmaceutical spray drying

Instead of making medicine in stop-start steps, the project links preparation, drying, and testing into one smarter continuous line that can run more smoothly.

Process optimization and closed-loop operational coordinationproposed integrated manufacturing workflow under development in pharmsd 3.0.
10.0

Expert decision support for small and mid-size firms adopting pharmaceutical continuous manufacturing

An AI copilot can turn expert manufacturing guidance into step-by-step advice for teams that do not already know how to run continuous drug production.

Knowledge-guided recommendation and workflow orchestrationproposed workflow strongly supported by the source’s emphasis on practical guidance for organizations lacking pcm expertise.
10.0

Free access to this report