iWave Insight / Process Intelligence
Turn production data into recommendations that help processes adapt as conditions change. Builder Agents develop and test the models behind run-to-run optimization; qualified automation applies the adjustments.
One Platform / From Data to Governed Models
Builder Agents automate modeling within a lifecycle that connects data lineage, model versions, qualification and monitoring.
Frame the problem and examine the data.
Address data gaps and prepare the inputs.
Establish how candidate models will be evaluated.
Develop and test competing hypotheses.
Review the model and evidence behind the result.
Engineers approve the problem framing, preparation and benchmark, then inspect the result. Approved model search can run autonomously between those gates.
Run-to-Run Optimization / The Operating Loop
When a process drifts, the next run should benefit from what you have learned. Insight turns process history and measurements into a recommendation. Qualified automation closes the loop.
Based on an actual manufacturing application: correcting post-epoxy drift with Insight recommendations and Autopilot adjustments. Follow the die-bonder example beneath each step.
Bring the completed run’s measurements, process conditions and target into context.
Measured Result + Process TargetPost-epoxy inspection results reveal a developing drift in the die-bonding process.
The qualified model evaluates the evidence and proposes the next adjustment, with its limits and lineage.
Evidence-Backed RecommendationInsight evaluates the drift and determines the recommended offsets for the epoxy gun’s pressure and height.
Autopilot or another qualified system checks operating limits and applies only an authorized adjustment.
Approved Equipment ActionAutopilot applies the qualified pressure and height offsets to correct the drift and prevent an epoxy error from being triggered.
Evaluate the new result against the target. Feed the evidence into the next recommendation.
New Evidence → Next RecommendationThe solution continues monitoring post-epoxy results, checking the correction’s effect and watching for further drift.
This is the basis of a qualified self-healing process: Insight provides the intelligence; the qualified automation system owns equipment action.
Explore iWave AutopilotiWave Builder Agents / Expertise at Work
The hard part is deciding what to model, what evidence is missing and which approach works. Builder Agents automate that investigation, drawing on manufacturing and data-science expertise.
01 / Define the Outcome
Upload sample datasets or connect your data sources. Explain what the data represents, the operating context and the outcome you want to achieve.
Data + Domain Context + Intended OutcomeThe walkthrough follows a chiller-warning example. The agent clarifies operating conditions and the useful warning horizon before modeling begins.
02 / Generate Hypotheses
Builder Agents develop likely modeling approaches, identify the evidence each needs and establish how the alternatives should be tested. Domain expertise shapes the investigation.
Agent-Program-Agent / Automating Expert InvestigationHypotheses can involve different models and feature choices, not just possible physical faults. Each candidate has a reason to be tested and a clear data requirement.
03 / Resolve Data Gaps
Builder Agents identify data gaps and show why they matter. Supply more evidence or choose a narrower scope; the agent does not invent the missing observations.
Evidence Gaps / Your Decision on How to ProceedAn unsupported candidate remains untested. In this example, the user continues with available evidence; the vibration hypothesis stays visible, awaiting data and excluded from ranking.
04 / Test & Compare
After the benchmark is approved, Builder Agents test model and feature combinations against the agreed objective. Compare the results and inspect why a candidate performs better.
Approved Benchmark → Autonomous Search → ReviewAssess the selected approach against subsequent readings, confirmed incidents and maintenance findings as evidence becomes available.
05 / Inspect & Recommend
Engineers inspect the result and its limitations. For run-to-run optimization, a qualified model informs an adjustment recommendation; Autopilot or another qualified system owns equipment action.
Insight Recommends / Qualified Automation ActsThe demo evaluates equipment-health warnings using example data. The die-bonder operating loop above shows how recommendations connect to a separate, qualified equipment workflow.
06 / Change the Objective
A new process or goal changes the data requirements, hypotheses and tests. Builder Agents automate that customization so expertise can serve the next manufacturing task.
AI Scalability / New Context, New EvaluationCompare the investigation needed for run-to-run optimization with the one needed for degradation warning. Each result is qualified for its own objective.
Change the objective. Builder Agents customize the data requirements, model hypotheses and evaluation to match.
The new objective gets its own evaluation and qualification.
Builder Agents automate the customization. The resulting model is evaluated for its own objective.
Builder Agents customize the data pipeline, model candidates and tests. Engineers inspect and qualify the result. Insight uses the qualified model to produce recommendations.
AI Scalability / Automating the Customization
A different process or goal changes the modeling work. Builder Agents adapt the data preparation, hypotheses and evaluation to suit it—bringing specialist knowledge to the next task.
Compare the Changed ObjectiveAnother Application
Insight can also use tool health to prioritize metrology sampling: lots from lower-health tools receive higher priority; lots from higher-health tools receive lower priority. Lower priority does not mean automatic metrology bypass.
Explore Health-Informed Lot SamplingYour Expertise. New Possibilities.