Workflows
Three workshop questions, not sixty features.
Jemba does two things only: spot what falls outside the ordinary, and find the operating range that holds. The three workflows below follow from that, and each one rests on a client deployment whose figures are published.
01 · Anomaly detection
Early fault detection
“Does a serious stop start within the next hour?”
- Input
- Can be as little as the run/stop event log the plant already keeps — no sensors and no process data required.
- Signal
- Recent micro-stop activity: how many, how long they lasted, and whether they arrive in bursts or evenly spread.
- Output
- One score per machine at a fixed cadence, against a threshold that adapts to that machine rather than a plant-wide constant.
- Deployment
- Only the past is ever used to produce a score, so it goes live as-is.
02 · Process optimisation
Root cause analysis
“Out of hundreds of variables, which ones actually matter?”
- Multivariate
- Abnormal conditions are detected across variables together, not one tag at a time against its own limit.
- Traceable
- Every flag is traced back to the variables behind it — not returned as a composite score nobody can act on.
- Automatic
- Thresholds, drifts and alarms are generated from the data rather than hand-set and then forgotten.
03 · Process optimisation
Process optimisation
“Where should the line run so the yield holds?”
- Screening
- The columns you already export are screened down to the levers that actually drive your target.
- Output
- A setpoint range per retained lever, plus the outcome expected inside each zone. Not a model for someone to interpret.
- Forward
- Predicts outcomes from parameter combinations you have not run — with anything beyond your production history labelled as extrapolation.
- Your call
- The only choice we ask is a process one: which variables are settings, and which are results.
Why this list is short
Three workflows, two published deployments.
Workflows 02 and 03 rest on the same case, rubber and polymer: the variables were screened first, then the operating range was drawn. Workflow 01 rests on the aerospace and defence case. We do not add a workflow until it has a published blind test of its own.
| Workflow | Published case | Result |
|---|---|---|
| 01 · Anomaly detection | Aerospace & defence | 273 h |
| 02 · Process optimisation | Rubber & polymer | 977 → 10 variables |
| 03 · Process optimisation | Rubber & polymer | +9.5 pts |
Next step
Name one line and one target variable.
We read the export you already produce and tell you within a week whether it carries the levers.
