Industrial Analytics

Cross-protocol, READ-ONLY OT analytics via the ot-aiops MCP server. oee_compute derives OEE = Availability × Performance × Quality from production inputs; downtime_events auto-detects and categorizes machine stoppages (changeover / material / mechanical / quality / break) from a state series; oee_multidim aggregates OEE across machine × part × shift; asset_inventory actively fingerprints configured devices (vendor/model/firmware) into an asset register; monitor_changes captures bounded change-of-value (only the changes). Use when the task is OEE, downtime/stoppage analysis, availability/performance/quality, asset inventory / device fingerprint / IEC 62443 register, or change-of-value monitoring. Not for IT/network/Kubernetes/hypervisor/backup.

The cross-protocol analytics layer of ot-aiops (read-only). Structured JSON for an agent to visualize. The OEE/downtime analyzers operate over provided/collected inputs (fully testable without a plant); asset-inventory and monitor_changes actively read configured endpoints.

When to use

• OEE / downtime / stoppage categorization; multi-dimensional OEE matrices.
• Active asset inventory / device fingerprint (IEC 62443-flavored register).
• Bounded change-of-value (deadband-report) capture on a point.

When NOT to use

• Single-tag reads → the matching protocol *-tap skill.
• "No data"/alarm-flood/tag-quality triage → industrial-diagnostics.
• IT/network/Kubernetes/hypervisor/backup → not this tool.

Tools

oee_compute(planned_time_s, run_time_s, ideal_cycle_time_s, total_count, good_count)

OEE = Availability × Performance × Quality. Returns {availability, performance, quality} (each {raw, value, capped}), oee, oee_pct, inputs, losses. Each factor is reported raw + clamped to [0,1]; a capped performance >1.0 flags an optimistic ideal cycle time.

downtime_events(series, category_map?, min_duration_s=0)

series:[{timestamp, state}] (state = string/bool/number). Detects every running→stopped span and returns {event_count, total_downtime_s, by_category:{cat:{count,downtime_s}}, events:[{start, end, duration_s, state, category}]}. Categories: changeover / material / mechanical / quality / break / unknown (keyword heuristics, or a {state: category} override).

oee_multidim(records, dimensions=["machine","part","shift"])

records:[{<dimensions>, planned_time_s, run_time_s, ideal_cycle_time_s, total_count, good_count}] → {group_count, mean_oee, worst_performers[...], matrix:[{dimensions, oee, oee_pct, availability, performance, quality}]}.

asset_inventory(endpoints?)

ACTIVELY connects to each configured (or named) endpoint and reads its identity (S7 CPU info, EtherNet/IP controller info, OPC-UA build info, Modbus FC43, MELSEC CPU, MTConnect device). Returns {asset_count, reachable_count, method: 'active_fingerprint', assets:[{endpoint, protocol, address, vendor, model, firmware, serial, reachable, last_seen}]}. Honest scope: ACTIVE fingerprinting (we connect to each device), NOT passive SPAN/tap discovery (roadmap).

monitor_changes(ref, endpoint?, duration_s=10, interval_ms=500, deadband=0, max_changes=100)

Bounded change-of-value: polls a point and returns only the changes (with timestamps), never every sample and never an open loop. Works over OPC-UA / Modbus / S7 / Mitsubishi MC / EtherNet/IP. Hard-capped by duration_s (≤120) and max_changes (≤500).

Example

oee_compute(planned_time_s=28800, run_time_s=25200, ideal_cycle_time_s=2.0, total_count=12000, good_count=11800) → {"availability":{"value":0.875},"performance":{"value":0.952}, "quality":{"value":0.983},"oee":0.819,"oee_pct":81.94}

CLI

bash

ot-aiops analytics oee 28800 25200 2.0 12000 11800
ot-aiops analytics downtime --input states.json
ot-aiops analytics oee-multidim --input records.json
ot-aiops analytics asset -e press1 -e cell5
ot-aiops opcua monitor "ns=2;i=5" -e line1 --duration-s 20 --deadband 0.5

Safety

Strictly read-only — never writes. asset_inventory adds light connection load to each device. 缺功能提 issue/PR 欢迎留言.