⌘K
INV-2049

Bosch WAW28620IN · Drum bearing

Vibration at spin cycle

Stage
Awaiting Validation
Triangulation
94%
Units-at-risk
42,800
Exposure
€18.4M
Loop age
Day 11
of an investigation that legacy took ~428 days
Origin
Loopback pattern engine
Owner: P. Iyer

Ranked Hypotheses

Generated by the agent, holdout-tested, ranked by posterior probability.

Human review & action
Validating routes into Production Response and raises the CAPA in SAP QM. No production action executes autonomously.

Evidence

For H1 · 87% confidence

Statement
Bearing supplier lots from weeks 32–38 (AxleWorks GmbH) show tolerance drift, consistent with undersized inner race or insufficient hardness.
Underlying data
  • Incoming inspection · BRG-W32..W38open →
  • Spec drawing 24.00 ±0.015mmopen →
  • SPC chart · inner-race diameteropen →
  • Matched serials (42,800)open →
Holdout test
Precision0.91
Recall0.86
FPR4.2%

Agent Reasoning Trace

Multi-step reasoning made inspectable. Each step is a real tool call against MES, supplier, PLM, and environmental data. This is the glass box.

Powered by TrustLayer
  1. step 1SignalTriangulator
    820ms
    Fuse SAP Service tickets, Home Connect vibration signatures, and Sprinklr social signals for WAW28620IN drum-bearing symptom
    Finding
    3,128 signals across 3 sources; triangulation strength 0.94; growth +42% w/w
    Updates ranking
    Pattern promoted to investigation
  2. step 2MES.query
    1240ms
    Retrieve production lots for implicated serials from Chennai Plant MES
    Finding
    78% of implicated serials built on Lines 1–3 between weeks 30–40, spanning all three shifts
    Updates ranking
    Shift/line hypothesis weight reduced
  3. step 3Supplier.incoming_inspection
    1810ms
    Pull AxleWorks bearing lots BRG-W28..W42 incoming inspection data
    Finding
    Inner race diameter shifted −1.8σ from spec centerline in lots W32–W38; hardness mean 58 HRC vs 62 HRC target
    Updates ranking
    Supplier hypothesis promoted to top rank (0.62 → 0.87)
  4. step 4PLM.query
    640ms
    Check for recent design changes on drum bearing housing or spec revisions
    Finding
    No design change in last 18 months; spec tolerance unchanged since 2022
    Updates ranking
    Design-change hypothesis dismissed
  5. step 5Environment.correlate
    720ms
    Correlate failure geography with regional water hardness and duty-cycle data
    Finding
    Bengaluru (soft water) leads failure rate; correlation with hardness is inverse
    Updates ranking
    Environmental hypothesis lowered to 0.04
  6. step 6Holdout.test
    2100ms
    Test leading hypothesis on 20% held-out serials
    Finding
    Precision 0.91, recall 0.86 on holdout; false positive rate 4.2%
    Updates ranking
    Confidence stamp: high