VitalAI: Failure Prognostics Model Generator for Predictive Maintenance
Iniciative
Project
Lead Organization
General Information
Challenge, Value & Description
Performance, Access & Contact
1 – General Information
Partners
Sectors addressed
AutomotiveMachinery equipmentMechanical engineering
Application categories covered
Ai as a service
Lifecycle level covered
Digital Engineering
Smart Maintenance
Customer Service
Circularity
Bertha-Benz-Straße 2, 64625 Bensheim, Alemania
Geographical Scope
- Europe
- DACH (Germany, Austria, Switzerland)
- Central & Eastern Europe (Poland, Czech Republic, Hungary, Slovakia, Romania, Bulgaria)
- Southern Europe (Italy, Spain, Portugal, Greece)
- Benelux (Belgium, Netherlands, Luxembourg)
- Nordics (Sweden, Denmark, Norway, Finland, Iceland)
- UK & Ireland
- Baltic States (Estonia, Latvia, Lithuania)
- North America
- United States
- Canada
- Mexico & Central America
- Latin America
- Brazil & Southern Cone (Argentina, Chile, Uruguay)
- Asia
- East Asia (China, Japan, South Korea, Taiwan, Hong Kong)
- Southeast Asia (ASEAN) (Indonesia, Malaysia, Thailand, Singapore, Vietnam, Philippines)
- South Asia (India, Pakistan, Bangladesh, Sri Lanka)
2 – Challenge, Value & Description
Challenge
- Slow project acquisition: Sales and engineering spend months aligning scope, legal terms and data‑access rights; data finalization alone takes 2-3 months.
- Data arrives fragmented: multiple clouds, formats, and qualities. This triggers iterative cleaning and rework, often from scratch
- Our predictive‑maintenance deliveries are typically one‑off engineering projects: we re‑clean, re‑feature and re‑train the models per customer.
- This results in low scalability and time‑consuming deployments tailored to each OEM’s tech stack.
Value
- VitalAI application via the dataspace will transform predictive maintenance projects into a subscription‑based Failure‑Prognostics Model Generator.
- Raw telemetry from data mature organizations flows via Catena‑X, 1000s of domain features are engineered and the best performing model is auto selected for future use by VitalAI.
- Non‑data experts can build predictive tools quickly to get VIN level risk scores that support existing planning systems to cut unplanned downtime, lift availability.
- Our method focuses on high risk VINs, thereby reducing warranty exposure and improving customer experience.
Description
- We exchange OEM telemetry through a sovereign, standards-based gateway using EDC technology.
- Because it speaks IDS protocol, security & legal approvals drop from months to weeks, and data remains under OEM control.
- Our semi‑automated AutoML tool, Vital AI, lets domain experts trigger pipelines via its UI — convert raw channels (e.g., battery pack voltage, current, temperature) into predictive covariates, inject domain-crafted features for model training.
- VitalAI provides via its UI a VIN level risk score table to support existing planning systems for scheduling maintenance events.
Data Value Chain Description
Infraestructure Elements
- Sovereign Cloud
- Public Cloud
- Private Cloud
- HPC
3 – Performance, Access & Contact Info
Performance
- The VitalAI backend, the frontend (UI) interfaces are working at the MVP level and ready to be deployed.
- We can use our pipelines quickly for prototyping ML models with new datasets, which is already an advantage in comparison to previous “from-scratch” development.
- The barrier of development of predictive models have been reduced within our team.
- Our Data Integrity & Integration team is locally instantiating the EDC Connector to simulate the exchange of data assets locally.
- Colleagues are also working towards defining data sub-models to be compatible with the requirements of the DSP.
- Next step is to make our EDC interface with data mature OEMs to exchange telemetry through the established EDC.