Standards-aligned AI

For time-series anomaly detection and predictive maintenance too,
we connect standards frameworks to design documents

Across SaaS and Edge HMI, EdgenomiX designs sensor data acquisition, condition monitoring, remaining useful life estimation and AI governance not as separate features but as a single operating framework. The standards below are not a “certification” marketing line but a reference framework that links requirements, data, models and operating criteria.

  • Time-series anomaly detection
  • Predictive maintenance / RUL
  • Data quality
  • Explainability
  • Robustness
// 01 — OVERVIEW
Reference areas

Standard areas reflected in product design

We have organized the criteria actually needed for time-series anomaly detection and predictive maintenance into four axes. The table below shows each standard's application points in more detail.

CM&D architecture and equipment criticality

By first defining the monitoring layers, data flow and equipment priorities, we structure “what to watch and how often.”

  • Data processing, communication and presentation structure based on the ISO 13374 family
  • Monitoring program design in the context of ISO 17359
  • Sensor and alarm priority adjustment based on equipment criticality assessment

Predictive maintenance and RUL

After anomaly detection, we connect condition assessment, deterioration trends and remaining useful life (RUL) review to operational actions.

  • Reflecting the ISO 13381-1 prognostics / RUL context
  • Separating the caution, warning and inspection-recommendation stages
  • Result presentation linked to maintenance priorities

AI/ML systems and data quality

Rather than looking at the model alone, we manage data, training, evaluation, deployment and quality reporting within a single framework.

  • Organizing ISO/IEC 23053 AI/ML components and flow
  • ISO/IEC 5259 data quality assessment and reporting framework
  • Quality management of missing data, synchronization, units, labels and history

Explainability and robustness

We build a structure that records explanation rationale and robustness reviews so operators can trust the results and act on them.

  • Reflecting ISO/IEC TS 6254 explainability and interpretability goals
  • Reflecting the ISO/IEC TR 24029-1 robustness assessment context
  • Accumulating alert rationale, reference data and drift inspection logs
// 02 — HOW IT WORKS
CM&D 5-layer blueprint

The flow from on-site acquisition to RUL and explanation logs

Predictive maintenance based on time-series sensor data does not end with “a single alarm line.” Only when acquisition, quality checks, state detection, health assessment, prognostics, explanation and operational records are all connected can it be used in real operations.

1
Data acquisition

Define the required sensors, tags and acquisition intervals according to equipment criticality.

2
Quality & manipulation

Check missing data, noise, units, time synchronization and label quality.

3
State detection

Generate early warnings using thresholds, rates of change, anomaly scores and comparison with similar history.

4
Health assessment

Connect cause hypotheses and impact to screens and explanations that operators can understand.

5
Prognostics & action

Link RUL or inspection recommendations to maintenance schedules, alerts and reports.

// 03 — SPEC
Detailed mapping

Application points by standard

ISO 13374Organizes the data processing, communication and presentation structure in the context of the CM&D five-layer model.
ISO 17359Defines equipment criticality and the scope of the monitoring programme first.
ISO 13381-1Reviews the prognostics / RUL estimation structure after anomaly detection.
ISO/IEC 23053Organizes the components and operational flow of AI/ML systems in a common language.
ISO/IEC 5259Evaluates and reports missing-data rates, synchronization, units, and label and history quality.
ISO/IEC TS 6254Defines explainability goals and how rationale is presented from the operator's perspective.
ISO/IEC TR 24029-1Reflects checkpoints for neural network robustness, sensitivity, drift, and false positives/negatives.
Deliverable examplesEquipment criticality matrix, tag definition sheet, data quality report, model validation document, explanation log screen, maintenance priority table
// 04 — FAQ
FAQ

Frequently asked questions about applying standards

Does reflecting a standard mean immediate certification?

No. The standard references on this site refer to design criteria and review items. Official certification, declarations of conformity, test reports and audit scope must be defined separately for each project.

Do you apply the same level to every project?

Not necessarily. In the PoC stage you can start with data quality, explanation logs and warning-stage design, then gradually expand the level of RUL, audit trails and documentation in the operational stage.

Why can applying standards be a selling point?

In industrial AI, the sentence “accuracy is high” alone is not enough. Trust comes only when you can explain by what criteria the data is reviewed, on what basis warnings are issued, and what the operator should do.

Which products is it most directly connected to?

Time-series sensor-based anomaly detection and predictive maintenance are most directly connected in EGNOX Edge HMI and EGNOX Cloud. If needed, this can be extended to an IoT gateway, MES and Custom Embedded Development.

Next step

Let's review your time-series anomaly detection and predictive maintenance project
together in a standards context

We can organize together the level you need, from equipment criticality, sensor configuration, data quality and anomaly detection approach to RUL output and explanation log scope.