Türkçe 
English 
Data Processing Management encompasses end-to-end rules, workflows, and infrastructure components for standardizing, enriching, validating, and preparing heterogeneous data collected in a BMS environment for analytics. The goal is to create a reliable data backbone that increases operational efficiency, reduces energy costs, and supports comfort/sustainability objectives. For details: Data Processing Management.
Data Processing Management: Concept and Scope
Architecture and data pipelines: ETL/ELT and streaming.
In a BMS, the data processing layer integrates two main pipelines: Batch ETL/ELT and Real-time streaming. The batch pipeline retrieves historical data from the archive, transforms it, and writes it to the data lake/time series. The streaming pipeline pre-processes telemetry from protocols such as BACnet, Modbus, KNX, and OPC UA at the edge, delivers it to the core processing engine via a message broker (e.g., MQTT), and triggers event-based rules. This approach simultaneously addresses both freshness-sensitive operations (alarm, setpoint optimization) and analytical and reporting requirements.
Standardization, unit conversion, and enrichment.
Data processing automates point naming , unit conversions (°C ↔ °F, kW ↔ kBTU/h), tagging , and feature engineering steps. Data associated with the device/zone/floor/building hierarchy is enriched with external sources such as occupancy, weather, and tariffs. Thus, energy and comfort decisions are made based on unified and contextualized signals.
Event rules and workflows
Rule engines define threshold, ratio, derived signal, and model-based rules using an “ if-this-then-that ” logic. Example: “If CO2 >900 ppm and occupancy=1, increase fresh air; if it doesn’t decrease within 15 minutes, create an alarm and notify the shift supervisor.” Time windows, state machines, and debounce / alarm deduplication mechanisms within the workflow reduce noise and generate actionable alerts.
Data quality, observability, and SLOs
- Verification: Range/business rule checks, outlier, gap, and bounce detection.
- Improvements: Interpolation, resampling, faulty sensor flagging.
- Observability: Pipeline latency, throughput, error rate, version traces.
- SLO/KPI: Freshness (<60 sec), accessibility (≥99.9%), accuracy (>98%) targets.
Security, compliance, and change management.
End-to-end TLS, certificate management, RBAC, MFA, network segmentation (OT/IT), and the principle of least privilege secure data processing. Audit trails, versioning, signed configuration, break-glass procedures, and backup/recovery policies support auditable operation. Masking, anonymization, and access permission policy-based controls are provided for regimes such as GDPR.
Privacy, retention, and lifecycle management.
The data lifecycle is managed through the steps: ingest → hot storage → cold archive → deletion. Retention periods are classified according to functional requirements and regulations. Privacy by design is applied to records that may contain PII (Process Information Indicators).
Edge processing and buffering
To prevent data loss during network outages, secure buffering and retry strategies are implemented at the edge. Time-critical rules (e.g., freeze protection) are executed at the edge to minimize latency.
Time series, data lake, and analytics preparation.
High-volume telemetry is stored on a time series database and/or data lake. Costs are optimized through downsampling and retention policies. Semantic layers and dataset templates provide input to BI tools (dashboards, reporting) and advanced analytics/ML applications (anomaly detection, fault prediction, optimization).
99.9% Processing line accessibility <60 sec End-to-center data freshness. >98% Data accuracy rate
Centra Approach and Implementation Roadmap
Modular components
Processing Engine:
Flow + bulk line, rule engine, windowing, alarm merging, orchestration.
Schema & Labeling:
Naming templates, unit conversions, metadata, and dictionary management.
Security:
TLS, RBAC, MFA, signed configuration, SIEM integration, audit trail.
Observability:
Speed, latency, error rate, data gap/outlier tracking, alerts, and SLA reports.
Integrations and interfaces
BACnet/IP, Modbus TCP/RTU, KNX, OPC UA, MQTT for the field; REST/GraphQL API, webhooks and scheduled exports for applications. Bidirectional data exchange with CMMS/FM, energy management and ESG reporting systems.
KPIs and success criteria
- Freshness: 95% of data is processed with a delay of less than 60 seconds.
- Accuracy: Outlier rate <2%; calibration agreement >98%.
- Operation: False-positive alarm ↓, recurring alarm ↓, MTTR ↓.
- Efficiency: 5–15% annual improvement in kWh/m² and peak demand.
- Compliance: Auditable transaction trail and data lineage >95% coverage.
Use cases
Failure Prediction (PdM):
Early warning of anomalies through attributes such as vibration/current, coil ΔT, and compressor cycle count.
Demand Response & Peak Cutoff:
Schedule-sensitive setpoint and equipment staging algorithms.
IAQ & Comfort Optimization:
CO2, VOCs, fresh air flow rate according to temperature-humidity balance, and dynamic ventilation.
ESG & Reporting:
Auditable data and evidence sets for carbon accounting; automated reporting workflows.
Implementation roadmap
Stage Explanation Outputs
1) Discovery & Design Resource/protocol map, data model and business rule framework, security requirements. Target architecture, schema/label glossary, security, and SLO plan.
2) Pilot Flow guidelines, rules, panels, and warnings for the selected zone/floor. Baselines, feedback loop, improvement list.
3) Scaling Multi-building/campus deployment, automation playbooks, versioning, and CI/CD. Standard templates, documentation, training.
4) Continuous Improvement Tuning, alarm hygiene, model updates, and maintenance rituals. SLA/SLO reports, audit trails, cost/energy savings report.
Frequently Asked Questions (short)
- Is it compatible with the existing BMS? Yes; it integrates with open protocols and APIs.
- Is cloud computing mandatory? No; on-premise or hybrid deployments are supported.
- Data loss? Minimized with edge buffering and retrying.
- Security? TLS, RBAC, MFA, SIEM integration, and audit trail are standard.
- Is it scalable? Multi-building/campus and multi-tenant scenarios are supported.
In short, Data Processing Management is like the brain of a BMS: it translates scattered signals into a single common language, enriches them with the right context, and provides operational teams with actionable, reliable insights. This allows operational excellence, energy efficiency, and sustainability goals to be measurably supported.
