Beijing-FANUC Mechatronics in China needed a faster data foundation for its factory IoT platform. CNC equipment, robots, AGVs, and sensors generated a continuous stream of operating data, but ingestion backlogs, slow analysis, and infrastructure costs limited how quickly teams could use it.
The team adopted TDengine TSDB to support equipment monitoring, machining analysis, traceability, and predictive maintenance. In the reported deployment, a three-node cluster handled peak ingestion above 800,000 data points per second, while server infrastructure costs fell by 60% compared with the previous platform.
Key Results at a Glance
| Metric | Reported result |
|---|---|
| Peak data ingestion | 800,000+ data points/second |
| Server infrastructure costs | 60% lower than the previous platform |
| Latest equipment status query | Under 1 second across 5,000+ tables |
Results reported for this Beijing-FANUC deployment. They describe the specific factory IoT implementation, not a global FANUC rollout.
The Challenge: Making Manufacturing Data Available When It Matters
FANUC’s digital manufacturing initiative brings together CNC technology, machining processes, automation, and production management.
As part of this initiative, the team developed a highly automated demonstration production line equipped with hundreds of CNC machining centers, collaborative robots, automated guided vehicles (AGVs), and industrial sensors.
These systems continuously generate operational data, including spindle speeds, load currents, machine positions, alarm codes, robot movements, machining parameters, tool-life counters, vibration readings, temperature, energy consumption, and production records.
For FANUC, the challenge was making this data available quickly enough to support daily manufacturing operations.
The previous infrastructure had four limitations.
High-volume ingestion and storage pressure. During peak production periods, the platform needed to process more than 500,000 data points per second. The previous system struggled with sustained write workloads, leading to data backlogs and occasional data loss. Rapid growth in historical data also increased storage costs.
Slow operational reporting. Production managers needed timely information about equipment status and utilization. However, the existing Hadoop-based offline analysis pipeline could take several hours to turn collected data into reports, limiting its usefulness for live monitoring and alerts.
Inefficient historical queries. Engineers regularly needed to compare spindle loads across similar CNC machines, investigate alarms, or retrieve sensor data associated with a particular workpiece. Cross-device and time-range queries were time-consuming, slowing down machining analysis and troubleshooting.
Complex maintenance. The existing architecture combined multiple databases, message queues, and computing engines. Operating and troubleshooting these components required expertise across several systems.
As FANUC described in its original customer-authored account, equipment data is a key asset across machine operations, machining processes, quality control, and production management.
The team needed a time-series data platform capable of ingesting large volumes of machine data, storing historical measurements efficiently, and supporting fast queries without adding further architectural complexity.
Building a Factory IoT Data Foundation with TDengine TSDB
From Equipment Data Collection to Manufacturing Applications
FANUC connected equipment data collection, MQTT transport, and batch ingestion to a three-node TDengine TSDB cluster.
TDengine stores the time-series measurements and serves queries to the manufacturing applications, supporting live monitoring, historical analysis, and data preparation for predictive maintenance.
The architecture follows five main steps:
- Equipment data collection: FOCAS, robot APIs, and standard industrial protocols collect operating data from CNC machines, robots, and sensors.
- Data transport: MQTT carries collected equipment data through the messaging layer.
- Data ingestion: The ingestion service batches incoming messages and writes the time-series measurements to TDengine TSDB.
- Data storage and querying: TDengine TSDB provides time-series storage and query access through interfaces including JDBC, ODBC, and RESTful APIs.
- Manufacturing applications: The customer’s applications use the stored data for equipment monitoring, machining analysis, historical traceability, and downstream predictive-maintenance workflows.
This architecture separates industrial data collection and application logic from the underlying time-series database.
TDengine TSDB serves as the data foundation. Equipment acquisition, monitoring dashboards, maintenance workflows, and predictive models are parts of the broader customer implementation.
Organizing Machine Data for Efficient Queries
The team designed its data model around TDengine TSDB’s supertable and subtable architecture.
Each physical device has an associated subtable containing its time-series measurements. Tags describe attributes such as device model, workshop, production line, and equipment type.
This model allows engineers to query an individual machine, compare groups of similar equipment, or retrieve the latest device status without assembling a separate dataset for every question.
For example, a production engineer can compare measurements from CNC machines on the same production line, while a maintenance engineer can retrieve the operating history of one machine during a suspected abnormal event.
The complete illustrative schema and SQL examples are included in the technical appendix.
Handling High-Frequency Data Ingestion
FANUC deployed a three-node TDengine TSDB cluster, with each node configured with 16 CPU cores and 64 GB of memory.
According to the customer case, the cluster handled peak ingestion exceeding 800,000 data points per second.
This result is distinct from the earlier requirement to process more than 500,000 data points per second. The two figures describe workload demand and reported deployment performance, respectively, and do not establish a measured percentage improvement over the previous system.
Query Performance for Monitoring and Analysis
The implementation also addressed two recurring query requirements.
Latest status across the factory
Manufacturing dashboards need current equipment data to present an overview of the production environment.
FANUC reported that retrieving the latest equipment status across more than 5,000 tables took less than one second.
Operating-state analysis over 24 hours
Production engineers also needed to review how CNC equipment states varied across production lines.
The customer reported response times below two seconds for hourly operating-state aggregation over a 24-hour period.
This analysis provides a basis for examining equipment running-state patterns. The reported query calculates a ratio of status samples and should not be interpreted as a complete Overall Equipment Effectiveness (OEE) calculation.
The database also supports data subscriptions used by the customer’s alert-management workflow. Equipment error conditions can be forwarded to the alerting application, where maintenance work orders are generated.
Together, these capabilities give the manufacturing applications faster access to current and historical equipment measurements.
How FANUC Uses the Data in Manufacturing Operations
The business value of the deployment is most visible in two everyday manufacturing workflows: understanding equipment operating conditions and investigating machining history.
From Equipment Status to a Production Decision
Production managers need to know which machines are running, which are idle, and where operating conditions require attention.
FANUC’s manufacturing application brings equipment operating states and production information into a centralized monitoring view, using data stored in TDengine TSDB.
FANUC’s manufacturing application displays equipment operating states using data stored in TDengine TSDB.
From this overview, managers can compare equipment running states and utilization over time, identify periods of low use, and review associated alarm or process history.
Engineers can then examine relevant spindle loads, temperatures, and machining cycle-time changes to narrow down their investigation.
This workflow helps teams move from a general picture of factory operations to the recorded equipment data behind a particular condition.
The monitoring interface is part of the customer’s manufacturing application, not a native TDengine IDMP dashboard.
From a Quality Question to Machining History
When a quality problem occurs, engineers need to understand what happened during the production of a particular workpiece.
Relevant questions may include:
- What was the spindle load during the machining cycle?
- Did feed rates or operating parameters change?
- Were temperature or vibration readings abnormal?
- How did the machine behave before and after the relevant production interval?
FANUC’s application allows engineers to select a recorded time range and retrieve the relevant equipment and machining parameters from TDengine TSDB.
Engineers review recorded CNC machining parameters over a selected production interval using historical data stored in TDengine TSDB.
Faster historical queries make it easier to compare operating conditions, examine abnormal events, and prepare equipment data for further process analysis.
This is also useful for predictive maintenance workflows, where consistent access to equipment history supports the preparation of data for downstream models.
TDengine TSDB provides the underlying time-series data infrastructure; the customer’s application and analytical models perform their respective monitoring and analysis functions.
Measurable Results
FANUC reported improvements in data ingestion, query performance, infrastructure costs, and predictive-maintenance workflows following the platform upgrade.
| Metric | Reported result | Measurement scope |
|---|---|---|
| Peak ingestion throughput | More than 800,000 data points/second | Three-node TDengine TSDB cluster, each node with 16 CPU cores and 64 GB RAM |
| Server infrastructure costs | 60% reduction | Customer-reported comparison with the previous platform; detailed cost categories and period not specified |
| Latest equipment status query | Under 1 second | Query across more than 5,000 tables |
| Operating-state aggregation | Under 2 seconds | Hourly aggregation over a 24-hour period |
| Predictive-maintenance model accuracy | 25% improvement | Customer-reported result following the platform upgrade; evaluation definition not specified in the published case |
The reported results reflect this specific FANUC implementation.
Lower Infrastructure Costs
Compared with the previous platform, FANUC reported a 60% reduction in server infrastructure costs.
The result indicates a substantial reduction in the server resources required for the reported deployment.
The published account does not specify the complete cost categories or comparison period, so this result should not be presented as a verified reduction in total cost of ownership, licensing costs, or all infrastructure expenditure.
Faster Access to Manufacturing Data
The reported query results show improvements in the accessibility of operational data.
Retrieving the latest status across more than 5,000 tables took less than one second, while a 24-hour operating-state aggregation returned results in under two seconds.
These are separate query workloads, and their response times should not be combined into a general benchmark for every manufacturing query.
The customer also reported more responsive monitoring and alert workflows, supporting quicker access to equipment conditions and historical operating data.
Supporting Predictive Maintenance
FANUC also reported a 25% improvement in predictive-maintenance model accuracy following the platform upgrade, with TDengine TSDB providing the underlying equipment-data foundation.
This is a customer-reported outcome. The published case does not identify the model’s specific prediction task, baseline accuracy, evaluation period, or whether the improvement represents a relative percentage change or percentage points.
For that reason, the result should be understood as an outcome reported alongside the broader platform upgrade, not as a direct performance claim about an AI model built into TDengine TSDB.
The customer also described improved support for maintenance workflows and the prevention of unplanned downtime. No verified downtime-reduction percentage was provided.
Future Plans
For FANUC, the deployment provides a faster foundation for equipment monitoring, machining analysis, and access to historical operating data.
The team plans to explore feature preparation closer to the database and deeper integration with industrial IoT, low-code, and business intelligence applications.
These are next steps beyond the implementation described here.
Build a Manufacturing Data Foundation with TDengine
TDengine TSDB supports high-volume machine-data ingestion, fast historical queries, and the time-series data foundation needed for equipment monitoring, machining analytics, and predictive maintenance.
Technical Appendix
A. Illustrative TDengine Data Model
The following example demonstrates how a supertable can organize device measurements and equipment attributes.
It is a simplified representation based on the published case, not a verified export of FANUC’s production schema.
CREATE STABLE devices_st (
ts TIMESTAMP,
speed FLOAT,
temperature FLOAT,
vibration FLOAT,
status_code INT
) TAGS (
device_id NCHAR(64),
device_type NCHAR(32),
model NCHAR(64),
workshop NCHAR(32),
production_line NCHAR(32)
);
An individual device can be associated with a subtable using the same schema.
CREATE TABLE device_CNC_01
USING devices_st (
device_id,
device_type,
model,
workshop,
production_line
)
TAGS (
'CNC_01',
'cnc',
'CNC_Model_A',
'Workshop A',
'Line1'
);
B. Latest Equipment Status Query
The customer case describes a query that retrieves the latest recorded values across equipment tables.
SELECT LAST(*)
FROM devices_st
GROUP BY tbname;
Reported result: Under one second across more than 5,000 tables.
C. Operating-State Sample Ratio
The following illustrative query aggregates equipment status samples into one-hour intervals over the preceding 24 hours.
SELECT
device_id,
COUNT(ts) AS total_count,
SUM(
CASE status_code
WHEN 1 THEN 1
WHEN 0 THEN 0
END
) / COUNT(*) AS running_sample_ratio
FROM devices_st
WHERE
device_type = 'cnc'
AND workshop = 'Workshop A'
AND production_line = 'Line1'
AND ts >= NOW - 24h
PARTITION BY device_id
INTERVAL(1h);
Here, a status_code of 1 is treated as running, while a value of 0 is treated as not running.
The resulting running_sample_ratio is the proportion of recorded samples representing the running state. It is not equivalent to full OEE.
It also does not necessarily represent time-based equipment availability unless sampling intervals and missing data are appropriately handled. Other status codes require defined treatment.
Reported result for the customer’s operating-state aggregation workload: Under two seconds.
D. Additional Manufacturing Screenshots
The following supplementary screenshots provide further context about the customer’s manufacturing application.
Appendix Figure A: Individual CNC Equipment Parameters
Appendix Figure B: Equipment Operating Time and Utilization
Appendix Figure C: Tool Efficiency and Parameter Trends
Appendix Figure D: Cycle-Time Analysis
These screenshots are from FANUC’s manufacturing application, which uses equipment data stored in TDengine TSDB.


