Cyber-Physical System power Consumption

Files

This dataset is comprised of 5 CSV files contained in the data.zip archive. Each one represents a production machine from which various sensor data has been collected. The average cadence for collection was 5 measurements per second. The monitored devices where used for hydroforming.

The collection period covered the period from 2023-06-01 until 2023-08-05.

Data

These files represent a complete data dump from the data available in the time-series database, InfluxDB, used for collection. Because of this some columns have no semantic value for detecting production cycles or any other analytics.

Each file contains a total of 14 columns. Some of the columns are artefacts of the query used to extract the data from InfluxDB and can be discarded. These columns are: results, table _start, _stop

results - An artefact of the InfluxDB query, signifies postprocessing of results in this dataset. It is "mean".

table - An artefact of the InfluxDB query, can be discarded.

_start and _stop - Refers to ingestion related data, used in monitoring ingestion. 

_field - An artefact of the InfluxDB query, specifying what field to use for the query.

_measurement - An artefact of the InfluxDB query, specifying what measurement to use for the query. Contains the same information as device_id.

host - An artefact of the InfluxDB query, the unique name of the host used for the InfluxDB sink in Kubernetes.

kafka_topic - Name of the Kafka topic used for collection.

Pertinent columns are:

_time - Denotes the time at which a particular event has been measured, it is used as index when creating a dataframe.

_time.1 - Duplicate of _time for sanity check and ease of analysis when _time is set as index

_value - Represents the value measured by each sensor type.

device_id - Unique identifier of the manufacturing device, should be the same as the file name, i.e. B827EB8D8E0C.

ingestion_time - Timestamp when the data has been collected and ingested by influxDB.

sid - Unique sensor ID; the power measurements can be found at sid 1.

Annotations

There are two additional files which contain annotation data: 

scamp_devices.csv - Contains mapping information between the dataset device ID (defined in column "DeviceIDMonitoring") and the ground truth file ID (defined in column "DeviceID")

scamp_report_3m.csv - Contains the ground truth, which can be used for validation of cycle detection and analysis methods. The columns are as follows:

ReportID - Internal unique ID created during data collection. It can be discarded.

JobID - Internal Scheduling Job unique ID.

DeviceID - The unique ID of the devices used for manufacturing needs to be mapped using the scamp_device.csv data.

StartTime - Start time of operations

EndTime - End time of operations

ProductID - Unique identifier of the product being manufactured.

CycleTime - Average length of cycle in seconds, added manually by operators. It can be unreliable.

QuantityProduced - Number of products manufactured during the timeframe given by StartTime and EndTime.

QuantityScrap - Number of scraped/malformed products in the given timeframe. These are part of the QuantityProduced, not in addition to it.

IntreruptionMinuted - Minutes of production halt.

scamp_patterns.csv - Contains the start and end timestamp for selected example production cycles. These where chosen based on expert users.

Jupyter Notebook

We have provided a sample Jupyter notebook (verify_data.ipynb), which gives examples of how the dataset can be loaded and visualised as well as examples of how the sample patterns and ground truth can be addressed and visualised.

Note

The Jupyter Notebook contains an example of how the data can be loaded and visualised. Please note that both data should be filtered based on sid; the power measurements are collected by sid 1. See Notebook for example.