Platform Setup/Getting Started
What preparation is required before implementing the Sight Machine platform?
- Identify a potential list of specific projects.
- Gain an understanding of the data sources on which Sight Machine will focus. Data sources include any databases, machines, and systems. You will also want to make note of information such as network locations and login credentials.
What level of access to shop floor systems is required?
- Sight Machine needs read-only access to all relevant digital data sources.
- Sight Machine can also work with indirect access (e.g., periodic query/file exports).
Will we need to change our existing processes or systems to work with Sight Machine?
No. Sight Machine takes a data-first approach to solving customer needs. This means that the platform does not typically require any process changes for customers to derive value. The platform pulls information from your existing Manufacturing Execution System (MES), historian, enterprise resource planning (ERP), and other production systems.
What support is needed from Infrastructure and Network teams to set up the Sight Machine platform?
Infrastructure and Network roles include:
- Manufacturing point of contact (Operational Technology, or OT)
- IT point of contact
Infrastructure and Network responsibilities include facilitating troubleshooting during the setup.
Do we need to develop any custom APIs or custom software to collect data from shop floor systems in order to work with the Sight Machine platform?
- No. There are no requirements for custom API access or custom software.
- Sight Machine can work:
- With existing data sources’ APIs.
- From direct database access.
- From file/log exports.
What are the infrastructure requirements on the shop floor, and any other on-premises or cloud requirements?
- Sight Machine is a Software as a Service (SaaS), hosted solution.
- On-premises data collection is provided by a Sight Machine appliance or in a virtual machine (VM) environment.
- If you select a VM environment, hardware requirements include:
- 2 CPUs
- 16 GB of RAM
- 250 GB of disk space
- Users interact with the software through a standard Web-browser.
- Bandwidth to transmit data is required, although Sight Machine’s data acquisition technology is optimized to mitigate typical bandwidth and network reliability challenges associated with production facilities.
What sort of tool is Sight Machine?
- Sight Machine provides a platform that focuses on data capture, contextualization, and operationalization of data and analytics for manufacturing.
- Extract, Transform, and Load (ETL) is a key component of Sight Machine and the Sight Machine version has been specifically built for manufacturing.
- The patented ETL engine (the AI Data Pipeline) allows Sight Machine to transform, clean up, and place data into pre-built manufacturing data models that are ready for analysis. These data models represent real-world parts, materials, batches, machines, operations, and their interactions.
- The ability to configure the AI Data Pipeline quickly and incorporate data from machine sensors (PLC), enterprise resource planning (ERP), Manufacturing Execution System (MES), Maintenance, Quality, and other systems is a key differentiator from other analytic tools that only do trend analysis and exploration with time-series machine data.
- Sight Machine has an edge connectivity layer called FactoryTX that captures information from disparate data sources at different locations and centralizes it within the Sight Machine (or your private) cloud.
- The Sight Machine product includes endpoint management, data security, and data compression.
- It is configurable through a Web browser and is extensible if new or custom plugins (e.g., proprietary data formats) need to be developed. This can be done by you, Sight Machine, or a third-party.
- Other products typically provide an API endpoint with which you must integrate, or are local point solutions that do not allow for comparison across multiple facilities and the inherent scalability of the cloud.
Data Acquisition and Ingestion
How is data collected and consolidated from the shop floor into the Sight Machine platform?
- Option 1: Data is collected by the customer and pushed periodically to the cloud. This option is recommended when there are limited needs for real-time data access.
- Option 2: A Sight Machine machine communicates directly with customer machines, offering modest sample rates across many sources that are connected to the same network.
- Option 3: A Sight Machine machine communicates with a Sight Machine interprocess communication (IPC) data collection node that is installed next to the edge devices for collection and preprocessing, offering sub-second sample rates.
- Data is collected by querying data sources and storing/forwarding messages to the cloud.
Does the Sight Machine platform support streaming data sources (IoT) or microbatching?
- The FactoryTX software can be configured to either stream individual messages or microbatch data in arbitrary increments.
- Sight Machine typically recommends microbatching at scale because it allows for significant bandwidth savings through the compression of messages that contain largely redundant data.
What data is needed by the Sight Machine platform?
The more relevant data that you can provide, the better. That being said, the type of data typically acquired consists of:
- Machine sensor information
- Product information
- Downtime information
- Defect information
- Batch and output information
How is data harmonized from discrete sources?
- Sight Machine supports multiple data sources from across your infrastructure.
- Sight Machine connects the data from these sources based on:
- Explicit metadata labeling – [e.g., a programmable logic controller (PLC) and interprocess communication (IPC) from the same machine]
- Timestamp joining
- Line configurations
- Part serialization
- Batch definition on serial numbers or time ranges
How quickly can data be collected and consolidated into digestible information for facility managers to review?
Typically, the data can be streamed from the plant and processed in near real-time. However, this is dependent on a variety of factors, such as edge device processing power and plant networking bandwidth capacity.
Does the platform have a Standardized Data model to consolidate the data?
Yes. Data is fed into the AI Data Pipeline and then onto Sight Machine’s patented manufacturing-specific data models (i.e., the plant Digital Twin).
What is the data storage layer?
The data is stored in the cloud and includes:
- cloud storage
Can the data stored by Sight Machine be accessed via API calls?
- Yes. API access to the data is available through both REST and SQL-like interfaces.
- Sight Machine can export high-level data into a variety of standard formats and systems, including SQL databases, data lakes, CSV files, and MES/ERP.
Does the Sight Machine platform offer the ability to configure visualizations or does it support any standard visualization tools?
- You can configure dashboards using Sight Machine’s dashboard builder, which is a point-and-click/drag-and-drop interface that runs within your Web browser.
- You can also develop more complex or highly configured dashboards by working with Sight Machine’s Data Engineering Team.
- Additionally, you can also use third-party vendors for visualization (e.g., Tableau, Qlik, etc.).
What visualizations does the Sight Machine platform offer for viewing the collected data?
- Sight Machine’s visualizations provide access to contextualized data built upon disparate data sources. You can specify visualizations based on an individual’s role using the same underlying data across an enterprise. End users can configure dashboards and visualizations without the need for coding or IT support.
- When necessary, domain-specific visualizations can be built and hosted on the Sight Machine platform using the software development kit (SDK).
- Sight Machine supports third-party visualizations from Cargill or other technology providers (e.g., Tableau, QlikView, etc.) via API.
- Sight Machine also has the ability to click through and link to other analytic tools to create data exploration and analytic workflows.
- The Data Discovery Toolkit, which is a part of the Sight Machine Enterprise Manufacturing Analytics (EMA) product, lets non-statisticians explore the data and identify potential variables of interest.
- The out-of-the-box analytics are generalized to serve a broad array of manufacturing verticals and operations.
- Sight Machine’s SDK and Web APIs can be used to build domain-specific analytics that need to be operationalized or integrated into statistical software for ad-hoc analysis.
Does the Sight Machine platform provide an aggregated dataset to the data scientist?
- Yes. However, the data scientist needs to build the models using external tools (e.g., R or Python).
- Sight Machine is architected and has been used, to house analysis of data through inferential statistics and other techniques. The SDK also provides an interface for easy consumption of manufacturing data models using R or Python. The data scientist can also implement other techniques or algorithms on top of this data.
If a data scientist builds custom algorithms that leverage Sight Machine modeled data, is it possible to display results back in the platform?
Yes. Sight Machine’s analytics service is extensible and allows for the deployment of customized algorithms back into the platform. A typical customer-produced analytic is first developed using the Sight Machine Python SDK. Then, customers work with Sight Machine’s Data Science Team to deploy the analytic on the platform.
If the software does offer advanced analytic options, how easy is it for a non-data scientist (say, an engineer at a plant) to build the models and analyze the data?
- The platform offers Web-based tools for setting up streaming modeling pipelines without writing code. It also provides visualization interfaces for consuming data and running standard out-of-the-box analytics.
- The left sidebar of the Sight Machine user interface is a constant fixture throughout the platform that allows for ease of use and simplicity in training new users. This ensures that users can access common parameters such as time ranges, assets, and facilities in a consistent way.
AI and Machine Learning
What is the role of artificial intelligence (AI) and machine learning in analyzing and harmonizing the data?
- Sight Machine’s AI Data Pipeline can best be described as an Expert System supported by Machine Learning, based on hundreds of different machine types.
- The platform leverages a set of learned experiences to establish boundaries, harmonize, and recognize a variety of characteristics about the data.
- Sight Machine creates a Digital Twin of each part and process through an automated and systematic data intake process that acquires, refines, and contextualizes data. Then this initial Digital Twin can be iteratively modeled with a Data Engineer and a subject matter expert (SME).