Digital twins have been sparking excitement in the manufacturing industry for their potential to accelerate industrial digital transformation and unlock the untapped business value within operations technology data.
The most common type of digital twin is the asset digital twin, which can model a single machine but is limited in its ability to enable comprehensive understanding of complex manufacturing environments. Whilst no single technology can resolve this challenge, purpose-built solutions have emerged over the past decade, bridging the gap between IT and operational technology (OT). These solutions bring coherence and context to OT data, opening a promising pathway to address manufacturing challenges on a scalable level.
Sight Machine’s Manufacturing Data Platform transforms streaming factory data into a data foundation that models the entire production process. This allows manufacturers to identify how production variables like speed, temperature, force and raw material variation interact across the line to drive core factory key performance indicators such as output and quality.
For example, consider a box of pastries. The product starts out as dough and proceeds through a series of machines for baking, sheeting, cutting and packaging. The mixer speed could affect the dough’s elasticity, affecting how well it rises, which in turn impacts the shape of the pastry before it is baked in the oven. Misshapen pastries create jams in the packaging machine.
There is a vast array of potential interactions among production steps and manually discerning correlations can be challenging. Sight Machine’s Manufacturing Data Platform creates operational digital twins, modelling entire lines, machines, production steps and the product itself as it is transformed step by step from raw material into its finished form. Through the lens of an operational digital twin, we uncover the ability to quickly identify and optimise the interactions that influence our factory’s performance.
For a digital twin of a production process to be effective, organisations must overcome these five constraints:
1. Insights must be delivered in real time
To understand, control and improve production, operations teams need to work with high-frequency data to know what is happening in the moment. To achieve real-time insight, a data platform must automate stream processing to integrate all factory data sources, from sensors to data lakes, including time series and transactional data, as well as late and missing data.
2. Plant data lacks context or meaningful metadata
Many manufacturers have collected years’ worth of data in lakes and historians. However, some of this data may not have been properly labelled when it was captured. Organisations face a challenge turning what seems like a ‘data swamp’ into actionable insights. Manually processing thousands of tags to derive meaning and identify relevant data is time-consuming and resource intensive. Consequently, teams tend to rely on a limited set of tags, utilising less than 10 per cent of available data to improve production performance.
To tackle that problem, Sight Machine developed an automated and artificial intelligence-enabled system that can label and map millions of factory data points. The system, called Blueprint, was jointly developed with funding from Nvidia and Microsoft.
3. Data heterogeneity is extreme
Plants have a wide range of data types, protocols, data sources, assets and hundreds of software solutions and systems. In Sight Machine’s data-first architecture, all streaming plant data is transformed into standardised information early in the process, versus a machine-by-machine approach, which is hard to scale.
4. Modelling requirements are also extreme
Plants have hundreds of assets and thousands of process steps. To understand and relate activities among all these steps, this variety must be commonly represented.
Sight Machine’s composable digital twins use standardised schemas, including assets, lines, systems and parts to represent not only the machines and processing steps, but also the product itself as it moves from raw material to finished product, relating each batch or discrete item to the machine settings and conditions in place as that product unit flowed down the line.
5. The data environment constantly shifts
Systems will change, machines will be replaced, and sensors will be added. Data solutions need to automate adjustments to these ever-changing data environments without the need for burdensome wrangling and model revisions.
An industrial data revolution
Sight Machine’s data-first architecture provides data transformation within a software platform that is cloud native and built to address these five constraints, whilst reducing the total cost of ownership and eliminating the technical risks associated with traditional methods.
The convergence of IT and OT within an integrated enterprise architecture, amplified by the force of AI and machine learning, promises substantial business value extraction from data, igniting a wealth of possibilities across the value chain. The accelerating momentum of data utilisation marks an unprecedented opportunity in the history of manufacturing.
Jason Nienaber is chief revenue officer at Sight Machine
This article was originally published in the Winter 2023 issue of Technology Record. To get future issues delivered directly to your inbox, sign up for a free subscription.