Your Manufacturing Execution System needs a sibling
April 23rd, 2024
Why Manufacturing Execution Systems alone don't cut the mustard for digital transformation, and where Manufacturing Intelligence can step up to the plate.
Walking into a factory is an invigorating experience. The constant action, the hum of machinery working at pace, products processed at speed with pinpoint precision. Engineering excellence, often at its finest.
But this routinely clashes squarely with the digital systems that accompany. Interfaces that come from the 1980s with hard-to-use UI. Limited means to collect, analyse, and infer powerful insights from data in real-time. Operators still recording batches & defects on pieces of paper that then have to be painstakingly entered into a system for record keeping.
This juxtaposition is fascinating because manufacturing is one of the most data-rich industries in the world. The logs from a single machine can generate 5GB of data per week . But it is also one of the industries that have taken the longest to digitise, and is only now beginning to leverage the deep insights from data that has been gathered over years.
The question is, why?
The OT-IT divide
Most industries (i.e. tech, services, finance) have one broad environment where they gather, store and analyse their data - namely their IT environment.
In manufacturing, there are two: the OT environment (standing for Operational Technology), and the classic IT environment.
The OT environment captures all the information flows and control systems on the factory floor: when machines are on, certain actions a machine can take under certain conditions, controllers for those machines (PLCs) etc.
These control systems needs to be absolutely reliable to ensure maximum efficiency. Therefore, they are designed to be robust & to have fewer features & logic that are battle-tested and proven to work. For example, PLC programming is a completely different paradigm to conventional software engineering (usually using graphical coding tools!). Programming languages which are known to induce subtle errors in run-time execution if you don’t manage data types carefully (i.e. JS, Python) are not a good fit here. Less is more in the OT world.
Partly because of this, OT systems often do not have enhanced security protections that are very common in IT systems. Therefore, they are subject to a risk of being hacked. And if you’re able to hack controllers and therefore underlying machinery, you can jeopardise production, which could kill a manufacturer.
Therefore, IT and OT systems have historically been separated, with the OT stack usually being separated from the public Internet entirely.
Bridging the divide
Over the last decade or so, there has been a strong push to connect OT and IT networks together. The advantages of doing this well are two-fold:
Business teams outside of production (i.e. sales, marketing, data, finance etc) can get powerful insights into how manufacturing is performing. This can help sales teams to understand whether SLAs will be breached, finance teams to learn which products are faster & cheaper to make, data teams on how to optimise operations.
Manufacturing teams in production can get enriched context from the rest of the business so that they can gain foresight into future production needs and adjust how they manufacture. From sales, they can see which customer orders are coming and when they’re expected. From finance, they can understand the cost structure of their SKUs, and identify opportunities to produce products more efficiently.
The challenge - as if often the case - is with implementation. How do you safely bridge an OT environment that is designed to be protected & highly reliable, with a sea of external IT systems (ERP, CRM, accounting etc) to ensure safe yet informative exchanges of data between them?
The advent of MES
Manufacturing Execution Systems (MES) first emerged in the late 1970s and early 1980s as a solution to bridge this gap between the factory floor and the management levels of manufacturing businesses.
They are designed initially for real-time control and to monitor, follow, document, and fundamentally control the process of manufacturing goods from initial raw materials to the finished item. In effect, it is a digital representation of the flow of production through a factory.
These early “bridges” were extremely complex requiring significant configuration, consultants, and system integrators. It took years and years of implementation to get them up and running. And because back then there was no cloud, all MES systems were on-premise which limited the ability to truly bridge the OT and IT systems.
They are also incredibly hard to use. The original MES systems were designed to be technical-first with deep functionality. Thinking about how operators or production teams could interface or interact with a system was not considered. The value of putting the right data, information & insights in the hands of those who could take action at the coalface of production had not yet come to the fore.
This was a shame because with such information, production teams can (as an illustrative example):
Leverage real-time information to firefight production issues. See downtime there and then, log stop causes, track issues with scrap & defects, monitor KPIs. Get a complete view of every aspect of production to keep the flow of goods smooth & at full efficiency.
Identify long-term trends to reorganise the way production is carried out. Which production lines or pieces of equipment are more efficient than others? Which shifts perform better than others and why? How do we manage energy spend over production cycles?
Dive into production performance by product or SKU. Often one piece of equipment can produce many different types of products (i.e. different bags of pet food). But very often, the efficiency of that machine varies significantly between different products. Understanding this is extremely important as part of organising the flow of production in the most optimal way possible.
Close the loop between production planning and actual performance. Being able to predict how long production actually takes from real historical data is much more powerful than ‘finger in the air’ estimates. It helps align production and sales teams to identify precisely which products can be reliably produced to meet customer requirements.
Keep to the highest standards of compliance. This may sound boring, but in many manufacturing industries, there are tight regulations around monitoring of processes, all with the aim of protecting you, the customer. Manufacturers historically have kept many of these records on paper, detailing individual batches, defects and downtime issues. Paper records are extremely cumbersome for QA teams (or frankly, anyone!) to comb through.
Despite bringing OT and IT systems closer together, manufacturers have not been able to translate this into long term productivity gains. In fact, official US labour productivity figures have shown that manufacturing labour productivity is beginning to fall since 2010!
The role of Manufacturing Intelligence (MI)
How can we turn this around? How can we empower production teams with the tools & insights to be able to manage their operations to peak performance? How do we build a real bridge between the OT and IT worlds to enable all teams within a manufacturer to fully leverage the value of their data?
This is where manufacturing intelligence (MI) systems come in. MI is a new category of manufacturing software that has arisen to tackle the shortfalls with legacy MES. A next-generation MI system is able to subsume vast amounts of data generated from both the factory floor as well as a manufacturer’s other IT systems. They prioritise providing deep insight into production performance with intuitive applications & workflows for frontline teams & more to run their operations more efficiently. And they serve to complement existing legacy systems for ease of implementation, to ensure that there is no interruption to day-to-day manufacturing.
The components of Manufacturing Intelligence
Let’s get down to the specifics. What is manufacturing intelligence exactly? At Ferry, we view MI as the comprehensive oracle of insight behind everything that happens on the factory floor, designed to be accessible by anyone regardless of skill or background.
To achieve this, a true MI system needs to cater to various user groups (i.e. operators, quality control, management etc), each with their own needs & requirements. And a MI system simultaneously needs to be technically sophisticated enough to be able to interface with any OT or IT system that a manufacturer is currently using, able to digest & analyse information at speed.
We believe that MI systems should be highly modular - Ferry itself is designed in this spirit, a bit like Lego - and so we break down MI into three human-centric building blocks that work in tandem:
Understand what is happening together (“Connected Worker”). This is about bringing the right information to the right people at the right time, fostering collaboration & joint problem solving. Examples include:
Live monitoring of operations: for frontline teams to have a real-time view of operational performance to reduce downtime & maximise productivity.
Equipment tracking: for operators & maintenance teams to work together to monitor critical equipment & assets to prevent unplanned maintenance.
Production attainment: to align scheduling, frontline teams and sales around how production is progressing in line with customer orders and SLAs.
OEE monitoring: track critical KPIs and efficiency metrics, providing transparency across the factory floor as a starting point for continuous improvement.
Make better decisions (“Augmented Worker”). This is about bringing powerful tooling - powered by the latest innovations in generative AI - to teams to improve the quality of decisions that they need to make, and in relevant cases automate routine decisions to allow people to focus in higher value tasks. Examples include:
Enhanced scheduling: leverage insights from the factory floor to feed directly back to production planning in real-time for dynamic adjustments and to avoid over-scheduling.
Root cause analysis: when issues do occur, deep dive into metrics & data pools to triage the problem, remediate and place guardrails to prevent future occurrence.
“Digital” engineers: automate routine analysis of production lines & systems that takes up valuable time for process & automation engineers. Discover anomalies and alert relevant personnel in real-time with proposed solution paths for efficient troubleshooting.
Predictive maintenance: automate always-on analysis of critical assets to ensure full uptime and trigger maintenance when actually required, not on schedule.
Plan for the future (“Empowered Strategist”). This is about leveraging the history of a manufacturer’s data for long-term planning of future operations. Examples include:
Resource allocation: Reorganise shifts, ensure that lines are always fully fed with input, adjust the product-SKU mix in line with production performance.
Capacity expansion: Identify bottlenecks in your process flows, and understand where capex investment needs to be routed to maximise both efficiency & production.
Process reorganisation: Make the most of existing resources, and simulate how a readjustment of production impacts both throughput and the bottom-line.
Manufacturing Intelligence, under the hood
What powers these three human-centric components of MI behind the scenes? Manufacturing intelligence relies on building a data fabric that connects together disparate siloes & sources of data (across OT and IT), making information & insight accessible to MI human-centric applications.
This is based on two foundations:
Connectivity. Being able to connect to a wide range of machinery, PLCs, automation equipment, HMIs, SCADAs, sensors & more on the OT side. And additionally being able to connect to a wide range of systems on the IT equivalent: ERP, CRM, MES (especially!) & more.
Data pipelines. Being able to interface with systems is one thing; being able to transform & make use of that information is another. This is often a very difficult challenge because data needs to be processed in real-time (which can include AI inference), joined with data from other systems, and then routed to the required destinations (i.e. databases, IT systems etc).
MI solutions should always be able to connect with existing MES systems. They can gather & process the vast amounts of data thrown off from these systems, whilst still preserving flexibility, usability and a focus on providing insights for teams within a manufacturer across divisions.
In this set-up the MES becomes the system of record for tracking the flow of production from raw material to finished goods; the MI is the system of intelligence that helps manufacturers make sense of the information that their MES collects.
They are, to all extent, siblings.