This week, research director Mike Evans discusses deployment of the next generation of smart shop floor machines.
Experts at industry analyst and market consulting firm Cambashi contribute a regular blog for TM titled Silos Changing exploring how new software applications enable manufacturers to implement business initiatives for the new economy.
As embedded software in machines becomes ubiquitous most observers agree how shop floor machines will evolve.
There will be more sensors and actuators. Materials will self-identify to machines with techniques such as Auto-ID. Location of assets like tools will be sensed in real time. Every aspect of the shop floor will be networked to allow disparate equipment to integrate applications. Machines will be field upgradeable and continuously improve.
It’s likely that price performance will increase as standard industry software, hardware and networking components replace custom software and hardware. That should mean that adoption will accelerate, perhaps with a burgeoning refurbishment sector.
There will be a huge potential stream of information coming from the shop floor in near real time. Maybe the current shop floor control architecture and nomenclature will struggle to cope. There is a big debate to be had about where to analyse, process and act on that stream of shop floor information.
Until now, manufacturers have used ISA-88 and ISA-95, a hierarchical structuring of the assets of an industrial company.
ISA-88 focuses on the process cell and ISA-95 focuses on the site and the area. In this nomenclature, levels 0, 1 and 2 are control modules, equipment modules and units that execute production processes.
Level 3 is Manufacturing Execution Systems (MES) that prepare, schedule works orders, monitor quality and track completion of the process executed in the lower levels. Level 4 executes in enterprise applications like Enterprise Resource Planning (ERP) where longer term planning and scheduling and the associated logistics take place.
It’s already the case that the boundary between level 3 and level 4 has blurred. In recent years, more powerful computers capable of dealing with bigger databases mean it is more practical to exchange information between the ERP, MES and control systems. The granularity of planning and scheduling has improved showing more transparency of plant capacity, allowing more flexible use and reducing waste.
One school of thought advocates smarter machines making better decisions about control, with only key parameters exchanged with higher levels of control. Another advocates sending almost all the information to higher levels where data can be mined to identify patterns that, when detected, would drive future decision making. However, much of the information – for example, an instrument reading time series – is unstructured and difficult to store in traditional structured relational databases.
This matters because true optimisation requires near real-time knowledge of the status of the whole industry network. Decisions made at the machine or process level can often sub-optimise.
Consider Maintenance, Repair and Overhaul (MRO). At present the works order for machine planned maintenance is generated after time spent or cycles used. All the additional sensors could detect wear or drift from normal values and then generate a repair works order. However, the machine does not know the big picture.
The works order might be better scheduled when that work centre is not on a critical path, only the Site and Area level applications could see that. Also, stored detailed sensor information could be analysed for patterns. Identifying a pattern that previously led to a failure could allow intervention before a problem occurs.
We think that the medium term trend will be to pass the whole stream of data to central servers rather than ever increase the smartness of machines.
The computing and storage capacity needed to implement this philosophy is huge. There are solutions, Stratus and Teradata provide hardware that can handle large amounts of data. Dassault Systèmes Exalead, and HP’s Autonomy can handle unstructured information. But it will take time before applications to exploit these technologies come to market.
In the meantime, we expect machines to get smarter.
Some manufacturers will commission professional services firms to implement custom software to implement best practices. Then commercial products will emerge.
In future blogs, we will go on to write about other potential deployments that respond to consumer and business demand for smart products and devices.
