Transcript

Today we're going to talk about digital twins.  The topic of digital twins has gotten quite fashionable recently. When I look at the current state of digital twins it seems that the science of simulation has come a long way. However, I thought our listeners may be interested in an historical perspective on digital twins. 

When I have examined what people are calling digital twins, it really seems that people are looking at simulations. The science of simulation has been around for some time. The two broad categories of simulation are deterministic, with the outcome for a known input there is a known output; and stochastic, where there may be a range of outcomes for a known input, each with a given probability. There is an interesting heritage as we move through generations and different domains in which simulation has attempted to provide a twin in the digital world of something in the real world. 

Simulating the Factory

MRP

Material requirements planning, sometimes known as MRP1, is a methodology for working out the materials required to meet a master production schedule.  It is simulating the factory,  or at least some of its inputs, and some of its outputs. It is deterministic because the quantity of any component required to manufacture an assembly on the master production schedule is known from the bill of materials for that assembly. Material requirements planning has been part of manufacturing theory since the 1950s.  MRP one, or small MRP, is now generally part of a larger simulation of supply and demand in MRP 2 or big mrp.  

MRP II 

Expanded the entities within a factory that were simulated.  Demand was simulated through a forecast, that formed part of the Master Production Schedule.  Loading on Human and Machine resources was simulated in the Capacity Requirements Plan.  The likely movements through the Shop Floor, into and out of inventory were simulated. The Requisitions for likely purchase orders we simulated.  All these inputs were fed into a larger view of the factory.  The quality and service level outcomes also formed part of the model. 

Zooming out from the factory

SCM 

Multi Plant MRP took into simulated transport between factories and distribution centers through Bills of Distribution. 

ERP 

ERP was an attempt to simulate more of the impact on Financial and Human Resources in a single Model.   From what I saw in my working life the connection between those domains was never very strong,  Your capacity plan in MRP may have shown you that you were in need of another skilled manufacturing engineer, but there was generally a human that was making the case for headcount and raising jobposts.  

All of the above was a deterministic model once the Master Production Schedule had been set.  In the time when I was being trained, there was an acknowledgement that variability occurred and reducing variability meant that a process was under control.  It acknowledged that the real world may not match prediction, but if your predictions were consistent, you could improve your model over time.  This was all largely done with tabular forms, or visual metaphors for pegging boards, that pegged supply of demand.  

Animated Simulations

The stochastic models of the factory, the models that focused on a range of outcomes and a probability, tended to be conveyed through animations of the factory.  They were fun to watch and rounds of simulation show queues building at work centers as run times are delayed or setups take time.  This became popular when the Theory of Constraints was in vogue. You would look for where ques were forming and focus on building capacity at that point in the process. 

Zooming in to the Factory Floor

Factory Simulation 

The movement to animated representations of the factory, meant that the Computer Aided Design companies were and continue to be at the fore of this kind of simulation.  

Work Cell Simulation 

One area of simulation that always amused me was in work cell design, where avatars are placed in the work cell such that you can test the manufacturability of the part in simulation without having to go through the expense of creating the manufacturing layout.  

Predictive Maintenance  and Enterprise Asset Management

Another area close to the factory floor where digital twins have a role is in equipment maintenance.  To model the probability of a machine breakdown based on something deterministic like a maintenance schedule or stochastic such as predictive maintenance, where a sensor reading may be giving a probability of a component failure within a given timeframe.  

Zooming out to the Enterprise

Financial Modelling

Zooming out from the machine on the shop floor, to look at the simulation at the financial flows of the enterprise in the planning and budgeting.  In my experience getting the analysts to give up excel for that kind of modelling has been very hard.  The capital budgeting in Portfolio Planning tools has generally had to simulate outcomes for a project and give some notion of sensitivities.  (A measurement of how wrong you can afford to be and still have the right answer) 

Zooming out to the Market

Consumer Modelling

Finally zooming out to the digital twin of the typical customer, (or customers in aggregate in the market) through sentiment analysis.  Looking at the balance of words that indicate positive or negative emotions and attempting to measure the emotional state of the market.   

Summary and Close

This has been a very business applications view of digital twins, and I have seen simulations of residential buildings, traffic flow through cities, or even air flow within the atmosphere all be appropriately described as digital twins, but my background has really been in business applications.  If our background in business applications can help you understand how digital twins create value for your business, please reach out to us at info@softwarestrategyconsulting.co.uk or click the Contact Us tab above.