Process improvement phenomena
Most of us have heard the phrase; ‘What gets measured gets done’. MES projects without the ‘E’ invariably yield excellent results. Without the ‘E’ means that the system is a measurement only system, a system of reports and indicators, production performance highlights and whatever it takes to improve the process when the skilled workers action change with good data. This even applies to business processes void of technology. It seems that when we focus our attention on measuring and improving anything we are more likely to improve it. It is important to realise that significant business improvement is a direct result of better information. It is even likely that with the right information, process improvement targets may even be met without implementing the more challenging ingredient ‘E’.


Best practice
BPR (Business Process Reengineering) encourages the selection of KPIs which will most positively improve your business. Our advise is, don’t automate first: work on your KPIs first. Understand the process and its dynamics and what to tweak to tune the process engine. In analysing data produced by production systems, be careful of the ‘cause and affect’ traps. Association does not imply causation.

As an example of poor association, in the early part of the twentieth century, it was noticed that, when viewed over time, the number of crimes increased with membership in the Church of England. This had nothing to do with criminals finding religion. Rather, both crimes and Church membership increased as the population increased.

During WWII it was noticed that bombers were less accurate when the weather was more clear. The reason was that when the weather was clear there was also more opposition from enemy fighter planes. Association does not imply causation, at least not necessarily in the way it appears on the surface.

Some rules of thumb to follow for best practice are to trigger important events in 100 per cent available environments; involve process experts in process improvement, not vendor architecture specialists; don’t try to compensate for a poor control system with back office application decision engines – understand the root cause and sort it out!; and learn from the past, understand transients, boundaries and the limits of design.


How important is the data historian?
With all this talk about better data and better information to make decisions with, how important is the data historian? Very! Often relational databases (RDB) are used for small to tiny control systems. The power of relational databases is not a good enough reason to commit to using a RDB for process data. To summarise what a relation database offers, it maintains a set of separate, related files (tables), but combines data elements from the files for queries and reports when required. The concept was developed in 1970 by Edgar Codd, whose objective was to accommodate a user's ad hoc request for selected data. Almost every business database management system (DBMS), including Oracle, DB2, SQL Server and MySQL, is a relational DBMS (RDBMS). These systems were never designed as high performance data acquisition systems for your plant equipment.

Knowing what to look for in a process historian is another discussion, however the need for such is paramount as size and performance becomes increasingly important. The closing comment on historians is this: if a RDB could cut the mustard, then process historians simply would not exist. So for seriously large applications, even the automation vendors have borrowed technology dedicated to being the ‘serious’ plant historians from the big end of town.

The hard truth
Assuming that it’s not as easy as 1, 2 and 3 to implement an MES system, what are the things to look out for? 
• End users may not have the same agenda as software vendors; however there is some synergy between mature vendors and end-user requirements 
• For end users, ROI can be improved or at least measured; a view into the production data; better planning and execution; plant can be leveraged if it is better utilized; and reduced cost for maintenance 
• For automation vendors, reinventing their space in a vertical market; competing with other automation vendors for perceived product sale opportunities; and recover investment in companion research and development efforts 
• Back-office MES applications are not mature 
• Some vendors make claims that are simply not true using VB (Visual Basic) applications to produce a GUI (Graphical User Interface) and Glue enterprise applications to SCADA or HMI 
• The real issue is often the control system or in other words the lack of planning 
• Minimal process improvement occurs if the system is inflexible or has ‘garbage’ going in 
• Sometimes enhanced reporting is the single most important requirement to improve the business providing a new view into your business assets.


Modelling your system
The first step is to model the system to manage the enterprise. This starts from the plan. This would include telecommunications, data communications, RDBMS, SCADA/HMI (Human Machine Interface) performance, RTU/PLC/DCS controls and business application requirements. All of these must be considered in unison, not in isolation, and not necessarily from a product perspective. The plan is about the bottom line, throughput, production, volume and time. This will ultimately provide a criterion which will determine if technology will provide the necessary performance.

Process modelers are an essential breed of people. They help develop bench mark specifications, not the system architecture. In order to guess their requirements, the modelers need to say what they need to do their job, not what they don’t like after they see someone else’s effort. If there is no buy in from the process modeler, the model will fail.

We all agree good information helps with making good technical and business decisions. Why is good information so critical to MES? Automation systems don’t like wild cards, that is, out of band values or variables outside the boundary of control. This is just as important to MES with ‘E’ as it is to a traditional control system.
Finally, when the model is signed off and technology is being selected, important questions should be asked: will the front office application cope with the volume or demand from back office applications? Can you get the data out in a timely fashion and will we be able to control the process if we are relying on databases to feed the decision-making application (MES)? This is a fundamental focus for the entire design process, that is, where should each sequence be performed; in the controller, HMI or MES?


Technology choice
How serious your desire to improve your process control system will partially determine the tools the MES must provide. Expectations from technology products could typically include some of the following: 

• New tools using latest development technology, for example, net support, xml and other standard interfacing 
• Workflow or production specific triggers (out of the box for things like OEE and downtime) 
• Better connections out of the box 
• Seamless integration to SAP (Systems, Applications and Products), ERP and other business applications 
• SPC (Statistical Process Control) 
• Future configuration ease (object-based point structures for modeling plant and equipment).


Enterprise system formulae
So what is the perfect model when designing an MES system? Building an enterprise solution is not about a shopping list. Building an enterprise solution is about understanding the business first, defining the outputs, defining the measure of the outputs, and defining all the inputs, and then measuring and controlling the workflows accordingly. A simple enterprise solution may not require a specific MES software application to deliver the promise. One thing is sure; you need good data to get good information. You must couple good information with your knowledge of how the business should run, well before you can design a fully-automated execution system.



The golden rules
• Mission critical decision-making should be executed on devices which are virtually 100 per cent available 
• Reality check: would you run your motor vehicle on a windows platform? Why not? Put controls, recipes, production set-points and error correction in the appropriate place 
• Should you automate decision engine rules using invalidated data? Why not? Validate the data 
• How vital is the process operator or system analyst and how should they be linked into my system? What process or executions should they validate? 
• Systems can be validated using operators or very tightly defined rules. Systems analysts are very helpful in discovering trends within seemingly non-sensible data sets.



The big plan
Often, the conservative approach is to make a history of the entire plant process either in real time or using time series data. This means creating an MIS, that is, MES without the ‘E’. Data must be raw if possible, not dead-band filtered, capturing all relationships, even the ones you don’t know yet. This may come at a price. It might mean you cannot use a relational database. Embed the process experts in the data repository with advanced reporting tools to truly understand the system’s ethos.

Convert your data sets into information before you apply or develop the rules. Organise the data using structure. It is as important as listening before you start to give the answers.

Consider the requirement for real time or time series data. For manufacturing it is almost always adequate to use real time only. For utilities it’s the opposite. What is the difference between time series and real time data? MES is normally about real time execution. Automation is traditionally about real time only. Real-time leads to a fast and crisp network and architecture. When distributed, infrastructure costs become very significant. Geographically-distributed but managed assets rely on time series tasks. These systems are much more complicated as latency must be considered. To suggest MES in these environments as challenging is an understatement.


Is MES a system or a product?
It can be both, but it is not a product only. MES should be designed and engineered to produce a solution to perfectly-fit your business model. MES is not a single product; it is an environment of tools developed using appropriately-matched technologies. With the MES industry mature in marketing but immature in delivery, it is vital to choose the right bits and pieces. These days it’s not so important to be as careful when choosing HMI or SCADA independently, but it will be so if your choice is to have a fully-integrated HMI and MES experience.


The biggest challenge
Poor quality conventions in the control system lead to a disaster in the information system. For MES this also means a greater-than-optimum engineering effort will be required. So often implementation of an information system highlights process faults never before known, but always there, which can lead to significant rework in the control system components.

Once the solution has been developed and implemented the greatest challenge is the discipline to maintain the integrity of the system. Given that the information system will highlight differences and anything out of the norm detected on a process centric HMI screen, the information system will greatly aid in maintaining instrumentation calibration and integrity and ultimately assist the overall quality factor of the enterprise’s control systems.


The big picture
In a maturing industry with product offerings changing frequently, the first timer (which is almost every MES customer) must avoid Vapour-ware. That’s the pretty pictures produced with savvy graphics designers and a marketing team with imagination. Don’t get me wrong, the graphic designers and the marketing team are not the wolves in sheep’s clothing – you just need to be alert to ask questions about how mature the offering is, where these systems have been implemented before and ask to speak with references. If you are the pilot testers and you are helping to develop first time tools and features, well that’s healthy too because at least you know what you are in for.

Good planning and MIS will yield excellent results. To be strictly analytical about it, the value of MES with the ‘E’ is only realisable or measurable if the information component is provided first and measured and then the ‘E’ is implemented and measured. MES demands discipline and planning from outputs to inputs and then the business rules. The products need to be robust if you intend to implement unattended ‘E’ functionality.

MIS might be the answer. It is for the conservative. MIS is a migration step of MES. MIS provides good data upon which excellent plans can be developed. If you already know your plant intimately then you are probably already ready for MES.