Digital Transformation - A Better Way

Plants successful with digital transformation have not done it the way it was first thought it should be done. They’ve found better ways. We explored some ways they do it differently in the earlier Think Big, Start Small, Scale Fast essay. Here are a few other important practices in plants successful with digital transformation that must be mentioned. Here are my additional personal observations:

Digitalization False Start

An April 2019 article from McKinsey suggests digital transformation (DX) should start delivering impact within three months or else it will stall. From what I personally can tell, deploying an additional “single software platform” on top of your historian and ERP system could take years and thus it will also take years to see returns on the investment. Hence layering yet another platform or “data lake” on top of others is not a good approach. Don’t let digital transformation (Industrie 4.0) become yet another “single software platform” layer on top of the other layers. Successful plants instead use their existing historian as the platform, simply extending it with a database for reliability data like vibration spectrums etc. for asset performance. These plants can start immediately, digitally networking sensors, and plug in software apps as needed, solving one problem at a time, with fast return on investment. Yet, they think of the big picture, using common infrastructure based on standards like WirelessHART, FOUNDATION fieldbus, and OPC-UA to accommodate a broad spectrum of applications with a common Digital Operational Infrastructure (DOI). Energy savings solutions deliver impact immediately. Reliability solutions start predicting failure immediately but it may take a while before you see reduced cost and downtime thanks to fewer surprise failures. Production process solutions create situational awareness immediately but it may take some time before reduced off-spec product, or travel cost and feedstock cost reduction becomes apparent. Safety solutions too prove themselves over time as reduced incidents. Solutions in these four domains are the backbone of digital transformation for operational excellence and top quartile performance. For many other solutions like tablet and smartphones, drones, and crawlers where the savings are harder to quantify so these should not be the core of the digitalization effort.

Some plants start digital transformation (Industry 4.0) with the wrong projects; where the difference between the old way and new way of working may not be that big. For instance, crawlers, drones, as well as tablet computers and smartphones may not always be the best examples of digital transformation – because they still need a human being to operate them. A human must control the crawler or drone. A human must carry the tablet or smartphone, inspect, and type the data into the device. It is still manual. Not fully automatic. There are digital transformation solutions which are fully automatic: using a permanent sensor to automate the data collection, digital data communication, and software to automatically interpret the data. For this reason, it may be better to put drones and robots in the later part of the digital transformation roadmap.

Dashboards and Augmented Reality (AR) are displays for humans, which is the last step in the data-information-knowledge-wisdom chain. Before a useful dashboard or AR overlay can be created, you need the right analytics to provide the information to display in the dashboard or AR device. And before you can get reliable analytics you need useful data. Therefore plants must start with sensors. If not, the dashboards, on desktops, tablets, or smartphones, will not have the useful and reliable information the personnel need to do their job better, and it if doesn’t they won’t use it.

Plants successful with digital transformation have started with digitally networked sensors; fieldbus or wireless.

Some consultants propose “analyze 10 years of data for 3 months to see what correlations we find and insights we uncover”. Such an approach not involving the ultimate end-users in the plant has a risk distracting digitalization towards problems that don’t exist. But, if a solution doesn’t solve a real problem, not making people’s work easier, they will not use it. There is a chance the solutions fall into disuse and transformation stalls.

Instead, plants successful with digital transformation use another approach. Plant personnel know what the problems are because it causes them pain; disruption, consuming their time, and resources. These challenges are easily uncovered in a discovery workshop with the ultimate end-users in the plant. Start by solving these problems. This will make people’s jobs easier. This ensures they use the new digital tools.

Digital Transformation Success Metrics

Digital transformation is often justified based on Return on Investment (ROI). The ROI may often have to be proven afterwards to declare success or if government grants are used. Track ‘before’ and ‘after’: maintenance cost, downtime, energy cost, flaring, incidents, product yield, etc. when working on investment justification. Spend some time thinking about how savings and other improvements will be measured and demonstrated. Do you already have a baseline cost or other metrics from your current way of operating?

Reduction in energy consumption such as thanks to steam trap and heat exchanger monitoring is easy to confirm as reduced utility consumption per ton of product produced or processed. If the solution is only implemented in a single unit or area of a very large industrial complex it helps if utilities like water, air, gas, electricity, and steam (WAGES) are measured at that area or unit level to see the real difference. Indeed area-wise and unit-wise utilities consumption measurement itself for ISO 50001 energy management is often part of digital transformation.

Reduction in material losses such as thanks to relief valve (PRV) monitoring is easy to confirm from the mass-balance between feedstock and product, or seen directly as reduced flaring per ton of product produced or processed if flow to flares is measured.

Reduction in surprise equipment failures such as thanks to condition monitoring can be seen by looking at maintenance records. The cost avoidance can be estimated based on reduced incident count and typical repair cost. Unscheduled downtime and opportunity cost for this downtime can also be tracked.

Safety incidents are also tracked so HS&E records can be used to demonstrate incident reduction; such as thanks manual valve monitoring, safety shower monitoring, and leak detection etc.

Production metrics like amount of off-spec product produced is usually recorded and can be used to demonstrate a reduction.

Reduced operating cost like transportation such as from fewer travels offshore or to other remote sites can be seen from financial records.

However, results from deploying tablet and smartphones with the associated supporting infrastructure could be hard to measure.

I personally do not believe in measuring system logins and usage time as a KPI for project success, as personnel may login just to meet the metrics. The metrics must be business outcomes like those above.

Sustaining Digital Transformation

Getting input on plant challenges to solve through digital transformation from all operational departments is a best practice. However, remember that when you form a temporary ad-hoc digital team pulling in personnel from various departments on loan for a digital transformation programme they will eventually have to go back to their regular job. However, digital transformation is forever. Therefore, digital transformation needs a long-term owner responsible for the automation and continual improvement.

Digital transformation is not a project, it is a lifestyle

From what I have personally seen, the individuals charged with running the digital initiative should be the I&C team for the Digital Operational Infrastructure (DOI) which is the major part, and the IT team for the ERP system part. The I&C team works with the operational departments like maintenance, reliability, integrity, production, quality, HS&E, and process/energy etc. to help them drive DX projects in each domain. The I&C team will own and support the DOI perpetually just like the DCS and historian. Make sure the I&C team is clear on this extended long-term support role from the start of the digitalization. They work with the other domains on individual projects: steam traps, pumps, safety showers, manual valves etc. but the I&C team is the perpetual owner.

If plant personnel don’t change their traditional ways of working it is because the new digital tools are not helping them. If the new digital tools are working the way they should, they will be used - and there would even be dismay if the tool was removed forcing users to go back. Similarly, if the tools are truly helpful, use will be self-sustaining. Personnel will not go back to their old ways.

Plants successful with digital transformation have deployed tools which truly help, are easy to use, with little effort. For instance, automating manual data collection by deploying sensors so personnel don’t have to go to the field with portable testers or read gauges is very successful. Once you have deployed sensors and display data, or information derived from the data, at people’s desk, they will not choose to go out to look instead. If you only display the data in an inconvenient location like a control room you have not solved their problem so it will not be used, and adoption will stall. Therefore, successful plants now ensure the information reach their people’s desktop, tablet, or smartphone.

Similarly, if you try to do equipment analytics using only existing process data you will not get early warnings and reliable results. Personnel will soon distrust the solution and go back to their old ways of working. Therefore plants use direct sensing on equipment to get more predictive and reliable analytics to ensure continued use. Plant floor engineers have a good sense of which sensors will be required, and it can also be found in documents like FMEA, fault tree, or fault matrix.

Management and line leaders communicating the importance of digital transformation is a must. However, for the change to sustain it must be for the better for the ultimate end-user in every way. It can’t add the burden of entering data in a tablet on the plant floor for somebody in the admin building to get a dashboard. Data must be collected automatically so there is no manual data entry; and it must benefit not just those in the admin building but also those on the plant floor. Personnel will go back to their old ways if the solution is not helping them, and maybe even resign if it becomes a burden. Custom programmed software which cannot be finished off with all the required features and a user-friendly interface due to the high cost of making iterations of changes and additions will lead to frustration. Analytics which give unpredictable results because the model created with machine learning cannot be verified, and not having the right sensors will result in the solution not being successful causes people to return to their old ways of working. If tools are not useful or too difficult to use they will not be used.

Plants successful with digital transformation instead use readymade software, with engineered analytics based on robust and verifiable first principles (1P) and FMEA, and direct sensors. When the new tools are truly useful they will be used.

Everybody wants the information, nobody wants to enter the data

Digital Transformation Culture

Digital transformation starts by conducting an operational excellence discovery session involving all the operational departments in the plant. The outcome of the discovery session is a report with highly actionable solutions to real plant problems. Most of these solutions are already tried and tested in other similar plants so there is no long development periods, testing, and proof of concept trials. Tried solutions can be implement immediately without delays and see returns quickly. Soon thereafter the solutions can then scale across the plant. Even the infrastructure is scalable; unit, area, to the entire plant and do not require massive infrastructure up front. With a few quantifiable successes under the belt such as reduced steam loss and product flaring, you can next do applications for which it is harder to quantify benefits. Some new solutions may have to be created.

Apart from uncovering pains that can be solved through digital transformation, the discovery session also teaches personnel to think in a new way; a new way of problem solving. Going forward they will innovate with sensors and software as a digital solution when they encounter problems in operations. When plant personnel think digital this way, that’s when you know your culture has changed.

No Data Replication

If you start digital transformation embarking on implementation of a “single software platform” or “data lake” you may spend months or years and millions of dollars before you get to deploy the first actual use-case solution that can make a difference. There is a risk digital transformation will stall and be abandoned before it completes.

Plants successful with digital transformation instead use their existing process historian and extend it with a reliability historian. This way data is not replicated in additional places. A distributed architecture means not replicating the data from various systems databases in a multi-million dollar “single software platform” central database but instead using OPC-UA as a “virtual platform” to access the data directly from each source database without replication. A modern distributed data architecture.

Overcoming Resistance

Some vendors propose changes to IT infrastructure like adding another “single software platform” above the ERP which can be a struggle because these platforms cost millions of dollars, takes years to implement, and duplicate the functionality of the existing historian adding little or no new functionality. Moreover, it would increase the IT annual support cost and burden. So making such IT infrastructure changes can be a struggle. Since most plants already have a single software platform; their historian, and plants have invested millions in their plant historian, and many millions in their enterprise historian at the global level, they don’t want to have to support one more. And because such a big project requires many staff and a lot of money, plants find themselves with too few technical staff and unforeseen demands on budgets.

Plants successful with digital transformation instead use their existing historian as their platform avoiding the IT infrastructure struggle. This way plants also avoid conflicts with legacy support requirements; by using what they already have, in order to move the digital solutions into full production with scalable access to data from across the enterprise.

Readymade Solutions not Custom Programming

This may be counterintuitive at first, but once you analyze it you will understand the logic: custom made solutions may stall. When a system integrator (SI) writes custom software apps it is new and unproven, so there will invariably be shortcomings in the functionality because the SI and users in the plant cannot think of and specify all the features required and desired in advance. Therefore the plant will spend many months, and maybe even years, with programming and testing multiple versions of the software before it becomes adequate. Shortcomings in analytics prediction accuracy can be expected from a new algorithm or model because it is not finetuned since it takes testing in actual plants in varying operating conditions to eliminate false positives and to not miss developing problems. Lastly, shortcomings in usability is common in custom software because nobody has used the custom-made software before, so there is no feedback from prior users which has gone into the product. Therefore, custom made is usually not better than readymade off-the-shelf software.

Plants successful with digital transformation use readymade software with a proven track record. Perfected over time, full featured, finetuned algorithm, and intuitive. This way there is also no need to contract programmers.

Nobody will attempt to write their own ‘better’ word processor - because readymade software is lower cost, easier, and more feature rich

Note that even without custom software programming, ‘agile’ methodology can still be applied in evolutionary development of digital transformation solutions; not at the low level rewriting and testing software code, but instead at a higher level; deploying and testing another analytics app, adding a sensor if needed, change the default settings etc. You can deploy the basic setup early, and then continually improve. Configuration options in the software provides flexibility to quickly respond to requests for changes.

Seamless Data Access

Proprietary Application Programming Interfaces (API) and web services associated with a “single software platform” is one reason why such projects stall. You must custom program a driver for every data source and every app, then maintain that software driver with constant upgrades for the life of the system. It is not a practical way to make data accessible. Too much dependency on the system integrator (SI).

Instead, plants successful with digital transformation use standard OPC-UA for data accessibility.

The Right Data for Digital Transformation

Another reason for not being successful could be reliance on existing data; without adding the necessary sensors for direct measurement of leading indicators on problem areas. You cannot solve equipment problems using only the process data plants have. Process data is not a reliable indicator of equipment problems so you can’t reliably predict equipment problems with process data.

Plants successful with digital transformation use direct sensing to get reliable data. In the new digital way of working in the plant, with a high level agile methodology to problem solving, data is continuously enriched by adding sensors for direct sensing of the required variables for each application. The most practical way to do it is using digitally networked sensors, wireless or fieldbus, preferably non-intrusive.

Digital Transformation Ease of Use

The effort and expertise required to develop accurate machine-learning models plus the effort of first cleaning up and classifying years of free-text maintenance records made by different people described in different ways makes machine learning algorithm training hard as described in a October 2018 McKinsey article. The effectiveness of such systems can be undermined by poor housekeeping such as incomplete or inaccurate maintenance records because when training the algorithm it may find the wrong correlation or miss a pattern resulting in nuisance alarms or no alarm even though failure symptoms are present. Because machine learning is based on data science, it may require the plant to hire data scientists to setup and maintain the system adding to the cost. Only alarming “anomaly” is not so actionable as it doesn’t describe what the problems is.

Plants successful with digital transformation instead deploy direct sensors and engineered analytics based on first principles (1P) and FMEA. When data science is used, it is important the resulting algorithm is verifiable as a ‘sanity check’ to make sure the variable used and rules makes sense. The most interesting fact is that in many cases you don’t even need any analytics – just a sensor and an alarm threshold as also explained in the October 2018 McKinsey article. While training is important, equally important is ease of use. Reliability engineers, maintenance technicians, and other operations personnel are not programmers and data scientists – so a key is to deploy easy to use readymade software with engineered analytics based on familiar and verifiable 1P+FMEA with a simple user interface that doesn’t require extensive training. There is often a concern about skill sets and competencies; with complex solutions that would indeed be a problem, but with the right software plant personnel can easily learn.

Enable Digital Workflow

Once analytics predicts an issue with a piece of equipment, such as predicting bearing failure in a pump, action must be taken. Maintenance engineers may not sit in front of their computers all the time, so for analytics (diagnostics) to be successful, notifications can be sent to their smartphone wherever they may be, and a workorder ticket issued to the CMMS/ERP system. The maintenance planner accepts and releases as a workorder which the maintenance technician in the field receives on their tablet through CMMS/ERP web browser interface. After finishing the job the technician closes the workorder on the tablet. These are the new digital ways of working: new digital Standard Operating Procedures (SOPs). Therefore, the predictive equipment analytics at level 3 (L3) must connect to the CMMS or ERP system at level 4 (L4) of the enterprise architecture ISA95/Purdue reference model.

There is no need to merge the IT and I&C (“OT”) departments to integrate the analytics with the CMMS/ERP, they just need to collaborate. This is made easier when the integration between Digital Operational Infrastructure (DOI) and the ERP is made through a standard off-the-shelf software rather than through custom programming. That is, IT and I&C meet at level 3.5. Merging the IT and I&C department against their will would not be a good idea.

IIoT Cloud or On-Prem / Edge

Cloud is not a requirement for digital transformation. Cloud integration requires connection to the Internet which in turn requires cybersecurity experts to be hired if you wish to connect the core process control system. The required security studies and network architecture hardening required could take months to complete. Therefore, connecting the core process control system to the cloud may not be a good way to start digital transformation as it can take long before the results are seen.

Most plants do not use cloud for digital transformation. 99% of plants instead do digital transformation on premises (“on-prem”) within the plant perimeter, without internet connection to the cloud. This allows the plant to get started with digital transformation very quickly and to see results soon after. Therefore, on-premises deployment is a better way to start.

Having said that, cloud connection enables Industrial Internet of Things (IIoT)-based Connected Services; where a Knowledge Service Provider (KSP) has a pool of domain experts in areas like rotating equipment vibration, control valves, and process analyzers etc. – any area of sought-after expertise - even small little steam traps and relief valves. Such services can be extremely valuable to sites which do not have sufficient inhouse expertise in these domains. The service provider does not run the process, but help ensure the equipment in the plant is healthy. This might be the greatest value of cloud connection.

For certain IIoT applications cloud connection can be made a lot easier by not connecting the core process control (CPC) system to the cloud. That is, the digital operational infrastructure (DOI) for Monitoring and Optimization (M+O) is completely independent.

This approach is not possible for every IIoT application, but is possible for quite a few, because there are many applications that do not require any process data at all. For example:

• Vibration of rotating equipment (compressor, turbine, pump, fan/blower, cooling tower, air cooled heat exchanger, etc.)
• Corrosion
• Erosion
• Steam traps
• Relief valves (PRV)

In these cases, the add-in sensors can be digitally networked through an edge gateway and on to the cloud without passing the core process control (CPC); without any data connection to the DCS or historian. That is, since the DOI is independent from the CPC, it does not add any security risk to the CPC even though the DOI is connected to the cloud.

Digitalization Protocols

Control engineers, instrument technicians, and other people in the organization are not trained on message-oriented middleware protocols like MQTT/AMQP/CoAP/XMPP unfamiliar to the plant environment. The tools that people in the plant use, such as handheld field communicators, laptops with modems, interfaces, and Intelligent Device Management (IDM) software do not support these protocols. Moreover, MQTT, AMQP, CoAP, and XMPP do not provide semantic interoperability, so there is no single tool that supports devices using these protocols from multiple vendors; that is, you will have to grapple with a plethora of tools; one for every vendor which is not feasible in a plant.

Plants successful with digital transformation use familiar WirelessHART, Modbus/RTU, FOUNDATION Fieldbus, and PROFIBUS-DP protocols and their IP version: HART-IP, Modbus/TCP, FF-HSE, and PROFINET-IO etc. These are the standard protocols which people in the plant are already familiar with. The plant already has the tools such as handheld field communicators, laptops with modems, interfaces, and Intelligent Device Management (IDM) software for these protocols. In short, the organization has the capabilities for these protocols. It is because these protocols (except Modbus) provide semantic interoperability it is possible to use a single tool for devices from multiple vendors.

There is a misconception that cloud connection requires MQTT or AMQP, but this is not the case. The IP version of fieldbus protocols like HART-IP, Modbus/TCP, FF-HSE, and PROFINET-IO etc. can run across the Internet straight into the cloud. There is no need for conversion to MQTT or AMQP. Moreover, conversion to MQTT or AMQP means valuable metadata and semantic interoperability would be lost. As a rule, do not convert protocols until the very end of the chain.

Full Throttle Digitization

Plants that have been successful with digital transformation have not done it the way it was first thought it would be done. They did it without hiring data scientists, without forcing IT and I&C departments to merge, without deploying yet another platform layer, without custom programming apps, without MQTT/AMQP etc., without custom programming API and web services, without machine learning, without cloud, and some of them even without process data. Make sure to also read the earlier Think Big, Start Small, Scale Fast essay if you have not already done so. So digital transformation can start sooner and move faster than originally thought. Well, that’s my personal opinion. If you are interested in digital transformation in the process industries click “Follow” by my photo to not miss future updates. Click “Like” if you found this useful to you and “Share” it with others if you think it would be useful to them.