Finished out 2016 on a relaxing vacation with my favorite people, read lots more books and enjoyed life.
Collecting data is easy.
Analyzing data is fun.
Preparing collected data for analysis is difficult.
Recently I started moving my websites off of Google Analytics. Instead of using Google, I’ve setup my own web traffic analytics. Collecting traffic data and links was easy. It took about an hour to set up.
Analyzing data is easy as well. There are a couple amazing projects that handle large amounts of data. I’m using Kibana from elastic.co. Kibana has a fun interface for creating all sorts of amazing looking graphs.
The difficult part is transforming the log files from my web server and loading them into Elasticsearch, the backend for Kibana.
The data is stored in individually compressed files created once an hour from each of my webservers. The log files are not consistent. The format for each log entry is roughly the same. Data might be appended or omitted depending on the information available at the time of the log event.
I need a complex piece of software to uncompress new log files, look for duplicate entries, combine files from each server into one file, and then load that file into Elasticsearch. Most of the time a Data Scientist isn’t doing data analysis. Most of the time is spent transforming data from one format to another.
The dashboard light on my car came on last month. Along with the light a message was displayed each time I turned on the car. The brake pad monitor was telling me the brake pads had worn thin. This type of predictive maintenance was built in when the car was new.
Most cars can’t tell when the brake pads are thin. It requires a special type of brake pad with a built in sensor. In order for an old car to take advantage of predictive maintenance it would need to be fitted with custom brake pads that include a sensor. Then a computer and wiring would need to be added to monitor the new sensor.
Marketing materials sell predictive maintenance as a magical device set on an old machine that automatically knows when to alert of maintenance issues. I imagine it to be like tossing an Amazon Echo into the car’s glove box and expecting it to report when the brake pads are thin.
Predictive Maintenance is not that simple.
Instead of adding sensors and computers, most car owners would benefit from taking their car in for inspection every 10,000 miles or 6 months. The mechanic checks 140 or so things and one of them is the thickness of each brake pad. Fancy IIoT is not required most of the time. Most of the time what is required is a clipboard.
IIoT is nothing until good daily practices are in place at the factory.
Like a writer staring at a blank page, engineers are overwhelmed when starting a new IIoT project.
What is important to measure?
How will we measure it?
How often will we measure it?
How long will we keep the measurements?
Where will we keep the measurements?
How will we display the measurements?
How will we query the measurements?
How much disk storage will we need?
Who will need access to the measurements?
How will we power the measurement devices?
How will we network the measurement devices?
And so many more questions
When I was a machine builder the best insights came from manually assembling the product. I’d gather all the materials and by hand I’d put the product together. Each step in the process would require unique movements. The process of handcrafting an eventually mass produced product taught me the subtleties that need to become part of the final machine.
The same is true with IIoT data. Manually querying and graphing the data is important for learning the shape of the data. Manual queries reveal the important variables. It becomes quickly apparent when a variable is missing or under sampled.
Before a factory can benefit from Machine Learning and Artificial Intelligence it needs to see its data. The shape of the data needs to be known. The type of ML or AI will be determined by the nature of the data. The nature of the data is best determined by querying and graphing by hand.
Seriously. Clipboards are the most advanced technology a factory can utilize.
Start becoming aware. To relate this to woo woo psychology, clipboards are mindfulness for factories. The process of recording numbers every hour creates a culture of knowing and measuring. It is the first step to true IIoT.
If a factory doesn’t know its numbers on a hourly clipboard level then adding IIoT will not be an improvement. IIoT is just as much about culture as it is about technology. IIoT requires a culture of measuring and improving. It is the next rung of the ladder.
The first ladder rung is clipboards and hourly records. The second rung is automated measurements. The third rung is automated corrections. The 3rd rung is where artificial intelligence and machine learning start to shine.
Get out the clipboards and start a culture of measuring everything.
Industrial Internet of Things is complicated. There are so many niches and opportunities. The only way to start receiving the benefit of IIoT is to just start. Pick a machine and add telemetry to it. Any vendor or system is fine. Just start. The telemetry data will reveal the opportunity.
Start with one machine.
Every factory needs a resident software developer. Information technology is moving rapidly. There are many small wins that an onsight developer can quickly capture. In some cases an opportunity can be identified and the ROI capture in under an hour. Most of these opportunities would be missed. The inertia of writing a spec, determining the scope of work and collecting bids would hinder the short-term benefit. An onsight developer can identify and implement in one sitting.
Business is tough. Once you get a customer you want to keep a customer. It turns out that a consistent way to keep corporate customers is to utilize inertial.
I want to go home early in the day. I don’t want to learn a new thing, debug a new thing, or install a new thing. I want to be as efficient as possible in my own job. I want my personal work to be the best it can be.
My goal of being the best I can is often in opposition to the corporation being the best it can. If a new technology comes along that will make the corporation more money, I personally have no incentive to install it, learn it and use it. I personally am doing great with the old tech. I personally am getting rewards and recognition for the old tech.
The new tech would destroy the expertise I have built. I would no longer be the go to person. I would lose my status in the company. Personally I have no incentive to adopt a new tech. It just so happens that I am also the best person to introduce and adopt the new tech.
Here’s the problem: the people who are the best to introduce a new tech are also the least incentivised to introduce that new tech. They are also in the position of most authority to reject new tech proposed by others.
What to do about this ROI depressing scenario?
Mandates from above could be issued. That assumes the CTO knows the minutia about the tech. The CTO has more important opportunities to pursue.
The CEO could introduce a culture of innovation. The problem there is innovation is vague and difficult to measure. This creates loads of tension and angst as people try to be more innovative.
Grass roots engineers could sneak the tech in. Small projects and low visible solutions could introduce the ROI improving technology. Stay out of site of the old guard. Then present the results to the stakeholders. Little by little these new technologies would replace the old inefficient tech stack.
I read a disturbing summary of a new IIoT product. The old world company planned to tack on IIoT capability to its flagship DCS product.
Don’t do that!
DCS has an important job to do and cannot be interrupted by IIoT. DCS data is predictable and secure. IIoT is mutable and shared with front office managers whose PCs are connected to the internet at large. Bridging DCS and office computers is a bad idea.
IIoT is easy after the marketing fantasies are cleared out of the way. Poor marketers get a lot of heat from… well everyone. This is especially true with IIoT where the topic is easily abused to get eyeballs to read a new website. In the process the core value of IIoT has been muddled.
The core of IIoT is simply to make factories better. The fantasies about artificial intelligence and machine learning are a small piece of the IIoT pie. These fancy technologies are not needed for a factory to realize immediate gains. The core connectivity of IIoT will immediately benefit every factory.
The core of IIoT is simple and immediately valuable.
IIoT should be forgotten. I should be able to setup my IIoT solution then forget it.
No fancy animations.
No daily emails.
No lights and flash.
IIoT should do its job. It should run and report. When I need it I should be able to immediately get only what I need. Then, move on to my next task.
Articles about the Industrial Internet of Things are filled with exciting technology and opportunities. I enjoy reading these articles. The articles often miss the true value of IIoT. The value of IIoT is boring. The technology that delivers the most value for IIoT is boring. There isn’t a science fiction related story or a moment of discovery.
The value of IIoT is simply seeing better what is already happening in your factory. IIoT is an incremental improvement. Instead of reading weekly reports about productivity, you can read hourly reports or even watch in real-time. There isn’t an amazing process improvement. There is the ability to make faster decisions with more complete information.
IIoT is simply another tool.
The Industrial Internet of Things has attracted lots of attention. Promoters have taken advantage of the attention with loosely related products and ideas. This has caused the scope of IIoT to balloon and dilute. IIoT now includes everything from Machine Learning, Artificial Intelligence, Lights Out Factories on one end; to strip chart recorders, OEE, Six Sigma, and MES on the other.
IoT is a dizzying spectrum of ideas and technologies. I’ve chosen to focus on one small piece of the IIoT topic. I’m focusing on the enabling technology of data acquisition. Machine learning, AI and Process Improvements depend on quality data acquisition. The physical network requirements for collecting data also enable controlling industrial processes.
There is a trend in the Internet of Things space that favors the easily achievable. Companies like SmartThings, Nest and Particle solve the easy problems. These problems are not valuable. Wall Street doesn’t think Nest is doing well. I think it is because Nest solved an easy problem and not a valuable problem.
We are at a point where hardware is cheap and easily produced. Advances in custom manufacturing make it possible to dream up a design in a weekend and ship it in a couple months. We have a deluge of half-baked devices. The long term value to the consumer isn’t present.
The Industrial Internet of Things is in a position to improve valuable industrial processes. Factories become more efficient when they collect and react to machine production data. IIoT promises to make production data quickly available. The flow of the factory improves when the upstream machines match the production rates of their downstream machines. Production waste and warehouse space are saved.
IIoT doesn’t need to be revolutionary or advanced. IIoT can ignore the buzz about Artificial Intelligence and still improve the bottom line. The core IIoT functionality is enough to make significant improvements to factories. We don’t have to wait for the quantum leap that the consumer IoT market is waiting for. Factories benefit from simply “turning the lights” on their production data. In the industrial internet of things even the half-baked ideas are poised to improve ROI.
Security is your top priority. All internet connected devices will eventually be compromised. If you are lucky the compromised device will be someone else’s and you will have time to patch your own. IIoT solutions need to be designed knowing that they will eventually be compromised. How will the system be made safe? How will the impact be contained? How will the system be patched?
The internet portion of IIoT is a bit misleading. In industrial settings we don’t typically connect our networks to the actual internet. Instead the internet is an intranet. An intranet is only available to people inside the company.
Networking industrial equipment and machines is a tough job. Industrial environments are electrically noisy, dirty, too hot, too cold, too wet, or too dry. Industrial environments are at the extremes of what typical electronics are designed for. The harsh environments require specialized electronics, enclosures and cables.
The industrial internet is just like the world wide web internet. The cables and equipment are more robust, but the core functionality is the same. The same architectures and software run the internet for your mobile phone and your factory. Counter intuitively the best people to design the internet side of IIoT are the same people who design websites and mobile phone games.
The industrial internet of things lives at the intersection between machine builders and web developers. It is an odd cross section of old industrial knowledge and modern web knowledge.
The industrial knowledge is required to know what to measure and how to measure it. Great machine operators have a sixth sense about the performance of their machine. They know the sounds and sights that indicate what actions they need to take to bring the machine to peak efficiency.
The expert industrial knowledge is required to know what sensors to use and how to install them. Industrial machines live within a unique ecosystem of standards and best practices. It isn’t feasible to slap on a hobby kit and start taking measurements. Solutions need to be built for harsh environments that include vibrations, moisture, temperature swings, high voltages, noisy RF bands and rough handling.
The production rate of every machine in the factory should be measured and recorded in one database. This is easily accomplished with 10 year old technology. Most machines can be instrumented for a couple hundred dollars. The database can be run from a leftover PC. This simple setup is a great proof of concept to get additional funding approved. The stakeholders will be impressed with the beautiful charts in Kibana.
IIoT is simple (and can be cheap… shhh, don’t tell Allen-Bradley, Siemens or Inductive Automation.) Yes, the Industrial Internet of Things is simple. Measure, analyze, and react.
Measure - Many machines don’t record simple metrics like widgets per hour. Put a photoelectric sensor at the end of the conveyor belt and start counting.
Analyze - Load Elasticsearch and Kibana onto an old laptop.
React - Look at the Kibana dashboard and make some decisions based on what you see
The industrial internet of things is an old idea. It has been growing since clipboards were used to record productivity. The manual process of measuring, analyzing and adjusting is the basis of the Industrial Internet of Things. A well organized factory records key metrics on a monthly, weekly or even daily schedule. The key metrics are analyzed over many months. Trends and opportunities are discovered in the analyzed data.
The processes that are used today to optimize ROI are the same processes that are being used in the Industrial Internet of Things. IIoT makes the processes faster and more efficient. If your factory doesn’t have a manual process then you need that first. Before you can get value from IIoT you need to get value from a good process.
Sarcastic Answer: It’s marketing hype.
My Answer: The Industrial Internet of Things is the belief that communication between machines improves the ROI of a factory.
The Industrial Internet of Things is a belief. The belief in IIoT takes many forms. Articles about IIoT share a vision of the future. The industrial internet of things is permission to dream about what could be possible. It is a platform for building fantastic visions for the factory of the future that we can build today.
The next 30 days are dedicated to writing about the industrial internet of things. I’m sharing my experiences in manufacturing, machine building and test & measurement. I’ve built machines for giant fortune 50s and tiny mom and pop shops that make window treatments. Every project had some sort of connectivity to a computer, the industrial internet.
The rules for writing about the Industrial Internet of Things for these 30 days are simple: text only. No pictures, no diagrams, nothing outside of standard markdown. This will be an exercise in clearly articulating my technical knowledge in English. Starting in January I’ll share code, diagrams and video. During the month of December I’m limited to text.