Today, a typical factory resembles a Zoo and ironically gives rise to an Information Jungle.

Here’s what I mean. An average manufacturing unit would have been in business for more than a couple of decades. At that time, the factory acquired the latest machines, or maybe one generation old. With time, as the business expands, the factory keeps doing the same, developing the latest or the latest-minus-one technology. Remember that no machines are getting discarded cause most machines have at least three to four decades of working life. If some of these machines have been bought recently and are of current technology, the factory has machines of the following variety:

• Latest intelligent machines with Robotic Controls. These machines connect directly to a Local Network and can take their instructions directly from a computer and communicate in real-time with the following machine in the assembly line. Of course, this will be possible only if the next machine is similarly enabled. They can also feed the data related to their workload, health, and several other physical parameters. Engineers can monitor all this data remotely and can subject this data to advanced analytics and assess and predict several valuable data points.
• The next category of machines is another intelligent machine that can monitor all the above data and display it locally on the machine. The only communication ports available on the machines have been used to connect it to a controlling computer/PLC and link it to the following machines on the assembly line.
• Then there would be the third variety with a limited level of intelligence. Such a machine would monitor and measure a few parameters but does not have any communication ports to connect with any IT system or any OT device. Such a machine will only display the information on a local screen and expect someone to monitor the parameters by sticking close to the machine. Such machines were designed in a time when the one-time CAPEX involved in procuring these machines was far more significant than the recurring Opex of the cost of labor over many years.
• Then there would be other machines which are old workhorses, capable of performing the machining tasks for many more years to come, but these are relatively dumb. They work, and that’s all they do. They do not monitor and parameters and have no information available to share.

So that’s the Zoo, I alluded to.

But how does it give rise to an Info jungle?

The workers and the supervisors alike must keep track of what a machine is doing and how it is behaving. Modern, intelligent machines can share a lot of data and info with the users/supervisors/management. Older generations of intelligent machines can display some relevant data locally. Still, older machines do not offer any such ready insight. These machines need to be continuously watched to sense current and impending problems.  

So what we have on our hands is a wide variety of data and how it is available. On one end, some machines can feed all the requisite data into a computer network to provide a convenient dashboard and generate alarms as per pre-defined business rules. The data can also be processed and subjected to machine analytics. On the other extreme, some machines only provide minimal visual or audio indicators that often need to be interpreted by someone who may be less than an expert. In between, there is a whole variety of ways in which this data is available with variations in insufficiency, timeliness, and convenience of data capture.

There is a severe lack of uniformity in the availability of operational and health data from a different set of machines, and this mass of data does not lend itself to any useful purpose. And that’s what I mean by a Jungle of Information.

Why is the Info Required

Let’s pause here for a moment to consider as to what is the compelling need for all this data to be made available in a uniform fashion.

Most of the factories are large campuses. There are many machines on the floor and a more significant number of people moving around. In addition, there is a constant movement of raw material across the shop floor machining activity, making a considerable noise level, coupled with workers having to communicate with each other in a loud voice, which adds to the chaos.

A factory manager is responsible for the productivity of the entire factory. He needs to be aware of the OEE value, and in doing so, he needs to be mindful of not only the current operational status of each of the machines but should have a tab on any impending issues. Under the circumstances described above, the manager cannot depend on visual cues from the shop floor. Traditionally, the source of all the info he needs is the workers deployed on the machines, which feed the information to the management thru a hierarchical channel.

The information eventually reaches the management but is prone to be incomplete, inaccurate, subjective, and is almost always late. Furthermore, this information moves on hard copies, making it very difficult for any manager to put it meaningful in improving the OEE on the fly. This info is primarily suitable for only post-facto analysis of incidents. By the time the data reaches a computer system to be analyzed, the incident would have already occurred.  

Sometimes I wonder if this is the real reason they call a database of factory operations data as a Historian!

In the past, manufacturing enterprises were able to live with this situation. The level of competition, the expectations of quality and timely delivery from the customers, and the pressure on margins have all been steadily increasing. So any successful factory today needs to know the following about each of the machines in the factory:

• Is the machine powered on or off?
• Is it in working condition, or does it fault if it is powered on?
• Is it being used?
• Is the rate of production from the machine within the defined tolerances?
• Are all the physical parameters related to machine health within limits?

And more importantly, all this information needs to be available in a real-time sense in a convenient way.  

What Would the Ideal Situation Be

If I were that factory manager, what would I want to see?

• I would somehow want to get all the info related to all the machines on my PC screen, in my office.
• I would want this info to be captured from the machines in real-time and made available to me without any delay of manual intervention
• I would want the info to be displayed on a convenient dashboard with drill-down facilities helping me to reach the root cause in a couple of clicks of my mouse
• I would want relevant parts of all this information to be available in different parts of my factory, displayed on a convenient large screen intelligent display.
• There should a business-rules based systems of alerts, alarms, and escalations related to all aspects of the machine’s working
• I would want all this data to be fed into a database where analytics can be applied and predictions relating to any impending problems.
• Ideally, I would also want the system to capture the production rates in real-time and display an OEE tracker to present OEE trends and compare shifts, seasons, specific individuals, etc.

How Can we Get There

The problem emanates from the zoo that we discussed above. What is needed is:

1. A suitable mechanism to enable sensing of data from the oldest machines
2. A way to capture data from older intelligent machines, so it can be made brought from the Edge to the computer network
3. A platform on which all the data from the various machines can be normalized to compatible structures and means of processing the same
4. A suitably designed dashboard with workflows and drill-down capabilities
5. An intelligent means of directing the information to relevant user interfaces for the concerned person to view the same conveniently  

Well, the above is a clearly defined set of tasks. But is it available for the taking?  

The answer is Yes, and No.  

Going by the very nature of work and the variety of components needed, the above is not available as a cut and dried solution. But the majority of features are available. They need to be stitched together by an experienced service provider. Let us examine the components and their availability.  

The Components of Getting There

Sensors

With the increased awareness of the need, the supply of sensors has been keeping pace. A wide variety of sensors is now readily available, and they come with a spectrum of price tags depending on the accuracy and longevity requirements. Most of the sensors operate on long-life batteries and come with built-in means of mounting on a machine. Many sensors come bundled with means of communication using GPRS, Bluetooth, or even Wifi. We need to keep in mind that while retrofitting, particularly on larger machines, we must carefully choose the specific spot where the sensors would be mounted. There may be a need to mount multiple sensors on one machine in some instances.  

Smart Boxes

Whether retrofitted or OEM supplied, the sensor’s output will often not be compatible with a computer network. This may be because of either signal strength, Voltage levels, or sometimes the sensor data output may even be analog. Therefore, some smart box needs to capture such data from the edge and bring it to a computer network, or “Fog,” often called. The term Fog roughly denotes a cloud close at hand.

A wide variety of such smart boxes are available from various manufacturers worldwide. Depending on the price, the quality and functionality will vary. The costlier boxes would typically be higher quality, handle a more significant number of input sources, and provide A2D conversion or even offer protocol conversion.  

Edge Computing

Often, the sensors come bundled with a provision of carrying out limited processing on the data on edge itself. This enables compressing the amount of data that needs to be transmitted on the computer network. This becomes particularly important where the orchestration application or dashboards are hosted on the cloud.  

Industry Grade Networking Equipment

The harsh manufacturing environment with its peculiarities puts particular demands on both the active and passive components of the computer network.

The active hardware components like routers and switches need to be designed to handle the harsh conditions of heat, vibrations, humidity, corrosive fumes, etc. They have to be beyond managing just TCP/IP and compatible with multiple protocols like MQTT/Modbus/DeviceNet/EtherCat etc. Furthermore, they need intelligence to detect the protocol of input signal and intelligently choose their settings to handle the same. Leading Networking OEM’s like Cisco offers a range of products aligned to such needs.

Similarly, the passive components (cables, connectors, etc.) must withstand the harsh environment. At the same time, they need to offer a much higher level of flexibility to fit into the cramped available space inside the plant. Often, the conditions dictate the cables to run along with the plant’s ceiling, and thus it even calls for particular methods and equipment to install the wires. And while installing these cables is such a big deal, it also makes sense that such lines come with a large number of spare pairs so that backup connectivity is readily available, obviating the need for an expensive installation exercise.  

Software Orchestration Platform

A few small-time entry-level orchestration platforms are available, with limited functionality and relatively restricted compatibility with edge devices. At the other end of the spectrum, robust & comprehensive platforms like Cisco’s Kinetic are available with a wider variety of functionality in accepting, normalizing, processing, presenting, and storing the data. As with anything else, the price points vary with the functionality of these applications.  

Dashboards and Multi-Channel alert Mechanism

To my mind, this is the proverbial proof of the pudding. These dashboards are the whole reason why the entire exercise is being carried out, and thus they must fulfill the exact purpose of the users. So while some limited dash-boarding functions are available built-in with some orchestration platforms, I would recommend a custom parameterized dashboard to meet the specific needs as perceived by the workers, the supervisors, managers, and the lead levels in the organization. The dashboards should also provide intelligent facilities to customize the display for several user categories and define business rules for alerts and escalations.

Building such dashboards using available platforms is not a very big deal, and they can be made in a short time without burning a hole in your pocket.  

Software Analytics

Based on the best practices, past data, and analysis, Analytics for machine environments are evolving. Many of the Analytics solutions offer self-learning facilities. These analytics come with a small set of functionalities built-in, and they continue to learn and develop as they are used. With the widespread use of analytic engines, as the ideal behavior for any given machine gets understood and documented, one would see the emergence of a more significant number of Analytics engines able to provide insights into a machine’s health off-the-shelf.

Most major software players offer some analytic engine with a finite set of built-in capabilities.  

Display Systems

Intelligent display systems with suitably sized screens and content management functionality to contextual display data from a centralized database are available. Such systems often provide split-screen features with part static and part dynamic displays. Such displays serve a handy function of the shop floor by presenting the standard version of the truth to all involved in the production process.  

System Integration

Last but not least, you need the support of an able system integrator who will

• Carry out the task of making all these components work together with each other for you
• Set up/enhance your computer network.
• Bring IT and OT to work together in an IoT fashion
• Integrate the smart boxes and the digital displays
• Ensure that the requisite hardware and software platforms for running the orchestration application are integrated with the entire setup
• Integrating the system with one or more cloud providers if you wish to store or process some part of your data on the cloud
• Provide project management and hand-holding support
• Provide software development and enhancement support.

In conclusion, the task of managing the Information Jungle, while being non-trivial, is certainly feasible. Most of the components required to put together the solution are available off-the-shelf. All you need to know is where you stand, where you want to be, and in what kind of time frame you want the entire set-up in place. Beyond this, you need to select the components carefully and have a worthy partner helping you with the implementation.

‍