Industrial digital transformation has its beginnings in the Industry 4.0 initiative, unveiled in Germany in 2013. Since then, industrial digital transformation has morphed into an imperative that underpins business vitality. However, according to the 2020 Industrie 4.0 Maturity Index in Industry[1], released by the German Academy of Science and Engineering (Acatech), more than 90% businesses are still only in early stages of industrial digital transformation. Most firms are still grappling with how to record and aggregate data generated by their equipment, systems, and workforce. This is a far cry from what decision makers at most businesses initially hoped for. Their vision is having big data analysis results—completed and readily-available—right on their screens so that business insights can be implemented to lower cost, increase efficiency, or drive business model innovation. If truth be told, many companies still have a long way to go regarding their industrial digital transformation.
This begs the question why are companies still behind in the industrial digital transformation race? One of the main reasons is that big data analytics, artificial intelligence (AI), and other disruptive innovations surrounding digital transformation, all of which have become buzzwords in recent years, are only implemented in the later stages of digital transformation. Why add these innovative solutions so late in the game? If you don’t obtain enough data in the early stages, then even the smartest AI or machine learning solution will have little value for you. As we know, In the industrial digital transformation landscape, data comes mostly from operational technology (OT) environments, for example, a drilling well in the middle of a desert baking at temperatures of 40 to 50°C, an oil pipeline system stretching for hundreds of kilometers in a freezing area, or a transportation system of a fast-moving and vibrating train. It does not require a huge stretch of one’s imagination to foresee how difficult it would be to capture data from these harsh environments. Hence, to kick-start a transformation initiative first requires a failproof strategy on how to accurately capture OT data from industrial automation equipment.
Furthermore, this issue requires real deep thinking. Against the backdrop of industrial digital transformation, OT data has shifted from monitoring-oriented to optimization-minded, looking not just at the present but also to the future. Errors in data collected from its sources may lead to defects in subsequent analyses. Therefore, the focus on just capturing “stable data” no longer suffices. It is safe to say, “quality data” will become the deciding factor f of any transformation program’s success. With three decades in connecting OT data, we at Moxa, a trusted OT data technology provider, identified four pillars that underpin data quality.
Insufficient Data Due to Data Silos
One of the challenges to data quality is insufficient data. This is mostly because automation systems were not designed for data analysis. Even in cases of data transmissions on a shopfloor, data is tapped to support control equipment operations only, which is not enough by any measure for distilling business insights. For example, a factory may have a bottleneck machine on its production line through which everything gets made or processed. If the machine goes down, the entire line shuts down. To minimize downtime, the plant needs to predict which key components inside the machine could fail and purchase the replacement components in advance. However, these devices seldom provide data regarding their key parts and components. Therefore, it is necessary to install sensors in them and convert the generated analog signals to digital ones via remote I/Os. The digital signals can then be sent to servers in the upper layer or to the cloud to enable predictive maintenance. This demonstrates the capability of OT Data Acquisition.
In this scenario, only one machine needs to be worked on. If you are dealing with an entire factory with myriad communication protocols, it goes without saying that the complexity of conversion will be much greater. Since OT systems are typically used for a few decades or more, equipment from various vendors is often applied in the same system. Moreover, each equipment has its own proprietary hardware design, communication interface, and communication protocol to deliver OT-worthy availability. This approach is effective to ensure system reliability and optimal performance if systems work independently. However, data silos have formed over time. When seeking to aggregate data from different systems, a factory will find each system speaks its own language. For instance, two production lines in the same plant may use different PLCs from two different vendors, each with its own communication language for the respective PLC.
Fortunately, the market is aware of this problem. Many solutions are also available today, such as implementations of consistent and open standards like OPC-UA or industrial protocol gateways, to allow the extraction of data from a machine using a protocol that users are familiar with. For example, with the help of Modbus-to-BACnet industrial protocol gateways, a heating, ventilation, and air conditioning (HVAC) system can obtain Modbus RTU data through the BACnet protocol.
Meaningless Data Has No Use for IT
The next challenge to data quality is unusable data. Equipment-generated data are raw data or values. IT analysts cannot make use of the data as-is, and manual data processing inevitably inhibits real-time responses. If OT data is converted into meaningful values first, data can then flow in the edge-to-cloud architecture seamlessly and quickly. OT data is structured as a series of time-related digits, each representing an event that happens to a specific device or sensor at a specific time, for example, the current magnitude of a certain motor every 10 seconds in the past 7 days. Contrarily, IT data is database-residing data with rigorous structures and descriptions and must be given a meaning before being applied for various analyses. Of the OT data mentioned previously, only the numbers 7 and 10 are shown, and preprocessing is required to provide the data with complete meanings (dates, seconds, etc.) by adding the missing context. Only then can further analysis be conducted.
Besides, for the sake of control precision, OT equipment often produces a piece of data in intervals of a second or a millisecond. If every piece of raw OT data is transmitted to an IT system, it will be too overwhelming for the IT system to do anything purposeful. Even worse, sending meaningless data to the cloud not only reduces operating efficiency but also increases data transmission and storage cost. To tackle this problem, smart IoT devices are used to regulate the frequency of data distribution. In doing so, OT equipment can work in alignment with the needs of IT equipment, such as uploading data once an hour, or processing data on the OT side first and only uploading it when a bigger deviation is observed. It takes these steps to excel at OT Data Preparation.
Myriad Sources Cause Incomplete Data
Digital transformation calls for more diverse and real-time data, and consequently, much more OT data to be transmitted. While OT networks traditionally transmit data to meet control requirements, industrial digital transformation necessitates data transmission for analysis and decision-making purposes. Take the smart factory as an example. To achieve zero failures, production lines must be able to provide immediate feedback every step of the way. When an aberration is detected—a sign of a problem in the previous station—the next station will instantly notify the previous one of the problem to prompt immediate reset, preventing small deviations piling on top of each other and ultimately resulting in failures. In other words, lots of data will have to move through OT networks, including control information and defect images. At the same time, a new challenge will emerge: How to avoid obstructing OT control data transmission with the addition of IT data?
Why is this a concern? It is because industrial Ethernet networks, the most used industrial networks, do not have real-time control mechanisms for mass data. The proposed solution has been to have two separate networks for sending images and control commands. The advantage is that the two streams of data do not compete for network bandwidth, but the cost of network implementation and maintenance doubles. Time-sensitive networking (TSN), the new-generation Ethernet, is thus designed to schedule transmissions according to the importance of the data, ensuring important data reaches the device at the scheduled time. This is what robust OT Data Transmission capability entails.
In addition, as various environmental disturbances, such as extreme temperatures and electromagnetic waves generated during the startup of a device, can cause network disruptions, data may be lost in route. Hence, contingency plans should be made for all kinds of incidents to avoid the loss of data in transit owing to disturbances. As an illustration, when a wired or wireless network is down, the network backup mechanism can immediately activate another section to resume transmission. Or, when the network is temporarily congested or disconnected, a certain amount of the latest data can be stored locally to ensure the data, if lost, will be retransmitted or retrieved to avoid delivering fragmented data.
Vulnerable Data Due to Security Weakness
OT data becomes not trustworthy mostly because of cybersecurity issues. In the past, OT systems did not need to be Internet-connected. Protection could be achieved simply through physical controls, such controlling access to an OT area or banning the use of USB sticks and personal computers. As industrial digital transformation takes off, Internet access becomes essential. With that, all vulnerabilities are suddenly laid bare to ruthless computer viruses or thrust onto the radar of profiteering hackers, providing channels to invade systems and even disrupt operations. When cyberattacks become all too common, data security and cybersecurity are emerging as required items on every digital transformation agenda. To safeguard production capacity and keep production lines safe from data-tampering attempts, firms cannot afford to let their operations succumb to the Achilles' heel of untrusted OT data.
A prevailing misconception among amongst businesses is that mature IT security solutions can be directly replicated in the OT world; in reality, security tools meant for IT environments are not entirely fit for OT protection. A case in point is ubiquitous antivirus software. As OT devices do not run the operating systems compatible with antivirus software, installation of such software packages is out of the question. Further complicating the situation is the paramount importance OT environments give to capacity availability. The fear of production capacity hurt by the possible wrongful blocking of data packets has kept many machines away from antivirus software solutions. Moreover, for the sake of connection stability and convenience, manufactures have instead deployed all devices on the same intranet. However, once ransomware breaks in, it can easily spread throughout the entire system. It is thus recommended to secure OT environments in three incremental stages: endpoint security, cybersecurity. and security management. To enhance OT Data Security capability, firms should:
- Apply Intrusion Protection System (IPS) technology to OT automation devices to secure critical infrastructure. An industrial-grade IPS monitors data flowing in and out of critical devices, segregates malicious traffic, and notifies administrators the instant an anomaly is detected.
- Take advantage of network layering to curb ransomware attacks. Firms will benefit from upgrading their Ethernet switches to managed Ethernet switches and activating the layering feature to divide an OT network into segments.
- Use network management software to overcome the interoperability hurdles among various OT communication protocols to effectively spot faulty or risky devices via visualization.
Let the Four OT Data Capabilities Come to Your Aid
As the old saying goes, “don't put the cart before the horse”. It is critical to get your priorities straight in industrial digital transformation. Don’t let poor-quality raw data undermine the results of your big data analyses. Before trying to source OT data, consider where you are in terms of data acquisition, data preparation, data transmission and data security. Armed with these four capabilities, you will be able to tackle the challenges head on and leverage high-quality OT data to lay a solid foundation for transformation.
Do you want to learn more about the secrets of OT data? Listen to OT Data Next here: