Before You Start
Industrial Internet of Things (IIoT) aims to take advantage of remote connectivity, large amounts of data (or Big Data) generated by factories, predictive analytics and expert know-how to develop applications (mainly cloud hosted) to bring value in terms of, among other things, plant energy efficiency, asset performance or productivity (potential improvements in availability and production throughput directly result in improved productivity).
But as a plant manager, to justify an IIoT investment or implement a project you need to understand your current operational base line and gap to close, the associated cost and, of course, to quantify potential benefits. So what if you are not sure about the current baseline of your operation? How do you determine the existing gap to bring your factory to an IIoT state? To answer these questions let’s focus on the example of Asset Management, which, according to ARC Group, is the key IIoT driver for industrial applications: Any improvement in this area can lead to throughput improvement, cost savings and, in the end, improved process and plant productivity.
In any IIoT roadmap for Asset Management the common first step is equipment monitoring (following the rule: first monitor then analyze then predict) to track their status and get necessary insights to improve operational performance and avoid unplanned downtime. Nevertheless we cannot assume that equipment monitoring and required data will always be available. For example the mining industry has been around for a long time and you can easily find facilities older than 20 years where, on the one hand, the necessary instrumentation to monitor equipment operation is there, so data regarding e.g. throughput, motor speed and PID loops are available, on the other hand however, variables such as vibration, temperature or lube oil flow – important for maintenance and downtime avoidance, i.e. Asset Management – may be not be monitored and the equipment and data lacking.
In this scenario an instrumentation project for the whole plant might or might not be a feasible option depending on budget, operational downtime constraints and required return hurdles. IIoT can indeed bring a lot of value but it can also require high risk management decisions, therefore before implementing any technology a baseline should be defined against which to measure and evaluate potential benefits and before starting a project we should conduct an initial assessment to ensure that we will capture sufficient real value.
Defining project scope
To start with it is important to define which assets, production lines or areas would potentially benefit most from an IIoT implementation by analyzing available data related to plant performance, e.g. operational process efficiency, production throughput, operating costs etc. A comprehensive decision tree summarizing the different KPI’s and how these are affected by asset performance can then be created and this can help decide where to begin. It is often the case (though still surprising) that only 20% of assets (the most critical) usually account for most of net production time and ultimately therefore most of plant profitability.
Evaluating existing infrastructure
Once the scope has been defined we have to examine the underlying system requirements:
- Instrumentation: Assess availability of required machine data and variables through existing instrumentation; Compare failure mode records and available instrumentation for each asset in a FMEA (Failure Mode and Effect Analysis) fashion in order to define the potential failures to be detected and determine what additional instrumentation is needed to achieve this.
- Plant automation network: Assess whether necessary data are available through the automation network or only in a local mode and what additional infrastructure (may be wireless) is required to make all needed data and variables available.
- Data handling and management: Identify where the data is currently being stored and the feasibility to pull this data out of the system. The data may be located at DCS OPC server, Historian server, etc., but what is important is to ensure that it is going to be available for the necessary applications.
- Remote access: M2M requires persistent / continuous communication through the internet which is usually not a default option for remote connections. Therefore it is important to understand the current IT policies and how a persistent remote connection can be enabled given IT security policies.
Experience has shown that most of the value that IIoT can bring strongly depends on understanding this initial baseline situation and the gap to close: Big data is not useful unless it is the right data and available in the right way at the right time.
Some vendors and plant operators may assume that an IIoT solution is kind of plug & play, easily implemented. But in fact this initial data driven analysis –essentially a consulting process, which can at times be quite tedious is essential. Only in this way can risks be controlled and minimized and probability of success improved. Furthermore with a data/information driven analysis showing quantified potential benefits in terms of productivity and performance it is easier convince decision makers of the merits of implementation.
Defining solution architecture
Once the technological status of the plant has been assessed and evaluated and the project scope has been defined, it is then possible to start developing an approach to Monitor/Diagnose/Predict by defining the technology platform and applications required.
Considering our initial baseline as a starting point, a solution in terms of hardware, software platforms and applications should be developed aiming to monitor equipment behavior, diagnose equipment problems by getting sufficient insights into current status and finally predict failure modes by applying predictive analytics. Where precisely you start to develop the IIoT solution really depends on current plant status: Some plants may have poor instrumentation and isolated networks while others may just need to pull out already available data from a Historian system, in other words: the higher the current maturity level, the lower the required additional investment.
The first step aims at establishing good basic monitoring, therefore all sensors required to generate missing data (e.g. vibration, temperature, lube oil flow, etc.) should be considered. This does not mean that every single piece of equipment will need a new sensor but rather a requirement to focus on the most important failure modes (biggest impact) and the equipment variables related to these. Data already available should also be taken into account in order to avoid sensor redundancy which can create significant problems due to communication protocols and legacy technology that will need to be integrated into the solution. At the same time we have to consider the best approach to integrate this data into the communication network for example through an I/O module -considering wired or wireless options-, in order to make this data available in the control system or also through the DCS OPC server.
In the second step we use the available data to visualize and get insights on equipment condition and health, the idea being to use analytics tools to identify issues that can impact equipment availability resulting in downtimes and production losses. Various approaches can be used here such as: trend visualization, descriptive statistics, frequency domain analysis, decision trees, root cause analysis or expert rules in order to develop the necessary applications. OEE on-line calculations, performance dashboards, or data correlation applications belong here as well and it is perfectly feasible to host this layer in the cloud so it’s not necessary to install and maintain additional hardware/software, but one can remain focused on the continuous improvement cycles and management tasks.
The final step aims at developing solutions to avoid downtimes by taking advantage of Big Data captured in real- time by the sensors and M2M communication by generating predictions, i.e. to move from “what’s happening” to “what’s going to happen” (from diagnostics to prognostics) through the use of predictive analytics tools. Again various approaches can be used, including clustering, classification or regression analysis. The objective is to predict failure events with the associated failure modes and therefore enable early warnings on equipment health –sufficiently into the future so that it is possible to react with minimal disruption to operations. These prognostics applications are complementary to the diagnostics applications providing additional information on where the focus should be, the priority subsystems that the condition monitoring software should analyze and the measurement ranges to be considered.
The goal of this IIoT approach is to develop an architecture that serves as a technological ecosystem for “Equipment Monitoring and Predictive Analysis” but also for “Operational Intelligence and Life Cycle Management” that supports a proactive maintenance strategy executed in a continuous work cycle, therefore allowing a dynamic and modular matrix maintenance where the tasks are prioritized, tracked or rescheduled under a data driven solution, expert engineers’ analysis and a whole view of the assets and plant performance.
Additional aspects to consider
The IIoT architecture will affect the whole plant operation and there are also other dimensions that should be considered in the solution scope in order to maximize potential benefits and minimiye risks:
- Implementation roadmap: IIoT implementation is (should) not be a sudden change but rather an incremental process where first the data is aggregated and then used to build at least 2 levels of applications. For sensor implementation it is probably necessary to shut down relevant equipment so required downtimes should be considered and planned in such a way as to minimize production losses. Initial monitoring and analysis applications need to be developed and customized based on historical data and first results need to be evaluated and processes calibrated before real data is used. So first benefits will only become visible after a minimum of three months or so. The software platforms’ inherent flexibility and strong implementation support are key to ensure fast and robust implementation.
- Change Management: While the IIoT solution may provide an excellent enabling framework for working towards operational excellence in asset management, nevertheless there will probably will be a gap to close in terms of know-how and/or the ability of the organization to adopt, absorb and actually effectively utilize the technology. A similar process as the initial assessment of the existing technology status, but this time focusing on internal processes and people should be followed. This will help set a baseline and define what training will be needed and which processes should be improved or changed and whether and how asset management and maintenance strategy and practice should be re-configured to be better aligned with and make the most of the IIoT solution. This can be a particularly sensitive point and managers must decide how they will go about this, understand the effort required and whether they will need external support. If the change process is not managed properly there is a high risk that the project will fail not only in realizing the promised benefits, but in fact the outcome might be worse than the initial status. On the other hand good change management can enhance the full potential of the solution with a direct impact on the bottom line.
With such an approach we can reduce the risk and maximize the potential of an IIoT implementation, taking into account all interrelated factors that can have an impact on the implementation process as a whole. Furthermore we gain clarity at a sufficient level of detail to be able to more confidently forecast potential benefits and costs of the project and therefore calculate returns and present accurate budgets and time horizons.
In another article we will discuss some common challenges in IIoT implementations and show some ways to overcome them.