How to properly implement IoT in production

How to properly implement IoT in production

Many experts call the Internet of Things (IoT, Internet of Things) a technology that can dramatically change modern business processes. Ilya Koshkin, Senior Manager of Infrastructure Services at Accenture in Russia, talks about the main success factors and the reasons for the failure of IoT projects at manufacturing enterprises.

DO NOT DO WITHOUT ANALYTICS

Today we can say that business still does not fully realize the potential of the IoT. There is interest in technology and the experience of some successful pilot projects, but there is no understanding of how to get the most out of the Internet of Things and replicate success throughout the organization. Companies lack the perception of IoT as a fully functional technology.

The team of industrial sensors search platform SourceMe believes – “Everyone knows that there are sensors, data from which can be collected, stored, and processed. But why this is necessary and how to apply these capabilities in practice, is understood by a rather limited circle of specialists within companies.”. Learn more about the resource here.

The overall concept of IoT is inextricably linked to the subsequent in-depth analysis of the collected data and the ability to make business decisions based on the findings. Roughly speaking,

INTERNET OF THINGS AT THE LEVEL OF SENSORS AND DATA STORAGE WITHOUT ADJUSTMENT IN THE FORM OF ANALYTICAL SOLUTIONS AND ARTIFICIAL INTELLIGENCE – A WASTE OF RESOURCES

Let’s take any industrial production. The IoT will allow you to collect a huge amount of data on the operation of equipment by recording certain parameters within the production cycle. However, all this information without analytics will be almost useless: it will not provide any discoveries by itself.

Businesses need to search for hidden correlations and patterns in routine workflows. Next – generate recommendations based on such findings.
Most often, the output will be hypotheses, and here it is important to involve specialists from business departments who have a deep understanding of the production process and are able to assess how relevant and valuable the provided conclusions are, to work with industrial IoT solutions.
In practice, this model is rarely implemented.

DISEASES OF GROWTH

Currently, IoT in industry suffers from a whole list of chronic “growing pains”. First of all, most manufacturing companies do not store their data in the required amount in order to effectively apply artificial intelligence (AI).

In addition, there are many procedural problems associated with the human factor. In the whole complex of IoT tasks at the enterprise, the final interpretation of the data processing results is often assigned to a specific specialist. But employees are not free from bias, are subject to professional deformation, and sometimes are not at all interested in creating an objective picture of their area of ​​work.

For example, to train an AI model on the collected data, you need to load into it sample situations that clearly demonstrate how a certain combination of factors led to a breakdown. In practice, it happens that employees often interpret such situations “in their favor” in order to retain their place in the company and avoid punishment for mistakes.

The third point: in large enterprises, systems and their elements, as a rule, were installed at different times, with a gap of 10-20 years. Many production mechanisms and assemblies are designed for a very long service life.

INTEGRATING DIFFERENT “HISTORICAL” PARTS OF THE GENERAL PRODUCTION SYSTEM IN A COMMUNITY IS VERY DIFFICULT, EVEN AT THE MONITORING LEVEL

Accordingly, only part of the production process is monitored. This may be the most critical area or the one where breakdowns occur most often. It is extremely rare to achieve end-to-end transparency of the process for specialists.

Equipping production devices with IoT sensors is a separate headache, since it is rather difficult to determine the configuration of suitable devices due to the fragmentation of the equipment, and even more so to make an economically viable installation (without stopping production).
As a result, only a case-by-case approach works in IoT for industrial companies (for each case separately).

DATA COLLECTION AND BOUNDARY CALCULATIONS

Another challenge for the Internet of Things in industry is the variety of data formats obtained from many production systems and various sensors.

As a rule, large systems like Siemens or Honeywell communicate using different proprietary protocols, which leads to difficulties in integrating with the rest of the enterprise IT systems, especially if we are talking about real-time monitoring.

IT IS NECESSARY TO ENSURE CONTINUITY OF DATA ACCESS, TAKING INTO ACCOUNT FORMAT CONVERSION TASKS

Here we turn to the issues of the correct organization of information collection. It will be optimal to use solutions of the class of edge computing, located directly at the production facility. In addition to collecting data, such solutions also carry out preprocessing, preparing them for deeper analysis using AI and machine learning (machine learning).

Today, major vendors such as HP, IBM and Dell have entire lines of equipment for edge tasks. They are constantly evolving and progressing, which, on the one hand, opens up new opportunities for the Internet of Things, and on the other hand, significantly complicates the process of their integration into a single IoT complex. All this leads to a shortage of consultants and integrators who are able to efficiently assemble all the elements into one working solution.

It should also be borne in mind that scaling itself is a costly process. Data flows increase significantly, integration becomes more complicated, models need to be retrained – these and other factors, with insufficient quality planning, can slow down the operation of the IoT system or provoke its failure.

As a result, not every company is ready to spend on an end-to-end IoT solution, starting the transformation of an enterprise from a small section of the production cycle. Here there is a risk to take the project and complete it, because this approach will give too small a sample of data and, most likely, will show inconclusive results. The preparatory process will consume a lot of time and money.

THREE STEPS TO SUCCESS

However, not everything is so gloomy and hopeless. With the development of technologies (networks, computing power, storage systems and algorithms for data analysis), the possibilities in the field of collecting information from a huge number of sources and real-time processing are growing every month. The 5G perspective makes the IoT one of the most promising areas for digitalization of business in general and industry in particular.

We are on the verge of a qualitative leap forward. The sum of the technologies on the basis of which the Internet of Things is developing allows us to extract practical insights from production data that simply did not exist before. For a breakthrough, you need to learn how to overcome the listed problem areas in the chain of information flow from production machines and systems to analytical conclusions that are useful in practice.

To do this, you should not lose sight of the next steps, which are guaranteed to give an IoT impulse to the development of the enterprise.

Planning. The goal is to identify the most important production areas from a business point of view, to identify areas that generate the most useful data. A pilot project is launched on them.

A pilot project with an eye on subsequent scaling. In the case of demonstrating promising results, you can proceed to the stage of scaling from the pilot site to the whole area or to the entire enterprise at once. Here you will need a prototype platform based on a pilot, taking into account the integration of all systems, all types of sensors and data formats. At the same stage, machine learning models, mechanisms for their integration, retraining opportunities, as well as types of dashboards (dashboards) are being worked out, providing the management with the final results of analytical data processing.

Development of a technical solution for the project and its implementation. A key prerequisite for working in this direction is understanding the gap between the engineering level (sensors) and the IT level of the final IoT solution. Relatively speaking, the people involved in sensors will not be able to tell you about edge computing and machine learning models. Conversely, the specialists in these solutions have no idea about the operation of specific sensors. Meanwhile, for a successful project, both are necessary. For example, sensors for IoT are a separate world with their own leaders, standards, requirements in terms of compatibility with IT components, etc. There are flagships in this market (National Instruments or HBM), but
FOR CERTAIN AREAS AND TYPES OF EQUIPMENT, SOMETIMES IT IS OPTIMAL TO USE VENDOR SENSORS OF THE SECOND OR THIRD ECHELON, INCLUDING FROM ASIA OR RUSSIA
This is the only way to build a solid foundation for an IoT project at a basic engineering level.

THE TEAM IS ALWAYS IMPORTANT

The key success factor in the IoT direction today can also be called the quality of cooperation between the production unit of the enterprise and the IT team in charge of the project. Without efficiently tuned cooperation between them, it will not be possible to generate working insights about production. Only an understanding of business processes and technical aspects will allow you to extract the maximum practical results from the IoT already at the pilot level and lay a high-quality foundation for scaling across production.

Yes, you can analyze “dry” data without involving “interpreters” from the business departments. But, in my opinion, this approach represents the next stage in the development of IoT and another level of analytical expertise that only comes with time, although it makes sense to try it within the framework of the pilot.

Today, the gap between the specifics of production, business goals and the technical component of the IoT is covered by an integrator consultant, whose work experience combines engineering and IT knowledge and allows you to implement projects from the design stages to commissioning.

WHEN MATURITY WILL COME

After the IoT solution is launched into operation at the enterprise, you can attend to the tasks of enriching production data with information from other enterprise systems.

In the future, industrial IoT projects can pay off through highly profitable enterprise integrations. Opportunities for continuous production control through an IoT solution allow you to clearly understand the volume of production, the level of quality, and predict them for a certain time horizon.
You can correlate this information with general economic trends, market dynamics and, using a mathematical model based on historical data, set a favorable market price for your products, ensuring maximum profit.

It is important to understand that IoT does not always have to open up fundamentally new directions or markets. Optimization of the existing production cycle will make it possible to seriously improve the existing processes in production, which will have a positive effect on the economic picture of the enterprise.

IoT in the industry today is a fully implementable concept, the maturity of which in the industry will come in three to five years.