Author: idbox

INTRO

Although Artificial Intelligence (AI) technologies began to be developed several decades ago, several definitions currently coexist that include concepts such as Big Data, data science, machine learning or IoT.

AI is defined as a set of tools and methodologies that allow machines to simulate processes and develop tasks that require applying human intelligence such as decision making, object recognition or speech understanding.

The industrial sector has been immersed for years in its digital transformation in what has been called the fourth industrial revolution or Industry 4.0. There is no doubt that this sector has a high potential to lead the application of AI technologies in the monitoring and optimization of its processes, providing an appropriate scenario to test its benefits, from design and manufacturing processes to the management of the value chain. Thus, large companies such as GE, Siemens, Intel, Funac, Kuka, Bosch, ABB, NVIDIA and Microsoft are investing heavily in the development of new systems to improve their production.

At this juncture, the concept of Industrial Artificial Intelligence (IAI) has recently emerged, which can be defined as the application of AI to the operations, processes and physical systems of a company, so that the behavior of these operations, processes and systems can be monitored, optimized or controlled to improve their efficiency and performance, providing them with greater autonomy. Thus, this concept includes applications related to the manufacture of physical products, production lines and warehouses, or operations related to different processes.

It is an end-to-end system, in which sensors generate data, which are sent, managed and analyzed by means of different algorithms and models that generate decisions in real time and whose results are returned for actual implementation in the actuators.

USE CASES

There are several different use cases related to three categories of applications depending on the degree of automation involved: monitoring, optimization and control.

Monitoring

Industrial processes need to monitor the performance of their systems and products to identify or predict failures and other situations that lead to unsatisfactory results. Some examples that will benefit from IAI technologies are as follows:

• Quality Control

Companies find it difficult to maintain high levels of quality and comply with regulations and standards due to the current short time-to-market for new products and services and their increasing complexity. On the other hand, consumers expect defect-free products, so companies must avoid the damage that complaints and defective products can do to the brand.

In this context, AI algorithms enable a new form of quality control. On the one hand, algorithms based on image recognition technologies warn production teams in real time about faults in production systems that can reduce product quality (e.g., recipe deviation, changes in raw materials). On the other hand, the implementation of these algorithms allows the prediction of maintenance and planning tasks that minimize the associated risks. Finally, the integration of such algorithms with IoT platforms allows the collection of data on the use and behavior of products during their useful life, information that can be very valuable when making strategic and design decisions.

• Predictive maintenance

The continuous maintenance of production machinery represents a major expense in manufacturing processes, so it is important to implement solutions based on AI algorithms that allow predicting future failures in a part, machine or system in order to drastically reduce unplanned downtime and increase the useful life of production systems.

Optimization

IAI-based decision making and planning systems allow users to design plans to optimize a set of business metrics.

• Process planning

Many industrial scenarios include complex work sequences whose order of execution can significantly impact factors such as cost, time, quality, workloads, supplies or waste. The application of AI-based optimization algorithms allows such sequences to be defined dynamically in real time.

• Generative design

AI is changing the way products and services are designed. Generative design uses AI algorithms to design new products based on their description, including parameters such as type of material, means of production, budget or time to market. These algorithms analyze different configurations before proposing the best solutions.

• New market adaptation strategies

The application of AI is not limited to the production plant; its algorithms make it possible to optimize supply chains or help companies anticipate market changes. All this is a great advantage for business management, moving from a reactive approach to a strategic one. Thus, estimates of market demand can be formulated by searching for patterns that relate location, socioeconomic and macroeconomic factors, weather patterns, political status or consumer behavior so that companies can optimize their resources or control inventory.

Control

Finally, control systems are at the heart of industrial process operations and are essential to reap the full benefits of automation. Examples of applications that benefit from AI include the following:

• Robotics

Traditionally, industrial robots have been explicitly programmed to move between a series of 2D or 3D points and perform specific actions at those points. New approaches such as collaborative robots or co-robots simplify programming by allowing the capture of these points based on the robot’s physical position. However, in both cases, the robot does not detect changes in the environment or in the position of the parts it is manipulating. Therefore, computer vision allows robots not to interfere with people or other robots, and to interact independently.

• AGV

Autonomous mobile robots (AGVs) are used in warehouses and companies to transport and collect materials thanks to the use of image-based AI algorithms that enable them to understand, map and navigate these environments more efficiently.

CHALLENGES OF INDUSTRIAL ARTIFICIAL INTELLIGENCE

Industrial AI presents a number of challenges that set it apart from other consumer-facing AI applications.

• Data acquisition and storage

IAI systems rely on data captured by sensors that seek to digitally represent the real world. The implementation of IoT platforms and cyber-physical systems (CPS) has enabled large volumes of data in industrial processes whose acquisition, management and storage has led to different architectures and storage systems.

However, it should not be forgotten that large volumes of data are often captured with a lot of noise, which makes it difficult to acquire and store the data for subsequent analysis. Therefore, advanced simulation techniques including digital twins are used to generate training data under different conditions.

• Data hybridization

A challenge of IAI is the generation of common formats for heterogeneous data coming from diverse sources (images, videos, dwg plans, …). For this, it will be necessary to develop semi-automatic models that help to structure (e.g. extraction of information from a plan, extraction of tags from an image), homogenize (e.g. spatial/temporal interpolation techniques to move from data in one unit to another) and/or harmonize data (e.g. product matching/record linkage techniques to associate data from different sources).

• Training

The correct application of IA algorithms is associated with the availability of annotated training data. Capturing this data can be complex in industrial environments, as it is often difficult to detect and reproduce some of the failures or lack of quality of products and services.

• Regulated environments

Industrial environments must comply with certain standards and regulations that impact their operations, such as product safety, public health, environmental impact or occupational safety. In some cases, regulatory controls can make it difficult to implement AI technologies.

As stated in the European Data Strategy1 published by the European Commission in February 2020, Europe must become a model of a data-empowered society for better decision making in the public sector and private enterprise. To this end, the European Union will promote a legal framework in relation to data protection, fundamental rights, security and cybersecurity. The aim is to generate an ecosystem of trust thanks to a regulatory framework for AI2.

SITUATION IN SPAIN

According to a study by EY for Microsoft, the majority of companies surveyed in Spain (65%) have plans, pilot projects or proofs of concept around AI. However, only 20% have solutions in operation, 12 points below the European average.

The “Strategy for Artificial Intelligence in R&D&I in Spain” includes a section oriented to the application of AI in the Connected Industry. As the document states, Spanish industry represents 13% of the country’s added value and employs 11% of the employed population, so the social and economic impact of AI technologies is essential.

Manufacturers most often use Artificial Intelligence (AI) to improve overall equipment effectiveness (OEE) and first-cycle output. Over time, manufacturers can use AI to increase uptime and improve quality and consistency, enabling better forecasting.

Like many other components of digitalization, implementing AI can seem like an overwhelming task. Additionally, concerns about how to effectively use and manage the billions of data points generated by AI computing power and connected machines are common among manufacturers. Many people don’t know how to start; moreover, they often attribute their caution in implementing AI to costs, IT requirements, or fear of not being ready for Industry 4.0.

To stay competitive, it is important for manufacturers to adapt to a more data-driven business model. This often involves reorganizing personnel, upgrading equipment and software.

AI, a concept often associated with the future, is now a reality and can be applied to your facility today. Here are 3 ways in which industrial AI is revolutionizing manufacturing and tips on how to implement it:

1. Predictive and preventive maintenance

Major equipment failures, whether mechanical or electrical, cause significant manufacturing downtime. Following recommended preventive maintenance schedules can prevent these breakdowns. However, preventive maintenance is often overlooked or not optimized for quick turnaround times. Thanks to the capabilities of IoT devices, sensors, MES data and machine learning algorithms, manufacturers can use multiple points of machine data to predict breakdowns. Optimize maintenance schedules before predicting breakdowns to maintain machines and ensure smooth production floor operations.

2. Supply chain automation

Today’s supply chains are ultra-complex networks that need to be managed, with thousands of parts and hundreds of locations. Moreover, AI is becoming an essential tool for the rapid delivery of products from production to the consumer. By leveraging machine learning algorithms, manufacturers can determine the optimal supply chain solution for all their products.

Furthermore, managing internal inventory can be a major challenge in itself. A production line heavily relies on inventory to maintain operations and meet production targets. Throughout the manufacturing process, operators must replenish a specific amount of components at each step to sustain efficiency. Ensuring the factory floor has all necessary supplies is critical, and artificial intelligence can play a pivotal role in optimizing this process. AI can learn component quantities, track expiration dates, and efficiently distribute them across the shop floor.

3. Production optimization

Optimizing a manufacturing process can be a data-intensive task that involves countless historical data sets. Determining which process parameters provide the highest quality products is no easy task. Production and quality engineers are constantly conducting dozens of experiments to optimize process parameters, but they can often be costly and time-consuming. By using artificial intelligence to process data quickly, engineers can find the best process recipe for different products. The AI will constantly learn from all points of production data to continuously improve process parameters.

Bonus: Augmented and virtual reality

As augmented and virtual reality technology improves every day, and with more and more large companies developing devices for this market, it’s only a matter of time before the manufacturing industry fully embraces it. Furthermore, virtual reality can help better train product assemblers to perform assembly or preventive maintenance tasks. Similarly, augmented reality provides real-time, machine-learning-based reporting in the factory or in the field, helping to quickly identify defective products and areas of improvement. The applications of AR/VR in manufacturing are endless and can play an important role in solving today’s problems.

The era of the smart building has arrived. Today, thanks to advances in technology, a building can not only provide all the services the occupants need. But, also, do it as efficiently as possible, minimizing costs and increasing energy savings over the life of the building. It’s a balance that will be key to business in the future.

What is a smart building?

A smart building uses technology to use resources efficiently and economically, while creating a safe and comfortable environment for occupants. Intelligent buildings not only use a wide range of existing technologies but also integrate future technological developments through design or retrofitting. Moreover, Internet of Things (IoT) sensors, building management systems, artificial intelligence (AI), and augmented reality are some of the mechanisms and robotics that control and optimize performance in a smart building.

The benefits of smart building analytics?

Smart buildings generate a lot of valuable data about how they are used. Analyzing this data can give you insight into usage patterns and trends, allowing you to make informed decisions to optimize building performance, providing the following benefits:

• Increased productivity.
  Creating a conducive environment with good indoor air quality, comfort, safety, sanitation, and efficient processes enhances employee performance. Therefore, understanding how people use and move within your building is crucial for optimizing space and reducing waste.Increasing the size of a cramped, high-traffic area can be a practical example of this.

• Reduced energy consumption.
  Smart buildings can improve energy efficiency and, in turn, reduce energy costs. For instance, by connecting IoT sensors that track occupancy to your building management system, you can automatically turn off lights or HVAC systems in unoccupied rooms or spaces. This reduces unnecessary energy consumption and the emissions that these aspects produce.

• Lower operating costs.
  Building overhead is a significant expense for any building owner/user. However, while they are a necessary expense for a business, the level of cost is often wasteful because it is not applied wisely. By identifying patterns of underutilization, you can reduce the footprint of a property to reduce costs.

There are many benefits to implementing smart systems in a building, from cost-effectiveness to making the structure more environmentally friendly. Today, smart buildings are relatively new. But, given the wide range of benefits they offer, they will soon become the norm.

How can you make your building smart with data?

The key to successfully turning your building into a smart and efficient space is understanding that only accurate and reliable data can help. Indeed, data is at the heart of smart building systems, determining how a facility is used. With this information in hand, you can identify areas for improvement. Either by integrating with other smart technologies and building systems that provide automation or by facilitating strategic decisions. The IDboxRT Operational Intelligence solution can help you monitor different types of usage data in real time.

Additionally, want to get started or learn more about how your building is being used? Request a Demo and talk to our experts.

One of our main customers, specialist in Energy Efficiency and Photovoltaic Self-consumption solutions for the industrial sector, EDF Fenice, has launched its Energy Efficiency Control Center, EnergyHub, thanks to the implementation of the IDboxRT tool. This installation is now operational and will enhance the performance of the energy assets of its customers’ production plants.

“From EDF Fenice’s new offices in Madrid, its energy efficiency engineers and technical support staff permanently monitor in real time all the energy vectors of its customers’ factories, which have already improved their energy efficiency, reduced their costs and minimized the environmental impact of their activity.” – Europa Press comments.

We often find ourselves in the situation of possessing or having access to information, a lot of information, but which, without any kind of organization, is of little use. What is the point of having thousands of data if I am not able to exploit them? Organized data becomes information, very valuable at times, and that is precisely what operational intelligence platforms such as IDboxRT offer me.

This ability to squeeze the most out of my data to obtain valuable information is presented to me in many different ways: the ability to graph and compare historical and real-time data, performing calculations to obtain new data, positioning this data on maps, reports…, but we also offer a more visual way to work with this information: the synoptic.

A synoptic is that “which presents the main parts of a subject in a clear, quick and summarized way”, as dictated by the dictionary. In our case, we use the synoptics to capture data in a visual way, being a schematic representation of reality, which allows us, at a glance, to obtain that valuable information and to know what is happening at the present moment.

Throughout all these years working with synoptics we have used many different approaches based on the nature and needs of each project and client, but experience has made us adopt certain techniques to improve these representations that are often repeated. This is the case of the drill-down approach, in which we start from a generic representation, a high-level visualization, where at first glance it is interesting to know general data, KPIs, and get an idea of where everything is located. From this first level we would navigate to the next one in which we could go into the detail of the selected area, and even obtain data of this in comparison with the rest of the analogous areas, and so, we would continue descending in level to go more and more into detail and get to the specific data.

A real example that we have developed with this approach could be: A first representation of a map of the world, dealing with a client that operates globally, where the top management can get at a first glance information on how their manufacturing plants are operating, and compare to make sure everything is working correctly. With a single click we could navigate to a second, more country-specific level to see the detail of the country’s plants, and in turn navigate to a plant, either because we want more detail or because something has caught our attention when reviewing the KPIs. Once we are visualizing the representation of the plant, we can consider the navigation to specific areas of the plant, and even continue down to the level of machinery or specific parts of each of the machines.

There are infinite approaches for this type of representation, although it must always be in accordance with the need for information. The synoptic representation loses its usefulness if the data shown are not understood or provide unnecessary information.

It is also important that the design of these graphic representations is carried out in accordance with the client, since usually each company has its own way of representing its assets (diagrams, maps, drawings, plans…) and it would be a contradiction to propose something that is unfamiliar to them and would involve a work of adaptation. In addition, we work to maintain the corporate identity of each client so that they perceive these representations as their own and feel “at home”. In this way, they can also make use of these representations at company level on information panels or videowalls. All this entails a previous work to the realization of the synoptic based on the study of the brand and its uses.

Following this approach we have worked on the representation of water treatment plants, road construction projects, industrial production monitoring, refining plants, energy efficiency in buildings, Smart Cities projects and so on.

On the other hand, it is important to highlight that all our synoptic representations use the standard SVG (Scalable Vector Graphics) format, which makes the representations scalable ensuring that there is no loss of quality, and allows the customer to make use of their own representations by importing them directly into the platform.

In addition, from IDboxRT we have an extensive library of pre-made elements of all kinds, so the user can make use of these forms with a simple drag and drop, and even add their own forms to this library.

Synoptics are undoubtedly one of the most powerful tools for data analysis within IDboxRT and are widely used by our clients, regardless of the sector, as they greatly facilitate the understanding of the data and are visually very attractive.

Understanding the differences between operational intelligence (OI) and business intelligence (BI) is crucial to contextualizing and taking action on the information and insights provided by your analytics toolset. While both operational and business intelligence are used to drive action and inform decision making, there are key differences that distinguish these two areas of analysis.

Business intelligence maintains a relatively narrow focus with an emphasis on finding efficiencies that optimize revenue or profitability. BI typically means taking a snapshot of data over a defined period of time in the past and reviewing it to understand how the organization might achieve better success in the future.

In contrast, operational intelligence focuses on systems, rather than profits. OI uses real-time data collection and analysis to reveal trends or problems that could affect the operation of IT systems and to help front-line workers make the best decisions about how to address those problems.

The differences between operational intelligence and business intelligence can be summarized as follows:

Business intelligence focuses on finding efficiencies that increase or protect profits, while operational intelligence focuses on maintaining the health of IT systems.

Business intelligence leverages more historical data, while operational intelligence relies on real-time data collection and analysis. Operational intelligence has been described as immediate business intelligence gained from ongoing operational functions, a definition that speaks to the real-time nature of data collection and focuses on the operational functions that characterize operational intelligence in an enterprise environment. While business intelligence typically runs within a specific data silo, operational intelligence helps organizations break down data silos to uncover trends and patterns of activity within complex and disparate systems.

Cementos Portland Valderrivas is a company of the FCC group dedicated to the manufacture and sale of cement, concrete, aggregates, mortar and other derivatives.

Within its global digitization project of different industrial plants in Spain, IDboxRT is proposed as a monitoring and integrating platform for all facilities, with the aim of creating a system that allows integration with SAP system data, reading of different types of files with analytical capacity for data visualization and connectivity with installed customer software, which makes it a collaborative tool within the corporation.

Main challenges

The main challenges encountered at the beginning of the project development focused on data acquisition, connectivity with the SAP system, extraction of historical data, as well as the creation of complex calculations for the FORTIA system and the synchronization of data transmission from the different systems.

Proposed solution

The starting point was seven Spanish plants, including also the data generated at the company’s headquarters.

The proposed solution was realized with the CIC IDboxRT Reporting Services system, LUCA, with the capacity to visualize the required data. On the one hand, production dashboards were needed through accumulated data by product type, and on the other hand, comparisons between the different plants of the company.

On the technical side, the project was developed in three areas depending on the data acquisition technology: through flat files extracted from SAP, Excel files from FORTIA and through the historical data tool, APIRest. For this purpose, three drivers were created for the collection of files in the shared directory. The data was loaded into the system and complex energy calculations of the grouped data were developed for later display on the platform.

Results obtained

As main results, the development of the project ensured the creation of Reports differentiated by the role of the users, the integrated acquisition of FORTIA data, as well as the generation of Reports that reduce the execution time of the tasks by the plant personnel and new data synchronization mechanisms. At the same time, a comparison of consumption and production results in the plants was developed, and access to data from different customer departments was generated from the same point.

Useful tips

To conclude, we would like to highlight several important conclusions for future developments: firstly, it is necessary to carry out data validation more efficiently, regardless of whether it is done by the customer or ourselves, since validation time may exceed development or implementation time; secondly, it is necessary to correctly capture customer requirements at the beginning of the project to reduce development time; and finally, it is necessary to seek and deliver a design that is better adapted to the company and brand image.

The technology key in improving the present future

The ability to synthesize sensory data, while preserving the desired statistical properties, is currently proving to be a great success in different industries.

Many examples are based on this concept and apply it to various industries using technology. One of the most prominent is the case of DeOldify, an artificial intelligence program that translates black and white images into color, or Nvidia, with its proposal to create realistic images of fake landscapes or non-existent people’s faces, from semantic sketches.

GANs: The most interesting idea in Machine Learning of the last decade

These systems are based on a very specific neural network architecture, called Generative Adversarial Network (GAN), an artificial intelligence algorithm based on the fact that synthesized data must maintain both statistical properties and be indistinguishable from real data, a process similar to a Touring test for data.

Such concepts have their origins in the past, where the comparison was made through simple visual inspection. Nowadays, classification models, called discriminators, are used to distinguish between synthesized data and real data. In a more intuitive way, this network can be understood as two competing networks: the first one is in charge of generating candidates, synthesized data, while the second one evaluates whether the data is real or synthesized. The goal of training is to increase the error rate of the discriminating network, i.e. to deceive the discriminator.

Monitoring, deep learning and its business benefits

The solutions described above are adapted to various sectors such as “Smart”, “Industry 4.0” and “Energy” among others. Real-time asset monitoring software is starting to use these technological advances to solve common problems, such as, for example, connection failures. It often happens that some sensors sending data are partially disconnected, which could have been avoided if a generator-discriminator model had entered the game, replacing the missing data with synthetic data. From here we could consider the possibility of replacing some sensors completely with synthetic parts, which would ensure the highest possible quality and reduce the hardware infrastructure required for effective monitoring.

Currently, Spanish companies such as CIC Consulting Informático, with its asset monitoring product IDboxRT Inteligencia Operacional, consider Deep Learning as a tool to make their customers’ lives easier.

The series of measures in the field of Deep Learning in monitoring, developed by CIC Consulting Informático, leads to significant positive results at several levels. First of all, it provides favorable economic conditions, allowing savings in the operation and maintenance of specific equipment, while avoiding serious losses of information. For this reason, there are advantages associated with energy efficiency, such as the reduction in energy consumption resulting from the reduction in the number of physical components.

Back to the future with deep learning

Deep Learning is expected to have a revolutionary effect on the way companies operate in the near future, making them more efficient in terms of consumption and profitability, optimizing all their processes and achieving tangible results on a global scale.

This is nothing new. We are not talking about a breakthrough technology that will appear with force in 2020. The IoT is something that has been on the minds of its precursors for many years, since the 1990s, simply waiting for communications and systems integration techniques to support their ideas.

And it is precisely this last word that is the key to understanding the reasons why its adoption in the industrial field has not been as rapid as predicted by the major global consulting firms. The truth is that we are not talking about a technology in itself, whose mere application is capable of solving a problem, but about a concept of application of several technologies in the service of a basic premise: an idea.

During the first years of the IoT boom, there have been situations in which medium-sized and large companies did not take steps, but real leaps towards the “application” of IoT in industrial environments. The problem is that these leaps were simply based on investing heavily in IoT devices, whose data turned out to be anecdotal or ancillary.

From the experience accumulated by our IDboxRT team, we subscribe to the maxim “What can’t be measured can’t be improved”. But the hype that IIoT is experiencing should not drag us into an unjustified eagerness to collect useless data, the counterpart to this maxim is that we should only and exclusively measure those parameters that allow us to achieve the expected ROI.

On this solid basis, a tide of device manufacturers, models, protocols, etc. appears before the leaders of these initiatives. In this sense, it is difficult to predict which of them will dominate the market in the medium term, so it is essential to have an open IoT platform that allows communication with a variety of devices in a simple way, as this greatly facilitates the choice of the right device for each use, without fear that certain data will remain isolated as the technology evolves.

At the level of communication protocols, we find a variety of lightweight protocols that allow communication with remote devices powered by batteries, whose duration can reach several years depending on the protocol used and the refresh rate. From the well-known MQTT, through CoAP, to other less recognizable protocols such as BACnet, we will find a multitude of protocols implemented by different devices, which may create doubts in those who have data processing platforms with low flexibility.

It is precisely this open nature what makes the Operational Intelligence tool that we developed at CIC Consulting Informático de Cantabria, IDboxRT, appear as one of the references in Gartner’s 2018 Competitve Landscape: IoT Platform Vendors. Without being able to be considered solely an IoT platform, the ability to ingest data from any device, regardless of the protocol, makes IDbox one of the safest bets in this regard, since whatever direction the industry takes, our customers will always be able to integrate their data, combining different protocols.

The possibility of combining data from our “IoT park” with process data directly collected from PLCs, SCADAs, databases or even third party WebServices allows IDboxRT customers to contextualize the information, implement mathematical models, combining all this data, and analyze the results to improve decision making in real time.

In short, from the IDboxRT team, we are sure that the implementation of IoT initiatives in the industrial field will undoubtedly bring substantial improvements both in terms of control and process optimization, as long as we focus on the value that each piece of data can bring to the heart of any initiative: an idea.