In previous posts, predictive maintenance was mentioned as one of the main digital enablers of Industry 4.0. Maintenance, linked to the industrial revolution, however, has accompanied us in our evolution as human beings.
Since prehistory, our ancestors have built tools that suffered wear and sometimes broke without prior notice. The solution was simple: to carve a new tool. By creating more elaborate mechanisms (e.g. wooden wheel), the natural alternative to disposal became the reparation by the craftsman. Mechanical looms of the First Industrial Revolution were even more complicated of repairing so specific professions emerged as precursors of current maintenance workers. During this evolution, the wear and breakdown of mechanical parts without prior notice continued as part of everyday factories.
Why this gear has broken? yesterday worked perfectly. Human brain can handle concepts such as linearity of events (seasons, day and night,…) or events that happen more or less at regular intervals. However, these unforeseen drove operators crazy. How can we ensure that gear does not break again? The answer was biologically predictable: “… let’s stop the machine every 2 days (for example) and let’s review gear wear…”
This tradition has resulted in the everyday maintenanceroutine that is applied in industry and in consumer products such as our cars. Our authorized dealer obliges us to make periodic reviews (e.g. each 10,000 km) to check critical elements (brakes, timing belt, …) and change pieces more prone to wear (tires, filters …). This is called preventive maintenance, and is applied in factories and other facilities (e.g. wind turbines) to avoid unexpected breakdowns. However, these faults cannot be eliminated (precisely, they are unforeseen) the only possible reaction is to repair them. This is called corrective maintenance and everyone hates it.
How to stop to all this flood of unexpected breakdowns, repair costs and unnecessary revisions? One of the disciplines with more experience since CARTIF‘s creation is predictive maintenance that seeks to mitigate (it would be unrealistic to assume that we will remove the unexpected) unexpected breakdowns and reduce machines’ periodic reviews. Again, predictive maintenance can be explained as a obvious biological response to the problem of unexpected breakdowns. It is based on the periodic review using characteristic signals of machine’s environment that may anticipate a malfunction. The advantage of this maintenance is that it doesn’t require stopping the machine like with preventive maintenance. For example, an electric motor can have a normal power consumption when it’s correctly operating, but this consumption may increase if some motor’s component suffers from some excessive wear. Thus, a proper monitoring of the consumption can help detecting incipient faults.
Continuing with the electric motor example, what should be the minimum variation of consumption to decide that we must stop the motor and a repair it? Like many decisions in life, you need to apply a criterion of cost/benefit, comparing how much can we lose if we do not repair this motor versus how much money the repair will cost. How to reduce uncertainty in this decision? The answer is a reliable prediction of the fault’s evolution.
This prediction will be influenced by many factors, some of them unknown (like we said it’s something random). However, the two main factors to consider for the prediction are (1) the kind of evolution of the damage (e.g. evolution of damage in a fragile part will be very different from a more or less tough or elastic piece) and (2) workload that the machine will suffer (a fan working 24/7, compared to an elevator motor that starts and stops every time a neighbor presses the button on a floor). A reliable prediction allows the maintenance manager choosing from, together with the forecast of factory workload, the more beneficial option, which in many cases is usually planning maintenance work without affecting production schedule.
Another beneficial effect of predictive maintenance is that a proper analysis of the measured signals provides evidence of what element is failing. This is called fault diagnosis and helps to reduce uncertainty in the more appropriate maintenance action. An example is the vibration measurement that helps distinguishing a fault of an electric motor having an excess of vibration because of an incipient short-circuit or due to a damaged bearing. But that’s the subject of another post.
Agriculture and husbandry are economical activities with high social value in some places around Europe; they have an important share in the economy of many European regions and the European Union devotes a significant part of its annual budget to farming and the related rural world. In spite of this, farmers usually have lower incomes than other citizens in the same social and cultural conditions.
Since the coming of the Enlightenment Age farming has enjoyed technical improvements that increased farming outcomes. During current century, Internet became a widespread technology and the Internet of Things is getting common. Both farming and husbandry will benefit from the Internet of Things. Is about machine communication and it relies on cloud computing and sensor networks. It is mobile, virtual and required reliable and fast data connections. It allows machines and processes to sense the environment and provides the intelligence needed to allow them to optimise by themselves.
Precision farming may be the first application of Internet of Things in farming. The key is to install sensorsto gather data from all the farming processes and to make decisions based on data in an automated way. Soil, plants, livestock, machines, weather can be monitored and actions can be taken to reach exploitation targets in an optimal way, as we reported here.
Although IoT can improve farming activity, we must keep in sight the prices farmers are payed depend on the market. Currently in Europe there is a market deregulation and therefore farmer incomes depend on the market whims. In this scenery, to organise the offer could help farmers to preserve their interests. Could IoT help to organise offer?
Imagine a region where all the farms use the IoT in their everyday activities. They use it to efficiently develop their work and they measure all the important parameters that allow knowing their state and performance. Imagine now that all the farms are connected and share the information gathered by the sensors. Finally, assume the network has intelligence.
Besides the farms information, that artificial intelligence receives information about who and where are the ones that potentially would buy farms products, how much the pay, how is production in other competitor regions, what are the forecasts for market and weather. Putting together all that information, that artificial intelligence would manage the farms by suggesting farmers different operations in order to maximise the delivery price. For instance, the artificial intelligence using available information may conclude that the maximum price for a given product could be reached if certain amount of tons is offered to a defined buyer a precise day. Among all the farms in the network, the artificial intelligence would choose those where the product is in the optimal maturation moment and would inform the farmers about the circumstances so they could proceed with harvesting and transport.
A schema like the one proposed would transform farms into things connected to Internet and smart enough to optimise the farming revenues by themselves. And it would be another technical innovation in the row started centuries ago that would improve farmers live.
The first post about Industry 4.0 indicated the need for key technologies that would make possible the 4th industrial revolution. These key tehcnologies have been called “digital enablers“. Each industrial revolution has had its “enablers”. The first one was made possible by inventions like the steam engine or mechanical loom. The second came started with breakthroughs like electricity or the car assembly line. In the third, disruptive technologies such as robotics, microelectronics and computer networks made their debut.
This post is intended as a shopping list to review those technologies considered highly relevant and key for this fourth revolution. Each brief description is linked to an extended information covered inside our Blog. In next posts we will complete the descriptions to have an overview of the full range of technologies:
Virtual / Augmented Reality: provides information to the operator adapted to the context (e.g. during a maintenance operation) and merged with their field of view.
IoT: internet for virtually any object, in this case, the ones we can find in a factory: a workpiece, a motor, a tool…
Traceability: seeks the monitoring of manufacturing operations (automatic and manual), products as well as the conditions that were used to create them (temperature, production speed…)
Predictive maintenance: an optimized way to perform maintenance in order to avoid unexpected stops and unnecessary waste because of periodic maintenance operations.
Artificial vision: provides the production process visual context information for quality control or assistance in manufacturing (e.g. automatic positioning of a robot to take a piece).
Big Data: generates knowledge and value from manufacturing data as well as other context data (e.g. demand for similar or related products)
3D Printing recreates of a three-dimensional copy of: existing parts, spare parts or prototypes with the same or different scale for review or testing.
Cloud Computing leverages on internet computing resources to undertake storage and processing of large data sets (e.g. Big Data) without the need of investment in own IT infrastructure.
Cybersecurity as physical and logical security measures used to protect infrastructure (manufacturing in this context) from various threats (e.g. a hacker, sabotage, etc).
Cyber-Physical Systems as any complex system consisting of a combination of any of the above technologies seeking improved performance, in this case, of manufacturing.
The strength of these digital enablers is not in their individual features but in their ability to come together. We as engineers love to look for the latest technology trend and then found a problem or area for its application. But to succeed in this revolution, it is necessary to face real challenges within the factories, using innovative solutions, and why not, combining several of the digital enablers shown above. Moreover, this terminology creates a common framework that facilitates a dialogue between technologists and manufacturers for undertaking successful projects seeking to optimize the factory.
If we think, for example, to optimize maintenance operations in a factory, the “predictive maintenance” will be one of the first enablers that comes to our mind. Also, this technology solution will benefit from a connection to a “Cloud computing” system where sensors’ data coming from different factories will be analyzed generating better diagnosis and predictions of the production assets under monitoring. In this type of cloud solutions, however, the security of information transmitted must be ensured via appropriate “Cybersecurity” mechanisms. We will, therefore, generate an Industry 4.0 cybersecure, multi-site, predictive maintenance solution.
The list of presented technologies doesn’t intend to be final. Also, technological evolution is continuous and incredibly fast. Like we have mentioned, the combination of different digital enablers generates a wide range of industry 4.0 solutions. In next posts we will discuss more scenarios where digital enablers can answer to different challenges in manufacturing.
In the 20th Century 80’s decade there was reborn interest in neural networks, both in academia and industry. A neural network is an algorithm that mimics the neural connections present in the neocortex. The interest was motivated by the rediscovering of algorithms to train the networks. Through training, a neural network can learn to do something. And since neural networks are implemented in computers, we have computers that can learn. This is an intellectual ability that people share with monkeys among other animals with neocortex. For this reason, neural networks are the backbone of machine learning, which according to some is part of artificial intelligence.
Neural networks can learn to classify objects and also to reproduce the behaviour of complex systems. They learn by examples. When we want to teach a neural network to differentiate between apples and oranges we have to present it examples of both fruits with a label indicating if it is an orange or an apple. The point is the neural network will be able to correctly classify oranges and apples different to the ones used during training. This is because a neural network does not perform a mere memorisation, but they are able to generalise. This is the key for learning.
But the interest in neural networks that raised up during the eighties faded as the following decade started because more promising machine learning methods appeared. However, a group of indomitable Canadian researchers managed to persevere and transformed neural networks into deep learning.
Deep learning is an algorithm family similar to neural networks, with the same aim and better performance. The number of neurons and connections is higher, but the main difference is the abstraction capacity. When we train a neural network to differentiate between apples and oranges we cannot present the items as they are, we have to extract some features that describe the oranges and apples, as the colour, shape, size, etc. To do this is what in this context we call abstraction. In contrast to neural networks, deep learning is able to do abstraction by itself. This is the reason why deep learning is thought to be able to understand what they see and heard and it is, therefore, a bridge between machine learning and artificial intelligence.
As it happened with neural networks, deep learning has gained huge interest among companies. In 2013, Facebook failed to buy company DeepMind, but Google succeeded one year later when it paid 500 million dollars for it. In case some body missed this irruption of deep learning in the media, it became mainstream in early 2016 when Google DeepMind software AlphaGo beated Lee Sedol, the go champion. This was an unprecedented technical success because go is much harder than chess. When IBM’s Deep Blue won Garry Kasparov in 1996, it used a strategy based on figuring out all the possible short-term movements. However, this strategy is not possible in go because the possibilities are infinite in comparison to chess. For this reason, Google DeepMind’s AlphaGo is not programmed to play go, it is able to learn to play by itself. The machine learns by playing many times against a human player, improving in every game until it becomes unbeatable.
Deep learning is not a secret arcane, anybody who wants to learn it can do it. There are free available tools, like Theano, TensorFlow and H2O, that allows any person with programming knowledge and the concepts in mind to try it. The company OpenAI has freely released its first algorithm, which has been made around the reinforced learning paradigm. There also companies offering commercial products onto which applications can be build. These are the cases of the Spanish Artelnics and the Californian Numenta. Deep learning is being successfully used for face recognition and verbal command interpretation.
Deep learning, besides other machine learning paradigms, could be an important innovation opportunity. It could be one of the tools to unleash the value hidden in the big data repositories. Moreover, in the industrial practice it could be used to detect and classify faults or defects, to model complex systems to be used in control schemes, and in novelty detection.
In July 2015 we were surprised by the news that a robot kills factory worker after picking him up and crushing him against a metal plate at Volkswagen plant in Baunatal (Germany). They insisted the death was a result of human error and not any malfunction on the part of the robot. A Volkswagen spokesman stressed that “the robot was not one of the new generation of lightweight collaborative robots that work side-by-side with workers on the production line and forgo safety cages”.
The application of robots in industrial processes is widespread in industry (mainly automotive), where they perform a multitude of tasks, mostly sequential, repetitive and at high speed. Accidents caused by robots are highly unusual. Many robot accidents do not occur under normal operating conditions but, instead during programming, maintenance, repair, testing, setup, or adjustment. During many of these operations the operator, programmer, or corrective maintenance worker may temporarily be within the robot’s working envelope where unintended operations could result in injuries. During normal operation, robots are confined in safety cages precisely to prevent incidents in contact with humans.
Without adequate safety measures traditional industrial robots can cause serious accidents to people by crushing and trapping (occur when a worker’s limb or other body part can be trapped between a robot’s arm and other peripheral equipment, or the worker may be physically driven into and crushed by other peripheral equipment; it can be deadly, as in the case of Baunatal), collision or impact (occur when a robot’s movements become unpredictable and a worker is struck by the robot) or by projection of materials (occur when parts of the robot , tool or product handled, breaks and fly off and hits a worker).
By rules applicable throughout the EU, it has been mandatory to provide a sufficiently large security perimeter to the entire workspace of the industrial robot that prevents access to the robot when in operation. When it will be necessary to enter to this area, the worker must perform some action to stop the robot, facilitating the access. Harmonised standards ISO 10218-1 and ISO 10218-2, “Safety requirements for industrial robots”, contain the minimum requirements for safe operation of these industrial robots.
This “separation” between workers and robots in an industrial environment is weakened through collaborative robots already available on the market (Universal Robots family of robots, ABB’s YuMi, KUKA’s LBR iiwa…) and the new technical specification ISO/TS 15066:2016, “Collaborative Robots”, that specifies the safety requirements for collaborative industrial robot systems and the work environment. The standard describes different concepts of collaboration and requirements needed to achieve them. The ISO standard also points out that the collaborative operation is a developing field and the new technical specification is likely to evolve in future editions.
Collaborative robots are designed to operate in a shared workspacewith workers without the need for conventional protections, safety cages or safety barriers. The main premise in the design of these robots is the safety of workers (Asimov’s first law of robotics: “a robot will not harm a human being”). These robots are designed to work side by side with workers.
The proximity of workers and robots requires a great safety design based on a combination of mechanical design and control measures, both the manipulator and the workspace. So rather than talking about collaborative robots, in CARTIF we prefer to speak of safe collaboration spaces (collaborative spaces). Besides the robot is safe, so it is the applications and working environments.
To ensure safety can be used different technologies and security measures. Lightweight manipulatorswithout shearing or cutting points, with rounded geometries, smooth surfaces and deformable or elastic components. Speed, acceleration and power can be limited. Current, force, torque sensors can be integrated to detect collisions. Real-time movement of the robot can be adjusted with proximity and tactile sensors. In order to be “aware of the collaborative space” it can be added visual systems based on 2D/3D computer vision technologies.
Usually, collaborative robots are similar to traditional industrial robots but smaller, lighter, less fast and powerful, cheaper and easier to install and configure. These robots do not need to be fast or powerful as they are specially designed to interact with workers. As experts say, in a collaborative space, the worker can bring skills, flexibility and, above all, ability to identify, understand and solve problems, and the robot provides repeatability, accuracy and endurance. Nevertheless, the ISO/TS 15066:2016 standard does not limit the capabilities of the robot in collaborative applications.
Augmented Reality (AR) is a technology that is little by little is making way in our daily lives. In broad terms, it consists in embedding digital information to the reality we have before our eyes through real-time superposition of this information about the image that provides our device, either a Smartphone, Tablet or a Smart Glasses.
Therefore, to use this technology is not necessary to have special devices. Thanks to the mobile devices development, equipment available to everyone as those mentioned before, can be the instrument that will allow access to the world of Augmented Reality (AR). It is true that in the last years are emerging new devices such as the Smart Glasses that are adapted specifically to this type of applications, and allow provide a better user experience by leaving freedom of movements of their hands. It is expected that with the development of new applications these devices will be entering in the consumer market in a generalized manner, as has happened with other devices such as smart watches, or activity tracker wristbands.
There are many applications of Augmented Reality (AR) and are varied. Most of them are oriented to leisure and tourism. As an example, there are applications that just taking a picture of a monument or a work of art, can provide you with information about it, its history, or oddities such as images of their appearance in the past.
New applications are emerging constantly related to leisure, mobility, marketing and advertising, including additional content that improve and enhance the user experience. These contents can be quite varied, including explanatory texts, links to supporting documentation, videos, images, indications for locating places or events, etc.
Equally, new applications are emerging for use in professional fields, such as architecture, to show designs and scale models, marketing and sales, for example catalogs that include QR codes to display additional content, medicine or education, which presents countless opportunities for the expansion of content through videos, tutorials or examples, allowing to learn in a more enjoyable and didactic way.
The industrial environments also are becoming an important field of application of Augmented Reality. Most experts expected that large-scale implementation in the industrial area will take place in the next three to five years as an increasingly robust technology. CARTIF is involved in some projects that apply the augmented reality to maintenance work in the industry.
Through the use of a Smart Glasses, the technicians are able toperform their tasks as they have always done, since they have complete freedom to use their hands, and at the same time they can access, for example, to technical specifications sheets for particular equipment, to the drawings of a facility, or the operating and malfunction history for a machine.
It would also be possible to guide the maintenance operator in performing a particular task, indicating the steps to be carried out and checking when and how has performed them. In the same way, they could be very useful in the training of novice operators for the accomplishment of these tasks, or even that the managers can monitor and evaluate the performance of operators.