In diverse situations I have been asked by people from different backgrounds about mechatronics. If you think it is difficult to explain to a friend of yours, convince a management board to adopt this strategy is not an easy task either. Here I have collected an overview over mechatronics, fuzzy logic and neural networks, what for me are the pillars of modern manufacturing.
Probably the most commonly definition for the word mechatronics is the one by Cetinkunt ‘The mechatronics field consists of the synergistic integration of three distinct traditional engineering fields for the system level design process…’
‘The old model for an electromechanical product design team includes:
1.- Engineers who designs the mechanical components
2.- Engineers who designs the electrical/electronic components
3.- Engineers who designs the computer hardware and software implementation to control the product
A mechatronics engineer is trained to do all of these three functions’ Cetinkunt, S (2005)
The rising of this technology have to be seen in the widespread availability of low cost microcontrollers, but it is also the natural evolution of the process of engineering disciplines applied to day to day electromechanical systems. Intelligent systems are usually asociated with human beings, this is directly related with mechatronics. Skeleton and muscles would be the mechanical discipline, while eyes, other sensors and connection between them would relay on electric/electronic disciplines. The functions of the brain, information processing and genertion of stimulus are linked with computer engineering.
‘Mechatronic strategies have been shown to support and enable the development of new products and markets such as the compact disc player as well as through enhancing existing products while responding to the introduction of new product lines by a competitor….the motivation for the adoption by a company of a mechatronic approach to the product development and manufacturing must be one of providing the company with a strategic and commercial advantage either through the development of new and novel products, through the enhancement of existing products, by gaining access to new markets or some combination of these factors’ Bradley, D (2000)
On the left hand side combination of resources implicit in mechatronics technology. On the right hand side a mechatronics aplication. The robot simulates the movements of a human hand captured by sensors, then processed, and reproduced by the robotic hand.
Mechatronics as an organizational strategy?
What benefits does it bring? What new oportunities? Is this a new and risky technology in the consumer electronic market? Some UK engineers visited Japan in 1993, searchig where mechatronics has taken the japanes industry since the company Yaskawa introduced the terminology in 1972. The results of this visit were presented in a conference at the at the Institution of Electrical Engineers in October 1998. Take a tour of mechatronics in Japan
‘The 1993 team recognized that mechatronics was the multidisciplinary combination of the “brains” of electronics, control and software with the “muscle” of mechanical systems. This approach was important so that Japan could add high value, functionality and flexibility to products, particularly as there were few natural resources. This was still relevant in 1996, but with significant improvements and additions to the technology involved. Tolerances in products had reduced to a few microns, rather than tens of microns, brought about by improvements in manufacturing as well as further introductions of micromachining, which then impinged on the assembly processes needed. Products were gaining in complexity, mechanically and electronically. Competitive pressures on the economy had resulted in mechatronic products incorporating new technologies in the multidisciplinary approach. The two most obvious were chemical and communications systems. As well as a refocus on core technology and business with less speculative research and developments, several companies had also looked for market diversification with their new technology’ Hollinghum, J. (1999)
Evidence of this raising technology can be seen in an internationsl initiative for research and developement for the next generation of manufacturing and processing technologies, where companies from the 5 continents are participating. This initiative is IMS, and aims to reduce R&D costs, and expand international markets.
‘Intelligent manufacturing systems are already widespread, and this optimization of the process through the use of intelligent systems is continuing, and is the theme of the multinational, industry-driven intelligent manufacturing systems (Intelligent Manufacturing Systems) initiative, which started in 1995. Currently projects with a value of $240 million are being undertaken through the IMS’ Hollinghum, J. (1999).
‘A neural network,… attempts to reproduce electronically through the use of artificial ‘neurons’ the functions of the human brain. The resulting network consists of the parallel combination of a large number of simple processing elements, essentially summing junctions, which, instead of being programmed, are trained how to carry out a particular task. This ability of neural network to learn and to modify its response as a result of training has been a major factor in their development…’ (Mechatronics, D, Bradley pp147).
Probably, a better definition is that one given by Gurney ‘A neural network is an interconnected assembly of simple processing elements, units or nodes, whose functionality is loosely based on the animal neuron. The processing ability of the network is stored in the interunit connection strengths, or weights, obtained by a process of adaptation to, or learning from, a set of training patterns’ Gurney, K (1997)
The analogies with the animal brain of this structures can be sumarized. Multiple inputs are connected to a node, emulating the multiple dendrites connected to a cell body (neuron). This cell body is in charge of weighting and process the info based on a learned patron, and produce a signal driven to the output. This ouptut (can be referred as the axon in the animal neuron) it is usually connected again to other inputs where the process continues. One of the interesting aspects of neural networks is their capability of ‘learning’. ‘In terms of processing information, there are no computer programs here the “knwoledge” the network has is supposed to be stored in its weights, which evolver by a process of adaptation to stimulus from a set of patter examples. In one training paradigm called supervised learning’ Gurney, K (1997) If a new pattern arises, the network is able to classify the core of the problem, and generate a correct answer. This process is called generalization.
Despite in the brain, different zones are in charge of different processes, if one of this parts is lost or damaged other parts of the brain can learn to carry that functions. At the same time, animal brains are capable of processing the correct information in noisy environments (the voice of the teacher can be disguissed even when a heavy machine is working outside generating noise). The same aspects are faced with artificial neural networks, that are able to provide a correct output also when the input is incomplete, noisy or some parts of the network are damaged. Thanks that the knowledge is spread over the network, and tolerance values are inherited in the patterns. It can not be undervalued the capability of the neural networks for processing different tasks at the same time or parallelism.
It is necessary to to clarify that for building a neural network it is not possible to grab a bunch of ressistors and IC’s for this (Although they can be necessary in some part of the process). Usually the networks are simulated on a PC or workstation, or at least, in a microprocessor.
Neural Networks nowadays provide enhancement of many processes as: speech recognition, image processing and analysis, processing of sensor data, market forecast. Let’s see a proper example. In some laboratory, are investigating over green mass (plant) generation. The traditional process from a vegetable have to be simulated like photosynthesis, transpiration, water released through the leaves, water absoved via the roots, carbon consumption, and others. The cuantities of water, generation of green mass and carbon consumption differ very much on the kind of plant, age and environmental conditions. A neural network has been proposed.
The paterns are set in base of the kind of plant, and their performance under determinated conditions and how other conditions could modify the devopement of the plant. The inputs that can be seen on the left of the diagram can be taken from sensors in a green house, or can be simulated from hypothetical scenarios. The learned patterns relay in the central nodes, while on the righ we have the outputs. This information could be used to adjust the conditions of the green house for an optimum developement of the plants (starting again the process as new inputs), or just evaluated over the desired performances.
Product development with Neural Networks
‘Aston Martin’s 2005 version of its DB9 has a mind of its own. The luxury sports car is equipped with an artificial neural network – a microchip modelled on the human brain – that has been “trained” to monitor its engine and detect misfires in its cylinders… The regular tightening of emission controls means car manufacturers such as Aston Martin are trying to find increasingly fast and accurate ways of detecting misfires. Aston Martin has so far used a conventional microprocessor running an algorithm to process data picked up by sensors around the engine to detect misfiring – but this method is becoming impractical… An artificial neural network can overcome this problem because it detects misfires differently. It looks at patterns in the data it receives and “spots” misfires by recognising misfire patterns it has already learned.’ Rubens, P. (1995)
‘Fuzzy logic is a broad theory including fuzzy set, fuzzy logic, fuzzy measure and others… Let us look at the application aspect of fuzzy logic. The earliest and most successful applications have been in the control field’ Tanaka, K. (1991)
Our environment, the world around us is analogue, and so are human feelings and language. We can feel very cold or really hot, depending on the situation (20 ºC can be hot in winter time, but not in August), and our conversations are plenty of ‘the tube was packed this morning’ or ‘there were just a few people in the pub’. Despite this ‘cold’ or ‘hot’, ‘packed’ or ‘few’ are not exactly determined, aproximate boundaries are asociated with this expresions, based in our experience. Fuzzy logic attemts to imitate this this human reactions to such undetermined information. For doing this, the information or universe must be classified into different sets. This sets can be related between them with different definitions as ‘union’, ‘intersection’ or ‘complement’. The logic will be applied based in if the values or membership functions belong to one set or not, typically assigning values of ‘1’ or ‘0’. Without forgetting the fuzzy characteristic, this values can be ‘totally 1’ or ‘just 1’ depending on the affinity given by the membership function. To perform fuzzy reasoning, ‘IF…THEN…’, ‘OR’, ‘AND’ rules are applied.
e.g. IF x>α THEN x=B
IF x<α THEN x=A
Fuzzy Logic configuration for two inputs and one output in two different situations
‘Fuzzy controllers have been shown to be particularly suited to applications where the process to be controlled is ill defined, is not well understood or changes with time but for which a clear statement or requiered performance indicator exists andwhere human operators have developed a control strategy based on their understanding of system behaviour’. Bradley,D. 2000
‘A turning point for fuzzy logic came in 1974. Ebram Mamdani of the University of London applied fuzzy logic to controls for the first time the control of a simple steam engine’ Tanaka, K. 1991
It was not untill the early 90’s when fuzzy logic was applied to home electronic products and the general population became aware of fuzzy systems. Nowadays, Fuzzy logic is applied to control washing machines, dishwashers, ABS in automobile industry, elevators digital cameras and many others.
e.g. The Fuzzy Washing Machine: A fuzzy controller for a washing machine can work as follows, the washing machine, always works during 10 minutes, to ensure that the user is satisfied, if the clothes are dirty or not. Then, it adds minutes to this base of 10 minutes, depending on how dirty the clothes are, adding 2 minutes to the program for each level of dirt in the water. After this process another fuzzy system gets involved. The washing machine rinses for 5 minutes, if too much washing powder has been added by the user, it rinses for two extra minutes untill to ensure the clothes are free of soap. In this example, the universe could be the clothes, and the sets are going to be how dirty the clothes are, and how much shoap we have added. Different information is gathered for the different sets (how dirty, how much shoap), and wheighted depending on the importance (this process is named fuzzyfication). An appropiate washing program is designed by a fuzzy control system (extra time, less rinsing), and then, through a process of deffuzyfication, actions are taken in order of the previous sets (continue turning, taking less water).
We can see how a state of the art fuzzy products company classifies one of their washig machine: ‘This machine’s intelligent fuzzy logic will detect when the laundry is out of balance and re-arrange it, to ensure minimum wear and tear to the drum bearings. The fuzzy logic also detects if too much detergent has been added and adds extra rinses if required.’ visit fuzzy logic in LG
- Bradley, D. Seward, D; Dawson, D; Burge, S. (2000) Mechatronics and the design of intelligent machines and systems. Stanley Thornes, Ltd. Cheltenham
- Cetinkunt, S. (2007) Mechatronics. WILEY, Hoboken.
- Gurney, K. (1997) An introduction to neural networks. University College London (UCL) Press, London
- Hollinghum, J. (1999) How are we doing in mechatronics? in Assembly Automation, Vol 19, Num1. MCB University Press
- Rubens, P. (2005) The sexy model with a brain takes to the road in Financial Times. (Published Feb, 18 2005), London. Financial Times web
Tanaka, K. (1991) An introduction to fuzzy logia for practical applications. Springer, New York