8
24
24
PCB
10
11
< 1
140
1.3
< 1
2
1.6
2
2
2
ACD
13
16
< 1
210
4.5
1
15
2.4
5
8
15
Inverter
29
38
< 1
4830
34.8
10
1280
33.0
19
57
76
Rectifier
20
17
< 1
53
3.8
3
160
10.8
15
40
64
Transformer
4
8
< 1
70
0.7
< 1
4
1.2
4
2
2
Backup Components
25
30
< 1
856
14.8
5
864
25.1
9
42
210
Main Bearing
3
5
< 1
191
1.7
< 1
3
2.4
3
0
0
As the results show while there are maximum computation times of around five seconds, the median of the distributions is located around and below ten milliseconds. Further, on average we can record runtimes of fewer than half a second. These runtime results as well as the automated modeling argue in favor of our approach as a feasible method for diagnosis in practical applications.
50.5 Conclusion
In this chapter we introduced the concept of Diagnosis As A Service where we argued for the necessity of the availability of diagnosis functionality especially in case of Industry 4.0 comprising interconnected manufacturing plants and devices to be flexible enough for being adapted accordingly to specific customers’ requirements. These underlying system requirements, i. e., distributed manufacturing and increased flexibility, demand a flexible diagnosis infrastructure. To serve this purpose we discussed the general diagnosis methodology abductive diagnosis where cause‐effect knowledge is used for identifying root causes for given symptoms. Such cause‐effect knowledge is often available in practice for engineered systems thus making the approach interesting for practical applications.
Besides discussing the basic principles of abductive diagnosis in this chapter, we also introduce requirements for implementing diagnosis as a service and present its underlying architecture. In addition, we recall the use of abductive diagnosis in an industrial context in order to show its feasibility for fault localization for several systems used in practice. In all cases except one the implemented diagnosis approach required less than one second for computing diagnosis candidates. Based on the diagnosis candidates the abductive diagnosis approach also allows for supporting the user in identifying the real root cause via asking questions about new symptoms. In order to come up with a diagnosis service the underlying cause‐effect knowledge has to be formalized, which can be done using a table as described in this chapter.
Acknowledgements
The work presented in this paper has been supported by the FFG project Applied Model Based Reasoning (AMOR) under grant 842407 and the SFG project EXPERT.
References
1.
Wikipedia, “Industry 4.0,” 17 06 2016. [Online]. Available: https://en.wikipedia.org/wiki/Industry_4.0.
2.
R. Reiter, A theory of diagnosis from first principles, 1 ed., Artificial Intelligence, 1987, pp. 38–44.
3.
J. De Kleer and B. Williams, Diagnosing multiple faults, 32 ed., vol. 1, Artificial Intelligence, 1987, pp. 97–130.CrossrefMATH
4.
G. Friedrich, G. Gottlob and W. Nejdl, “Hypothesis classification abductive diagnosis and therapy,” in Proceedings of the International Workshop on Expert Systems in Engineering, vol. 462, Vienna, Springer Verlag, 1990.
5.
L. Console, T. Dupré and P. Torasso, “On the relationship between abduction and deduction,” Journal of Logic and Computation, vol. 1, no. 5, pp. 661–690, 1991.CrossrefMATH
6.
R. Koitz and F. Wotawa, “From Theory to Practice: Model-Based Diagnosis in Industrial Applications,” in Proceedings of the Annual conference of the PHM Society (PHM-2015), San Diego, CA, USA, 2015, pp. 197–205.
7.
Y. Duan, G. Fu, N. Zhou, X. Sun, N. C. Narendra and B. Hu, “Everything as a Service (XaaS) on the Cloud: Origins, Current and Future Trends,” in Proceedings of the IEEE 8th International Conference on Cloud Computing, IEEE Press, 2015.
8.
B. Buchanan and E. Shortliffe, Rule Based Expert Systems: The MYCIN Experiments of the Standord Heuristic Programming Project, Addison-Wesley, 1984.
9.
P. Jackson, Introduction To Expert Systems, Addison Wesley, 1998.MATH
Further Reading
10.
M. Turner, D. Budgen and P. Brereton, Turning software into a service, 10 ed., vol. 36, IEEE Computer, 2003, pp. 38–44.Crossref
Part XII
Intelligent & Autonomous Enterprise
© Springer-Verlag GmbH Germany 2018
Claudia Linnhoff-Popien, Ralf Schneider and Michael Zaddach (eds.)Digital Marketplaces Unleashedhttps://doi.org/10.1007/978-3-662-49275-8_51
51. Preface: Intelligent & Autonomous Enterprise
Martin Hofmann1 and Stefan Meinzer2
(1)Volkswagen Group, Wolfsburg, Germany
(2)Volkswagen Data:Lab, Munich, Germany
Martin Hofmann (Corresponding author)
Email: [email protected]
Stefan Meinzer
Email: [email protected]
51.1 Introduction
The automotive industry is in the midst of a disruptive change. While the car itself was in focus of the customer and the manufacturer so far, digital customer services around customers’ mobility will be the future core value. The car must fit into the customer’ digital ecosystem providing maximum convenience and safety.
The exponential growth of computing power, as illustrated by Fig. 51.1, leads to unforeseen possibilities in the automotive industry. Intelligent and autonomous cars; new mobility services due to artificial intelligence; cross‐industrial knowledge transfer from other business areas like the health care [1] to provide the best safety features are three examples for the disruptive change. Therefore, new technologies, the corresponding skills and new IT models must be implemented. The automotive industry is following this path already [2]. In order to remain competitive, the companies must satisfy customers’ expectations within these areas. Maximizing customer satisfaction by services is one core competence of the automotive industry [3]. While the goal to secure highest satisfaction remains the same, the services themselves will change towards data driven services.
Fig. 51.1Exponential growth of computing power. (Kurzweil [4])
According to Moore’s Law, the possibilities of digital applications will be far reaching due to the almost exponential increase of calculations per second achieved by high end computer systems [5]. In future, the automotive industry will be one of the largest data generators and data provider within the service industries [6]. Latest car models are already creating more than 25 Gigabytes of data every hour. In the era of autonomous driving,
1 Terabyte is estimated to be produced every hour by sensors and services of the cars [7]. This estimation ends up in 2 Petabytes of data per car per year that need to be analyzed in real‐time in order to create customized services. With the growing importance of battery cars, sensor fusion will be one central area in order to apply machine learning algorithms or artificial intelligence [8]. Internal sensor data from cars must be combined with external data sources from various digital ecosystems in order to secure maximum mobility of pure electric cars.
The traditional automotive industry has to overcome various challenges in the era of disruptive change. The most important areas that are touched by the fast growth of technological development are presented in the following section.
51.2 The Disruptive Change of the Automotive Industry
The change within the automotive industry due to new connected technologies is already proceeding [28]. In general, the disruption describes the move from cars towards mobility. The first step of the change is the development of connected cars, as illustrated by Fig. 51.2. The path reaches from connected car via the connected customer and the connected ecosystem towards connected transportation [9]. Remarkably, new competitors of the traditional automotive industry already entered the market, such as Uber, Google or OnStar. Consequently, the existing manufactures must strengthen their core competences – producing cars and services around the cars and customers – enhanced by latest available technologies.
Fig. 51.2Path of disruption of the automotive industry. (Seiberth [9])
The change due to digitalization affects the complete supply and value chain of the automotive industry [10], as illustrated by Fig. 51.3. In the following, the focus is set on the most important business areas.
Fig. 51.3Overview on business areas of the automotive industry affected by the change of the digitalization. (Wedeniwski [10])
51.2.1 Autonomous Driving
The total number of cars is estimated to be more than 100 mio. worldwide [11]. A huge challenge and opportunity. In order to predict traffic or provide high safety standards, sensor based functionalities are already implemented that transfer signals for instance from radar or cameras. Machine learning models are implemented for features like traffic sign recognition or lane assists [11]. However, it is mandatory to enhance these functionalities to reach the goal of selling self‐learning cars. This is fundamental to be able to provide customer services, such as the mobility on roads or at a time with low traffic. One vision of the future, related to autonomous cars, is the elimination of traffic jam. Based on car‐to‐car communication, vehicles will be intelligent enough to navigate in the most effective way. As the navigation of these cars will be based on applications related to artificial intelligence, the safety and the efficiency of driving will continuously increase.
From a customers’ perspective, the value of autonomous driving is individuality. Services while driving and the driving behavior itself will be personalized. For the automotive industry it means Petabytes of data that need to be analyzed in real‐time. Latest technologies, such as quantum computing, will be needed to handle these optimization problems from an analytics perspective. To realize such new capabilities, companies need to change their existing structures, especially the IT, and develop new skills [2]. Companies, such as Volkswagen, are already pioneering these new technologies for example using quantum computers for traffic flow optimisation [26].
51.2.2 Smart Cities
The urbanization yields to new concepts of living. Smart homes and smart cities are the two main pillars coming together with the urbanization and the digital ecosystem. The competitive advantage of companies offering digital services will be the degree of integration of these services into the connected ecosystem of the customer [9]. The automotive industry will play a significant role in designing smart cities. Intelligent parking systems, battery charging on demand or driving home functionalities in autonomous driving cars are just three representative examples [12]. Finally, the goal of smart cities is to increase the humans’ quality of life. Taking the technological perspective, this goal can only be reached by integrating and combining the countless amount of transferred data [13]. Consequently, sensor fusion will be one of the most challenging and important capabilities to remain competitive and to enter this market. The city of Barcelona, as an example, is already changing their whole infrastructure towards a smart city concept. Today, more than 320 millions of sensor signals received from more than 1,800 different kinds of sensors produce more than 8 Gigabytes of compressed data each day [13]. The significance for the automotive industry shows the critical topic of parking. In mega‐cities, such as Barcelona or Singapore, every third driver is searching for a parking spot for around 8 min [12, 14]. In order to increase quality of life as mentioned earlier, smart parking will be a core digitalized service for the automotive industry. Therefore, internal sensor data from the cars, infrastructure signals from the cities, geo‐data, individual customer preferences and many more signals must be fused in order to offer the driver the best spot at the best location at the right time at the best price. Such scenarios are game‐changer for traditionally acting manufacturers, technologically as well as from a structural‐process perspective. While the analysis and storage of the internal data is already a core discipline, the analysis of merged internal and external data requires modern machine learning methods and state of the art computing technologies. In order to provide real‐time services in smart cities, technologies like scalable cloud computing must be a standard element of the future IT environment.
51.2.3 After‐Sales
The After‐Sales business is the most important, established core business of the automotive industry. The margin and profit that is generated within After‐Sales is already significantly higher than within sales [15]. An estimated revenue of 6–8 billion dollars is annually generated by After‐Sales services only in the US [16]. As the importance of the product car itself will shift towards the importance of digital customer services around mobility, the significance of the After‐Sales business will be even higher [17, 18]. Various applications based on machine learning techniques in order to improve the After‐Sales business are already established, such as the prediction of churning customers, the modeling of customer satisfaction or the rule‐based recommender systems [19–21]. However, following the trend of connected cars, autonomous driving and the integration of the mobility in customers’ digital ecosystem, new and individualized customer services will form the future After‐Sales business [27]. The classification of potentially dissatisfied customers before the customer-service interaction ends, is one representative example [27]. Predictive maintenance is one representative example. Generally, the goal of the After‐Sales business will be mobility maximization, downtime reduction, and individualized service packages while the customer visits a dealer [22]. The big advantage of the After‐Sales sector and thus the traditional automotive manufacturer, compared to the above mentioned new areas and competitors, is the fact that the fundamental basis for these new services is the already existing data. Years of service information, repair campaigns, etc. identify a vast amount of information [23] the automotive industry must make maximum value out of it. Therefore, new business approaches based on latest technologies are necessary to generate this value. Autonomous marketing campaigns, pricing, finance or controlling are exemplary areas where artificial intelligence needs to be applied to realize these new business potentials.
51.2.4 Production
The manufacturing process of the automotive industry is an area that produces a vast amount of data. It is an area that is changing rapidly towards an automated and autonomous industrial sector [2]. Due to the fast development of research in robotics and in particular human‐robotics cooperation [24], this field will require special technological attention in the era of Industry 4.0. The evolution of robotics research yields to faster production processes with lower probability for quality failures. Addi
tionally, from an economic perspective costs might become reduced significantly [25].
The integration of high automated manufacturing processes brings additional requirements for existing manufacturers. The importance of predictive maintenance, as already introduced as a customer service, will be fundamentally important in the field of robotics. With visionary applications, such as integrated factory optimization, new potentials arise [27]. Predicting potential downtimes, quality failures or missing parts for production are examples where the application of machine learning techniques is the basis. In order to cope with the increasing variety and complexity of car configurations, methods related to artificial intelligence will be needed in order to speed up the manufacturing process and increase the output quality.
51.3 Intelligent & Autonomous Enterprises to Remain Competitive
To remain competitive in the future, fundamental key requirements need to be fulfilled: Invest in new products, such as autonomous driving and electric cars; integrate the mobility and thus the car itself in customer’ digital ecosystem; offer digital customer services to maximize the mobility, minimize downtimes and offer individualized predictive maintenance; automate the manufacturing process to make it more efficient. Three representative examples to improve established processes fundamentally within an autonomous enterprise are the following.
First, the planning of sales activities is a big potential of autonomous enterprises [27]. Based on sold cars, future customer preferences can be predicted. Even more, the configuration of cars that have the highest probability to be sold can be determined. Calculating similarity matrices, future trends can be predicted. Adapting anomaly detection methods, shifts in model selections or configuration choices can be identified even earlier than humans could do it today. The potential of an autonomous sales planning is the time saving compared to the expert knowledge that is needed today. The model portfolio can be suited to country and customer groups as soon as switching customer preferences are detected [27].
Digital Marketplaces Unleashed Page 78