Automated application service detection for configuring in dustrial communication networks

FIELD OF THE INVENTION

The present invention relates to an automated method and an arrangement for configuring an industrial communication net- work using a digital twin.

BACKGROUND OF THE INVENTION

(Communication) Networks are a medium to transfer data. Net- works, however, vary based on the volume of data being trans ferred. Industrial communication Networks refer to networks that deal with transfer of data on a large scale. This means, they allow to connect various devices across large spaces and enable communication between them by allowing to transfer huge chunks of data between them.

Industrial applications usually have very specific require ments to the industrial communication communication system, such as determinism in communication, reliability, and many more. Such requirements are very process specific and use case specific, and usually cannot be easily "guessed" but need careful consideration with usually a large amount of manual work. Especially for application development, but also for system integration it is often not possible to completely capture or to easily describe by simple values. Unprecise or incomplete descriptions of requirements, or even wrong param eter settings will result in unreliable operations of indus trial applications. Another "extreme " but typical setup is the parameterization of the system with far too high require- ments, leading to an over-reservation of system resources.

The following small example describes problems arising for a typical question: When does a device interface need to send a data packet, so that arrival at a time X decided by a TSN (= time-sensitive networking) controller is adhered to?

Typical tasks and problems are:

Application or application end point A is a sensor and wants to send packets to application or application end point B with the sent interval of time X.

The application or application end point B is a control application and wants to send packets to an actuator C every Y Nanoseconds - within a given cycle time of the application.

Ideally, endpoint B intends to send after a packet from end point A is received

The underlying network must serve both requests, but the requirement to send after receiving a packet from end point A is not taken care of.

Current solutions require the applications or a configuration tool to send a service request to the control entity (= net work controller). The creation of such a service request is not easily possible in automation and production environ ments, as industrial applications have a very specific set of requirements - quite different from the requirements we know from the IT world.

To create such a communication request, the application de veloper usually provides mechanisms to capture the communica tion requirements (e.g., through user input) so that the ap plication can create and send the service request. An automa tion of that procedure is currently not available. The cur rent procedure is even more complex and error prone.

When formulating the requirements on the communication ser vice, the application developer must take timing constraints caused e.g., by network interfaces, network stacks, operating system, etc. into account, which is often difficult or impos sible for him. Particularly considering the development of portable apps, where the destination hardware and eventually operating system is variable.

SUMMARY OF THE INVENTION

The objective of the present invention is to provide a solu tion for an automated detection of application service in in dustrial communication networks.

To accomplish the objective, the present invention provides a solution according to the independent claims. Advantageous embodiments are provided in the dependent claims.

According to an aspect of the invention an automated support method to detect application services and their parameters in such an industrial communication system is introduced. The proposed method also gives a network controller information on the type of application that is sending the data packets. As typically network controllers (e.g., a TSN controller) does not automatically know how to handle and parameterize the data traffic of an application. The proposed method will detect data packet timing of the industrial applications' communication endpoints (e.g., sensors, actuators, applica tions). This parameter detection can be used in all types of setups, such as systems that follow the perceive - analyse - act paradigm, and in systems with asynchronous communication.

The mechanism described in this invention allows an automated assessment of application requirements, that must be ful filled by the industrial communication system. Thus, applica tion developers as well as system integrators can easily de ploy applications in industrial communication environments with a reduced risk of system misconfiguration. The invention claims an automated method for configuring an industrial communication network, comprising the following steps: deploying an application, which is deployable to the in dustrial communication network, to a digital twin, whereby the digital twin is set up to represent the application, to represent a digital representation of an industrial communi cation system which is comprising the industrial communica tion network, devices connected to the industrial communica tion network, and processes, and to represent physical inter actions in the industrial communication system, running the application on the digital twin, whereby data of the application's communication traffic is collected, application's end points and their corresponding com munication characteristics in the industrial communi cation network are initially detected and classified based on the collected application's communication traffic, the initially detected communication characteristics of the communication links used by the application are changed until reaching limits of pre-defined applica tion performance quality, and from thereof determined communication characteristics network parameters for the communication service re quired by the application are derived, and configuring the industrial communication network based on the determined network parameters.

This means, that the communication characteristics are changed until the application performance degrades under pre defined thresholds of required communication service require ments or fails to function properly, thereof finding the lim its of the communication characteristics the application can stand without degradation of process quality.

In other words, the application's communication is "pushed to the limits" and the Digital Twin will notify as soon as the application fails, e.g., a control loop is not able any more to control a physical process in the digital twin, results of processes are outside the required operational envelope.

The advantage of the invention is that by assessing the be havior of an application in a digital twin environment, the network and communication effects, the host system where ap plication endpoints are running, and other applications run ning in the system and causing data traffic are considered.

In a further embodiment of the method the communication char acteristics comprise delay, jitter, and packet loss.

In a further embodiment of the method the application perfor mance quality comprises pre-defined quality parameters and/or a pre-defined functional envelope.

In a further embodiment of the method changing the communica tion characteristics comprises adding of additional delay on links, creating packet loss, and/or re-aligning of send times of different streams in relation to each other by adding ad ditional communication services requiring resources from the industrial communication network or re-arranging network schedules.

In a further embodiment of the method devices comprise sen sors and/or actuators.

In a further embodiment of the method the application perfor mance quality (= QoS parameters of communication streams) is continuously monitored and in case of deviations from pre defined minimum requirements an alarm is generated.

In a further embodiment of the method status feedback data of real machinery the application is used for modifying the com munication characteristics. Whenever additional applications or components are added to a plant using the industrial communication network, or existing ones are to be modified or deleted, planning and design phase might be re-entered.

The invention further claims an arrangement comprising an in dustrial communication network and one or more computational devices, whereby the one or more computational devices comprises a digital twin, a Network Information Provider and Monitoring module, Network Traffic Classification module and a Network Controller module, whereby the digital twin is setup to run a deployed ap plication, which is deployable to the industrial communica tion network, and to represent the application, to represent a digital representation of an industrial communication sys tem which is comprising the industrial communication network, devices connected to the industrial communication network, and processes, and to represent physical interactions in the industrial communication system, whereby the Network Information Provider and Monitoring module is setup to retrieve network data traffic characteris tics from a communication network, a digital representation of such a communication network, or a simulated or emulated network,

It can retrieve measurement data from network devices and ap plication endpoints such as hosts or virtualized compute en vironments, such as virtual machines or software containers. Network data traffic characteristics can be obtained through active or passive monitoring.

Typical network data traffic characteristics are end-2-end (e2e) delay of data packets exchanged between application endpoints, bandwidth used by an e2e data stream, jitter of the packets within a data stream, packet losses, packet siz es, and many more. whereby the Network Traffic Classification module is setup to classify endpoint types of the industrial communica tion network and to classify detected application data traf fic,

Classification of endpoints can be done based on its data traffic characteristics and is a well-known domain in tradi tional Internet data traffic engineering. Nevertheless, these traditional approaches have their limitations and the compo- nent in this invention is designed to use also more recently developed approaches using artificial intelligence and ma chine learning, e.g., Thuy T.T. Nguyen, G. A. (FOURTH QUARTER 2008), A Survey of Techniques for Internet Traffic Classifi cation using Machine Learning, IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 10, NO. 4, to capture the special nature of industrial communication data traffic, and to take the spe cial needs of the end points into account. - whereby the Network Controller module is setup to con figure network devices in the digital twin as well as in the industrial communication network to embed data flows, and is setup to detect active topology, including network device properties and capabilities,

Configuration of network devices is typically done based on communication service requests coming from an end station (via User-Network-Interface, UNI), or via its northbound in terface from a central management unit. whereby the Industrial communication Network is a real net work setup to run and operate an industry plant,

Here the real applications are controlling the real industri- al processes and actuators based on the data from sensors, etc. whereby the network traffic classification module is setup to initially detect and classify application's end points and their corresponding communication characteristics in the industrial communication network based on the collect ed application's communication traffic monitored by the Net work Information Provider and Monitoring module, whereby the digital twin is setup to change the initial ly detected communication characteristics of the communica tion links used by the application until reaching limits of pre-defined application performance quality, whereby the network controller module is setup to derive from thereof determined communication characteristics network parameters for the communication service required by the ap plication, and is set up to configure the industrial communi cation network based on the determined network parameters.

In a further embodiment of the arrangement communication characteristics comprise delay, jitter, and packet loss.

In a further embodiment of the arrangement the application performance quality comprises pre-defined quality parameters and/or a pre-defined functional envelope.

In a further embodiment of the arrangement changing the com munication characteristics comprises adding of additional de lay on links, creating packet loss, and/or re-aligning of send times of different streams in relation to each other by adding additional communication services requiring resources from the industrial communication network or re-arranging commication network schedules.

In a further embodiment of the arrangement devices comprise sensors and/or actuators.

In a further embodiment of the arrangement the Network Infor mation Provider and Monitoring module is setup to continuous ly monitor the application performance quality and in case of deviations from pre-defined minimum requirements to generate an alarm. In a further embodiment of the arrangement the digital twin is setup to use status feedback data of real machinery the application is dedicated for to modify the communication characteristics.

Further benefits and advantages of the present invention will become apparent after a careful reading of the detailed de scription with appropriate reference to the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 shows a block diagram of the arrangement, and

FIG. 2 shows a flow chart of the configuration of an in dustrial communication network.

DETAILED DESCRIPTION OF THE INVENTION

As an embodiment of the invention FIG. 1 shows a block dia gram of the communication arrangement. The arrangement has three major components, a Network Information Provider and Monitoring module NIP, a Network Traffic Classification mod- ule NTC, and a Network Controller module NC, interacting with other components in an industrial control system, such as ap plications A, a Digital Twin DT, and the industrial communi cation network IN, itself. If the digital twin DT environment is not available and the real system is in testing and con- figuration phase, of course also the real system can be used for the training. All modules and the digital twin DT can be implemented in one or more computational devices CD.

Network Information Provider and Monitoring module NIP is a software-implemented component that can retrieve network data traffic characteristics from a communication network, a digi tal representation of such a communication network, or a sim ulated or emulated network. It can retrieve measurement data from network devices and application endpoints such as hosts or virtualized compute environments, such as virtual machines or software containers. Network data traffic characteristics can be obtained through active or passive monitoring. Typical network data traffic characteristics are end-2-end (e2e) de lay of data packets exchanged between application endpoints, bandwidth used by an e2e data stream, jitter of the packets within a data stream, packet losses, packet sizes, and many more.

Network Traffic Classification module NTC: In industrial com munication systems exist many different types of devices D (e.g., sensors, actuators, industrial PCs) as well as differ ent applications A (e.g., SCADA software, control applica tions, monitoring data streams). The Network Traffic Classi fication module NTC can classify the endpoint types and can do a classification of the detected application data traffic. Classification of endpoints can be done based on its data traffic characteristics and is a well-known domain in tradi tional Internet data traffic engineering. Nevertheless, these traditional approaches have their limitations and the compo nent in this invention is designed to use also more recently developed approaches using artificial intelligence and ma chine learning, e.g., Thuy T.T. Nguyen, G. A. (FOURTH QUARTER 2008), A Survey of Techniques for Internet Traffic Classifi cation using Machine Learning, IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 10, NO. 4, to capture the special nature of industrial communication data traffic, and to take the spe cial needs of the end points into account.

Network Controller module NC can configure network devices in the digital twin DT, as well as in the industrial communica tion network to embed data flows, etc. To enable that, it can detect the active topology, including device properties and capabilities. Configuration of network devices is typically done based on communication service requests coming from an end station (via User-Network-Interface, UNI), or via its northbound interface from a central management unit. Application A represents any type of application A that is present or should be deployed in an industrial communication network IN. Of particular interest for this invention are ap- plications with strict, even though inexplicit, requirements on quality of service for its communication relations, e.g., time-sensitive applications.

The Digital Twin DT represents the complete industrial sys- tern, including digital representations of the industrial com munication network, the devices, e.g., sensors, actuators, as well as digital representations of the applications A, pro cesses, and physical interactions. The Industrial communication Network IN is the real network that is used to run and operate the industry plant. Here the real applications are controlling the real industrial pro cesses and actuators based on the data from sensors, etc. FIG. 2 illustrates creation of an industrial communication network with its sensors, actuators, control processes and applications can be structured in four phases, with the main components' major functions: Planning and Design Phase

In this initial phase, application A that should be deployed to the industrial communication network IN is first deployed in the digital twin DT. After the application A is deployed and started in the digital twin DT, the network information provider and monitoring module NIP can start to collect data of the application's A communication. The data can already be used for initial classification of the application end point types and the corresponding data traffic. Whenever additional applications A or components are added to a plant using the industrial communication network IN, or existing ones are to be modified or deleted, planning and design phase might be re-entered. Detection and Analysis Phase

Now the Network Traffic Classification module NTC will iden- tify the type of the data flows and end points. Here differ ent approaches can be used, e.g., Fannia Pacheco, E. E.

(2018), Towards the Deployment of Machine Learning Solutions in Network Traffic Classification: A Systematic Survey. Com munications Surveys and Tutorials, IEEE Communications Socie- ty, 21 (2), S. pp. 1988-2014.

The classification results are passed to the software- implemented network controller module NC so that the network controller module NC can receive the relevant data traffic characteristics from the Network Information Provider and

Monitoring module NIP, in the same manner it receives commu nication service requests according to state of the art. As just assessing data from the application A running in an ide al environment is not enough to create communication pattern useful for the real world, the network controller module NC now starts changing the communication characteristics of the communication links used by the application in the Digital Twin DT. Changing the communication characteristics includes adding of additional delay on links, creating packet loss, re-aligning of send times of different streams in relation to each other, etc. by (virtually) adding additional communication services requiring resources from the network, or re-arranging network schedules. Thus, the application's A communication is "pushed to the limits" and the Digital Twin DT will notify the net work controller module NC as soon as the application A fails (e.g., a control loop is not able any more to control a phys ical process in the digital twin, results of processes are outside the required operational envelope).

An optional extension for that assessment can also be the use of real machinery, e.g., after adding a delay in the communi- cation of the real system, the surface quality after milling does not satisfy the quality requirements anymore. Then this real-world data can be used by the network controller module NC as well. By using this status feedback of the application, the network controller module NC gets a clear view of the ap plications' requirements for the communication channel(s) - without the application explicitly asking for them. The de termined parameter ranges can be used for further scheduling tasks of the network controller.

Configuration and Deployment Phase

After the detection and analysis phase, the network control ler module NC knows the upper bounds of delay, jitter, and packet loss the application can stand without degradation of process quality (= "its limits"). From that it can derive pa rameters for the communication service required by the appli cation A. These parameters will typically be significantly less demanding than the parameters observed by monitoring the "undisturbed" communication of the application in planning and design phase. The Network Controller module NC will then configure the physical industrial communication network IN, including end stations, based on the determined parameters. Afterwards, it can inform application A or an application de ployment framework (e.g., Kubernetes) about successful en forcement of their required communication relations, thus trigger productive operation.

Operations Phase

During operation, the real application A exchanges data through the physical communication network. From the Detec tion and Analysis phase, the network controller module NC knows the real QoS requirements for each communication ser vice of the application A. In a further embodiment, the Net work Information Provider and Monitoring module NIP continu ously monitor QoS parameters of relevant streams and alarm the network controller module NC in case of deviations from the defined minimum requirements.

Traditional approaches require a detailed specification of the applications' communication requirements, so that the network controller can configure and embed the corresponding data traffic flows accordingly. To specify the applications' communication requirements is a very complicated task as it involved knowledge from the applications' workflows, the in volved physical processes, and knowledge of the interfaces to other systems in the production environment. Such a knowledge is usually not available through one single person but must be consolidated by experts of different domains, such as the application developer, the subject matter expert for the pro duction process, the operator of the production, a system in tegrator.

The inventive approach offers automated support for describ ing the applications' A requirements and uses this infor mation to automatically derive the applications' A communica tion quality parameters required for the network controller to appropriately embed the data traffic flow.

The described method must handle applications that follow the "perceive - analyze - act" paradigm for process control, where the action based on a received sensor data reading is given in the same cycle - e.g., sensor data is received at the beginning of a cycle, then the reaction is calculated, and at the end of the cycle, the response command is sent to the process actuators. But of course, the described approach can also be used in parallelized processing of sensor read ings and control command - e.g., the cycle is here not the main criterion for sending and receiving, and control re sponses are calculated as soon as a sensor data reading is received (and this does not necessarily happen in the begin ning of the cycle). Although the invention has been explained in relation to its preferred embodiments as mentioned above, it is to be under stood that many other possible modifications and variations can be made without departing from the scope of the present invention. It is, therefore, contemplated that the appended claim or claims will cover such modifications and variations that fall within the true scope of the invention.

LIST OF REFERENCE SIGNS

A Application

CD Computational device D Device DT Digital Twin IN Industrial communication Network NC Network Controller NIP Network Information Provider and Monitoring module NTC Network Traffic Classification module