WO2020026256A1

WO2020026256A1 - Inspection entity and method performed therein for handling an output fault of a system

Info

Publication number: WO2020026256A1
Application number: PCT/IN2018/050504
Authority: WO
Inventors: Perepu SATHEESH KUMAR; Anusha Pradeep MUJUMDAR; Chakri PADALA
Original assignee: Telefonaktiebolaget Lm Ericsson (Publ)
Priority date: 2018-08-02
Filing date: 2018-08-02
Publication date: 2020-02-06

Abstract

An inspection entity (110) and a method performed therein for handling an output fault in a first output of a system (100) which is operable with multiple inputs and multiple outputs are provided. The method comprises: obtaining (S210) the multiple inputs and multiple outputs of the system (100); determining (S240), based on a dictionary, a time profile of an input fault associated with a time profile of the output fault, wherein the input fault is in a first input, the respective time profile of the input and output fault indicates how the respective input and output fault changes over time, wherein the dictionary comprises a mapping between the time profile of the input fault and the time profile of the output fault; and determining (S250), based on the determined time profile of the input fault, a time instant indicating when the input fault impacting the output fault occurred.

Description

INSPECTION ENTITY AND METHOD PERFORMED THEREIN FOR HANDLING

AN OUTPUT FAULT OF A SYSTEM

TECHNICAL FIELD

Embodiments herein relate to an inspection entity and method performed therein.

Furthermore, a computer program product and a computer readable storage medium are also provided herein. In particular, embodiments herein relate to handling an output fault in an output of a system which is operable with multiple inputs and multiple outputs.

BACKGROUND

Any system, in general may have multiple inputs and multiple outputs. For example, a robot is designed to perform various tasks such as walking, navigation etc., by taking several inputs from the environment or a user.

However a fault in an output may happen sometimes. For example, for a mobile robot, a command from the user may be an input to the mobile robot. An actuator system of the mobile robot may receive the input and pass it in a form of current to a servo motor attached to the wheels of the robot, in order to actuate movement. Hence, any fault in the current, i.e., the input, to the servo motor may impact the movement of the robot, e.g., the robot movement can display a zig-zag behavior. To correct such a zig-zag movement, the fault in the voltage input may need to be corrected, since the fault in the input has been propagated to the output of the robot.

Finding out the input which is responsible for the fault in the output is called root source of the fault identification. This process is straightforward for single input - single output system since there is only one input and one output. However, it is complex to find out the relation between fault of input and fault of output for multiple inputs-multiple outputs systems, i.e., a fault in one input may propagate to more than one output, and similarly a fault in output can be attributed to faults in two or more inputs. The propagation may also depend on dynamics of the system.

In brief, the fault in the output may be classified into two categories depending on the cause: (i) fault in the output occurred because of its own variable (because of a faulty sensor, for example)

(ii) fault in the output propagated from the input. In the case of fault propagation i.e. from input to output side, there are two possibilities: depending on the dynamics of the system, the fault may reach the destination, or may not reach the destination. Also, the time profile of the fault in the output may vary due to the dynamics of the system.

Here a challenge is to first classify the fault into one of these two categories: (i) fault occurred in its own variable or (ii) fault propagated from an input variable to one or more output variables. This classification of the fault is also known as fault attribution. In the latter case, the next challenge is to identify the root source of the fault i.e. to find which input is responsible for the fault in the output.

There is a plenty of literature in the field of fault attribution and root source identification. For instance, attributing the fault and identifying the root source may use sparse optimization techniques. First the fault in the output is identified and then a graph is constructed to identify the root source.

Furthermore, different faults in the same input may affect an output in various ways. For instance, a spike in the input may cause a step signal in the output, while a step signal in the input may cause a ramp signal in the output. Therefore it would be good to further identify the exact fault in the input, because different faults in one input, e.g., the spike and step signal in the input, may need different corrective actions.

However the prior art is only able to identify the input which has impacted the fault in the output, not be able to identify what an exact fault in the input that caused the fault in the output is.

SUMMARY

An object of embodiments herein is to provide a mechanism for improving the handling of an output fault in an output of a system.

According to an aspect the object is achieved by providing a method performed by an inspection entity for handling an output fault in a first output of a system. The system is operable with multiple inputs and multiple outputs. The inspection entity obtains the multiple inputs and multiple outputs of the system. The inspection entity determines, based on a dictionary, a time profile of an input fault associated with a time profile of the output fault. The input fault is in a first input. The respective time profile of the input and output fault indicates how the respective input and output fault changes over time. The dictionary specifies a mapping between the time profile of the input fault and the time profile of the output fault. The inspection entity determines, based on the determined time profile of the input fault, a time instant indicating when the input fault impacting the output fault occurred.

According to another aspect the object is achieved by providing inspection entity for handling an output fault in a first output of a system which is operable with multiple inputs and multiple outputs. The inspection entity is configured to obtain the multiple inputs and multiple outputs. The inspection entity is also configured to determine, based on a dictionary, a time profile of an input fault associated with a time profile of an output fault. The input fault is in a first input. The respective time profile of the input and output fault indicates how the respective input and output fault changes over time. The dictionary specifies a mapping between the time profile of the input fault and the time profile of the output fault. The inspection entity is also configured to determine, based on the determined time profile of the input fault, a time instant indicating when the input fault impacting the output fault occurred.

It is furthermore provided herein a computer program product comprising instructions, which, when executed on at least one processor, cause the at least one processor to carry out any of the methods above, as performed by the inspection entity. It is additionally provided herein a computer-readable storage medium, having stored thereon a computer program product comprising instructions which, when executed on at least one processor, cause the at least one processor to carry out the method above, as performed by the inspection entity.

According to another aspect the object is achieved by providing an inspection entity comprising processing circuitry configured to obtain multiple inputs and multiple outputs of a system which is operable with multiple inputs and multiple outputs. The processing circuity is also configured to determine, based on a dictionary, a time profile of an input fault associated with a time profile of an output fault. The input fault is in a first input. The respective time profile of the input and output fault indicates how the respective input and output fault changes over time. The dictionary specifies a mapping between the time profile of the input fault and the time profile of the output fault. The processing circuity is also configured to determine, based on the determined time profile of the input fault, a time instant indicating when the input fault impacting the output fault occurred.

According to still another aspect the object is achieved by providing a system operable with multiple inputs and multiple outputs. The system comprises the above inspection entity for handling an output fault in a first output of the system.

Embodiments herein provide an inspection entity which is able to identify an exact fault in the input that caused the fault in the output. Thanks to the time profiles of the input and output faults, the dynamic information of the system, e.g., the changes of the input and output faults with time is taken into account for handling of an output fault in an output of a system. By determining the time profile of the input fault which matches the time profile of the output fault, a precise fault in the input is identified, since different faults in one input may lead to different time profiles of one output fault. Embodiments herein enable a corrective action to the specific fault in the input. Embodiments herein are applicable to any dynamic system.

BRIEF DESCRIPTION OF THE DRAWINGS

Embodiments will now be described in more detail in relation to the enclosed drawings, in which:

Fig. 1 is a schematic overview depicting a new system according to embodiments herein;

Fig. 2 is a flowchart depicting methods performed by an inspection entity according to embodiments herein;

Fig. 3 is a flowchart depicting methods performed by an inspection entity according to embodiments herein;

Fig. 4a is a schematic depicting a network with two nodes according to embodiments herein;

Fig. 4b is a schematic depicting a 3D graph according to embodiments herein;

Fig. 5a is a schematic depicting a time profile of a fault in an input according to embodiments herein;

Fig. 5b is a schematic depicting a time profile of a fault in an output according to embodiments herein; Fig. 6 is a block diagram depicting an inspection entity according to embodiments herein.

DETAILED DESCRIPTION

As part of developing embodiments herein, a problem will first be identified and shortly discussed.

There is no prior art discloses construction of time profiles of the faults in the input and output. This is because some conventional techniques use only static information of the system to construct the graph, which does not include the transients or dynamics of the system i.e. the graph cannot reflect the dynamics information of the system, e.g., time profiles of the faults in the input side and output side.

Some other conventional techniques on graph construction focus on only the change of adjacency matrix with time, instead of the changes of the input or output with time. The dynamic information of the system, e.g., changes of the input or output with time, is therefore still not captured.

Embodiments herein relate to handling a fault in an output of a system which is operable with multiple inputs and multiple outputs. The embodiments herein assume that the fault has been attributed properly, and the root source of the fault has been identified, e.g. by using any conventional method.

The embodiments herein disclose determining a time profile of an input fault associated with a time profile of the output fault. Thanks to the time profile of the fault in the input, it is enabled to take the dynamic information of the system into account and take a necessary corrective action with respect to the specific fault in the input.

The embodiments herein also disclose determining a time instant indicating when the input fault impacting the output fault occurred. By knowing the associated time instant, a technical benefit is that the exact cause of the fault can be inferred since, both time and specific input has been identified.

Embodiments herein also disclose a library or dictionary, which comprises mappings between time profiles of the faults in the input and outputs. The dictionary brings a technical benefit of reduction in time to identify a fault in an input. Some embodiments herein may construct a plurality of adjacency matrices. Vertices of each adjacency matrix are the inputs and the outputs of the system. Edges of each adjacency matri are metrics, i.e., correlations, between each input and each output.

A system or device 100 which has multiple inputs

i_n and multiple outputs o₁- o_m is shown in Fig. 1. It will be appreciated that the system 100 is provided as an example of one embodiment and should not be construed to narrow the scope or spirit of the claimed solution in any way. In this regard, the scope of the claimed solution encompasses many potential embodiments in addition to those illustrated and described herein. As such, while FIG. 1 illustrates one example of a configuration of a system 100 for handling a fault in an output signal, numerous other configurations may also be used to implement embodiments of the present disclosure.

The system 100 refers to an electronic system or an electronic device. It should be understood by those skilled in the art that“system 100” is a non-limiting term which means any robot, machine, terminal, wired or wireless communication terminal, communication equipment, Machine Type Communication (MTC) device, Device to Device (D2D) terminal, or user equipment e.g. smart phone, laptop, mobile phone, sensor, relay, mobile tablets. The system 100 may be equipped with wired or wireless communication technology module, configured to communicate with an inspection entity 110, and/or to receive input from a user. The system 100 may comprise a camera and/or a sensor, such as a temperature sensor etc. The system 100 may also comprise e.g. one or more out of: a speedometer, a Global Positioning System (GPS) and an engine performance monitor. Examples of the system 100 may also include a vacuum robot, and a wearable device, e.g. a watch, a wristband, glasses, contact lenses, e-textiles and smart fabrics, a headband, a beany and cap, jewellery such as rings, bracelets, and hearing aid-like devices that are designed to look like earrings.

An inspection entity 110 which may be located internally in or externally to the system 100 is for handling an output fault in a first output of the system 100, particularly determining what fault in which input impacted the output fault. When the inspection entity 110 is located externally to the system 100, it may communicate with the system 100 with any wired or wireless communication technology.

The method actions performed by the inspection entity 110 for handling the output fault in a first output of the system 100 which is operable with multiple inputs and multiple outputs according to embodiments herein will now be described with reference to a flowchart depicted in Fig. 2. The actions do not have to be taken in the order stated below, but may be taken in any suitable order. Actions performed in some embodiments may be marked with dashed boxes.

Action S210. In embodiments herein, the inspection entity 110 obtains the multiple inputs and multiple outputs of the system 100.

Depending on different functions of the system 100, the multiple inputs and multiple outputs may vary. For instance, when the system 100 is a vacuum robot, the multiple inputs may comprise a voltage and current and also the commands given by the user etc., and the multiple outputs may comprise navigation of the vacuum robot, and performing some tasks like lifting the weights, cleaning the house etc.

Action S220. After obtaining the multiple inputs and multiple outputs of the system 100, the inspection entity 110 may identify the output fault in the first output.

The output fault in the first output may also be referred to as a fault in the first output or an output fault in short.

Action S230. Next, the inspection entity 110 may classify and attribute the output fault in the first output to a first input.

The above actions S220 and S230 may be performed according to any prior art method as mentioned above.

For now, the inspection entity 110 has the knowledge of the first input which has impacted the output fault, however is still not aware of what an exact fault in the first input is, and a time instant when the exact fault in the first input occurred. In order to obtain the above information, one or more of the following actions will be performed.

The term fault in the first input may also be referred to as an input fault in the first input or an input fault in short.

Action S240. The inspection entity 110 determines, based on a dictionary, a time profile of the input fault associated with a time profile of the output fault.

The respective time profile of the input and output fault indicates how the respective input and output fault changes over time.

The dictionary comprises a mapping between the time profile of the input fault and the time profile of the output fault. Constructing of the dictionary will be discussed later. By knowing the time profile of the output fault, it is possible to arrive at the time profile of the input fault.

Thanks to the time profile of the fault in the input, it is enabled to determine which exact fault lies in this input, and take a necessary corrective action later on.

Action S250. After knowing the time profile of the input fault in the first input, the inspection entity 110 may determine, based on the determined time profile of the input fault, a time instant indicating when the input fault impacting the output fault occurred.

The time instant indicating when the input fault impacting the output fault occurred may also be called the associated time instant in short, or the position of the input fault, The associated time instant or the position of the input fault may comprise one or more time instants. When the position of the input fault comprises more than one time instant, it may comprise a starting time instant and an ending time instant.

The inspection entity 110 may determine the position of the input fault by determining time instant(s) of the input at which the input is most correlated with the time profile of the input fault in the dictionary.

In other words, the inspection entity 110 may compare a whole time profile of the input fault with the time profile of the input fault in the dictionary, the latter normally has a shorter time length, e.g., 5 seconds, than the former. In the whole time profile, a part that is most correlated with the time profile of the input fault in the dictionary will be determined as the position of the input fault. Accordingly, the starting time instant and the ending time instant of the input fault respectively correspond to the beginning and the ending of the most correlated part of the whole time profile of the input fault.

Determining the associated time instant brings technical benefit of knowing an exact cause of the output fault can be inferred since both the time instance and specific input fault have been identified.

Action S260. To facilitate a correction of the first input, the inspection entity 110 may send the time profile of the fault in the first input and the associated time instant, to the system 100 or other device, so that proper corrective action to the specific fault in the input can be performed.

In order to handle the output fault in the first output of the system 100, the inspection entity 110 may construct the dictionary. The method actions performed by the inspection entity 110 for constructing of the dictionary according to some embodiments herein will now be described with reference to a flowchart depicted in Fig. 3. The actions do not have to be taken in the order stated below, but may be taken in any suitable order. Actions performed in some embodiments may be marked with dashed boxes.

Action S310. The inspection entity 110 may obtain multiple inputs and multiple outputs of the system 100 under a test. There are at least one fault in the multiple inputs under the test and at least one fault in the multiple outputs under the test.

The actions in Fig. 3 may normally be performed during a test phase, and prior to Fig. 2. The multiple inputs and multiple outputs of the system 100 obtained here are those under the test, which may be different from those in the actions S201.

Action S320. The inspection entity 110 may construct the plurality of adjacency matrices at different time instants based on the multiple inputs and multiple outputs under the test.

The constructing of the plurality of adjacency matrices at different time instants is also referred to as constructing a 3D graph, which is an efficient way of indicating the dynamics of the system 100, see Fig. 4b.

Each adjacency matrix may be constructed by determining all elements of the matrix at that time instant. Each element may indicate a matric, e.g., either conditioned correlation or conditioned mutual information, between a pair of vertices. Due to all the inputs and outputs under the test are already known, the element may therefore be determined by computing the matric between the pair of vertices. The construction of the 3D graph may require a lot of computations. However, the computation is only a one time task unless operating conditions change. Hence, the 3D graph may only be constructed when the operating conditions change. For example, a non-linear system may be linearized only around operating conditions. Hence, for the case of non-linear system, the 3D graph may be constructed only in a specific time interval which can be system specific so that the construction of the 3D graph is linear over time.

Action S330. The inspection entity 110 may construct the dictionary based on a plurality of adjacency matrices, wherein each adjacency matri specifies metrics between the multiple inputs and multiple outputs at a specific time instant.

The dictionary may comprise a plurality of mappings. Each mapping specifies a correspondence between a time profile of one input fault and a time profile of one output fault.

Constructing the dictionary herein may be referred to determining the plurality of mappings.

For each mapping, the inspection entity 110 may derive an algebraic expression defining the mapping between a time profile of each input fault and a time profile of each output fault. The algebraic expression may be derived based on the above determined metrics in each adjacency matrix. Due to each adjacency matrix is associated to a time instant, the algebraic expression is therefore able to reflect the changes of each input and output fault with time.

The embodiments herein enable to construct the 3D graph only once to obtain the mappings in the dictionary. The embodiments herein are computationally less complex when compared with the conventional method, as the construction 3D graph is performed only once to obtain the mappings in the dictionary. Hence, necessary corrective actions can be done based on the dictionary without constructing the 3D graph again. An easy solution is therefore achieved herein. The dictionary enables about 50% reduction in time to identify the fault in an input, due to the mapping helps to reduce the time spent in identifying the fault in an input, type of fault in the input etc.

In the embodiments, some embodiments are described in context of single-integrator system, however the skilled person will appreciate that the embodiments herein are also applied to any type of system. Before explaining a 3D graph according to the embodiments herein, it is illustrated a generic concept of the 3D graph. If we say, an adjacency matrix is a 2D graph depicting correlations between vertices. A 3D graph is able to reflect dynamic information of the system 100. The 3D graph refers to a plurality of the adjacency matrices constructed at different time instants.

Normally, a network may be represented as a graph in a form of an adjacency matrix. For the ease of understanding, nodes of network will be used as an example of vertices of the 3D graph herein. The adjacency matrix contains information of nodes which are connected to other nodes. For example, let us assume a network with a two nodes A-B network as shown in Fig. 4a.

An adjacency matrix AJ is

For ease of writing, we are skipping the nodes’ information written above the adjacency matrix.

The nodes A-B of the network form the vertices of the directed adjacency matrix AJ. Edges of the directed adjacency matrix indicate connections between nodes. In the directed adjacency matrix AJ, the rows correspond to sources and the columns correspond to destinations. The adjacency matrix can be understood as follows. For example, if the first element is‘G, this means the node A is directly affecting the node Έ’. If it is zero it means the nodes are not connected. It should be noted that the value will be 1 even when there exists a connection at any time i.e. irrespective of dynamics of the system.

A problem with this representation is that dynamic information of the network is not reflected. The dynamic information may comprise a delay and/or a gain factor, etc. For example, there may be two dynamic aspects of the network:

(i) A delay of 2s for information transfer from the node A to the node B;

(ii) A gain factor of 0.5 for information transfer from the node A to the node B.

It is noted that though some embodiments are described in context of 2 seconds delay and 0.5 gain factor, however the skilled person will appreciate that the embodiments herein are applicable to any delay and any gain factor. The above adjacency matrix does not present any information about the dynamics of the system. Hence, we propose here to construct a plurality of matrices at different time instants, in order to present the dynamic information of the system along time.

For instance, at a time instant t = 0, the adjacency matrix A_t is

The adjacency matrix A₀ may be constructed by computing the directed correlation between the value of the vertices A and B at the time instant t = 0. In this case the directed correlation between the vertices A and B is 0 since the vertex B contains total different data when compared with the vertex A due to the delay is 2 seconds. No information is received at the node B, when information is sent at t = 0 from the node A. Thus the element from the node A to the node B is 0.

In the case of A₁, the adjacency matrix may be constructed by using the conditional correlation between the value of the vertices A and B at one lag difference. At a time instant t = 1, which indicates, e.g., one second after sending the information, the adjacency atri A₁ will be the same as A₀, since a delay of the transferring the information from the node A to the node B is 2 seconds.

An adjacency matrix at lag‘M’ may also be referred to as an adjacency matrix at a time instant t=M. The adjacency matrix at lag‘M’ may be calculated by using the data of the vertices at the M time instant

At a time instant t = 2, the adjacency matrix A₂ will be

At this time instant t = 2, i.e., two seconds after sending of the information, the information has been received by the node B, the element from the node A to the node B should be 1. However, due to the edge weight is 0.5, therefore the metric from the node A to the node B is l*0.5=0.5. For the adjacency matrix of higher orders, i.e., higher time instant, i.e., t > 3, all elements of the adjacency matrix will be zero. This is the case with the system having pure time delay. Adjacency matrices of different time instants would be constructed until the adjacency matrix does not change or all the elements become zero. Let the final order be P. In this case, the value of P is 2. The plurality of adjacency matrices would be stacked as shown in Fig. 4b. The plurality of adjacency matrices at different time instants form a 3-dimensional array, therefore it is referred to as 3D graph. For systems with transients, the adjacency matrix will be complex. Here, we may propose to“stack” all the matrices in 3-dimensional format as

A = [A₀ A, A₂] (300)

From the 3-dimensional matrix A it is possible to know how a signal is getting distorted passing on from one node to another node. All the adjacency matrices at different time instants may be constructed from the data of all the nodes, i.e., the entire approach is data-driven.

Henceforth, we call the 3-dimensional array a 3D graph. In practice, the adjacency matrix is constructed by filling the elements of matrix with either conditioned correlation or conditioned mutual information. The usage of these metrics gives the relation between their variables exactly conditioned on other variables.

The 3D graph according to the embodiments herein may present the following dynamic information of a system on per input and output basis:

(i) How much time an input will take to affect an output.

(ii) How much gain factor indicating distortion an input will undergo in an output.

The method actions performed by an inspection entity 110 for handling an output fault in a first output of a system 100 according to more detailed embodiments herein will now be described.

According to the embodiments herein the multiple inputs and multiple output of the system 100 are the vertices of the 3D graph. For the reason of simplicity, the following embodiments will be described in context of two inputs i_lt i₂ and two outputs o₁, o₂ , however the skilled person will appreciate that the embodiments herein are also applied to any number of input and outputs.

The system 100 is operable with two inputs i_lt i₂ and two outputs o_lt o₂. The system 100 is a discrete one, and a k^lh time instant refers to t a time instant at K*T, where T is a time interval, e.g., 1 second.

At a 0^th time instant t = 0 The inspection entity 110 may obtain multiple inputs and multiple outputs of the system 100 under a test at this time instant.

For each input and output, the inspection entity 110 may compute a metric (either conditional correlation or conditional mutual information) between the input and the output.

The computed metrics will be used as elements to construct an adjacency matrix

A₀ at the time instant t = 0.

As an example the adjacency matrix A₀ at the time instant t = 0 may be constructed as below:

The inputs do not impact each other, therefore the elements indicating metrics between inputs are 0.

The element, e.g., indicating the metric between and o₂ is 0, that is because the input does not impact the output o₂ at the time instant t = 0.

In this example adjacency matri A₀, the element in the first row has a metric 0.869 between the input and the output o_c. It indicates that at the time instant t = 0, there is a direct impact on the output o₁ from the input i_{1 ,} and a magnification factor there is 0.869. Similarly, the element in the third row is the metric 0.869 between the input and the output

Oi .

At a I^st time instant t = 1

The inspection entity 110 will perform the same actions as those at the 0^th time instant t = 0.

As an example an adjacency matrix A^t the time instant t = 1 may be constructed as below:

As you can see all the elements in the matrix are zeros since there is no dependency on this lag. In other words, all outputs are not impacted by any input at this time instant. At a 2^nd time instant t = 2

The inspection entity 110 will also perform the same actions as those at the 0^th h time instant t = 0.

As an example an adjacency matrix A₂at the time instant t = 2 may be constructed as below:

(603)

In this example adjacency matrix A₂, the element in the second row has a metric 0.5 between the input i_x and the output o₁. It indicates that at the time instant t = 2, there is an impact on the output o₁ from the input i₂, and a magnification factor is 0.5; also there is an impact on the output o₂ from the input i₂, and a magnification factor is 0.64.

The adjacency matrices A₀ - A₂ are symmetric matrices, however they are not necessary symmetric, depending on a design of the system.

At a M^th time instant M > 2

For all the other lags, the adjacency matrices will be same as the system input is affected from two lags of the data.

After having the above adjacency matrices at different time instants, the 3D graph may be constructed by stacking the constructed adjacency matrices as shown in Fig. 4b.

A = [A₀ A₁ A₂] (700) Back to the computed metrics at different time instants, taking the input i₁ and the output o₁ as an example, time profiles of faults in the input i_x and the output o₁ will be obtained based on the computed metrics at different time instants. Fig. 5a illustrates the time profile of a fault in the input i₁ under test following sinusoidal. Fig. 5b illustrates the corresponding time profile of a fault in the output o₁ impacted by the fault in the input i₁ Based on the computed metrics at the k^th time instant, as an example, the algebraic expressions specifying the relationship between the time profile of the fault in the input and the time profile of the fault in the output may be derived as below:

o /c] = 0.5 ki^k] + 0.25 i₂ [k - 2] (400) o₂ [k] = 0.75/e²t₂[/e - 4] (500)

Wherein,

2: a delay, e.g., in second, transferring information from an input to an output. In other words, how long time it will take that the input i₁ impacts the output o_x;

0.5: the magnification factor/weight indicating the input i_x amplifies by 0.5 times in construction of the output o₁ ;

0.25: the magnification factor/weight indicating the input i₂ amplifies by 0.25 times in construction of the output o₁ ;

0.75: the magnification factor/weight indicating the input i₂ amplifies by 0.75 times in construction of the output o₂ ;

k²: an amplification factor, which varies along with a value of the time instant k, so it may also be called a dynamic amplification factor in terms of time instance, e.g., seconds. The dynamic amplification factor indicates that the input i₂ amplifies by k² times in construction of the output o₂ ;

4: a delay, e.g., in second, how long time it will take that the input i₂ impacts the output o₂.

From the equations 400 and 500, it is understood that any fault in input i₁ may affect the output o₁ and fault in input t₂ may propagate to the outputs o₁ and o₂. The input may be correlated with each other.

Next, the inspection entity 110 may construct the dictionary based on the plurality of adjacency matrices, particularly, based on the computed metrics.

An example of the dictionary may be as Table 1 below.

Table 1

It is noted that in Table 1 time profiles are described in text as examples, alternative, they may also be presented as figures, like Fig. 5a-Fig 5b.

Taking the first row of the dictionary as an example, it indicates a mapping of time profiles between an output o₁ and an input i₁. In other words, how the input (fault) i_x impacts the output (fault) o₁ along time. For instance, at the output side, the time profile of the output o₁ is in a sinusoidal shape, with a time length, i.e., a time window, 10 seconds. A frequency of the output (signal) o₁ is 0.1 Flz with amplitude 1 unit. Meanwhile at the input side, the time profile of the Input i_x is also in a sinusoidal shape, with a time length, i.e., a time window, 5 seconds. A frequency of the input (signal) i_x is 0.1 Flz with amplitude 0.5 unit. As an example here the time window of the time profile of the input i₁ is different from the output o₁, alternatively they may be the same as shown in Fig. 5a-Fig 5b. Choice of the time windows is a design option, it may vary depending on the dynamics of the system 100. The magnification factor may be dawn as follows: magnification factor = output amplitude/input amplitude. Thus the magnification factor in this example is l/0.5=2.

Similar to the first row, an example of a mapping of time profiles between the output o₂ and the input i_x is also given.

After having the dictionary, the inspection entity 110 may obtain multiple inputs and multiple outputs of the system 100 and try to identify the time profile of the fault in the input i₁ given the fault in the output o₁, and the associated time instant as follows.

First the fault in the output o₁ may be identified and attributed to the input i_x by using any prior art method. Next, the inspection entity 110 may determine, based on a dictionary, the time profile of the fault in the input ^ which is corresponding to the time profile of the fault in the output o₁. The inspection entity 110 may then determine a time instant when the input fault impacting the output fault occurred.

Consequently, necessary corrective action may be taken (not part of the proposed solution).

Use case Robots Case Study: In any robot, there may be many inputs and outputs. These outputs facilitate the functions of the robot such as navigation, and performing designed tasks etc. The inputs these robots take may be actuator commands such as voltage supplied to a motor etc. Any fault in the voltage to the motor may result in a fault at the output of the robot. Such faults may result in the robot displaying undesirable behaviour.

For example, a fault in the voltage given to the motor of one leg may result in improper navigation of the robot. The same can occur if there is a fault in the camera of the robot. In this case, improper navigation is an example of a fault in output, the faults in the voltage and in the camera are examples of faults in input. Hence, identifying the root source of the fault occurring during navigation requires the identification of the fault in the input from which the fault is propagated. In addition, due to system dynamics the faults in the different inputs may impact the output differently. Hence, by virtue of understanding the time profile of the fault in the output, it is possible to more correctly identify the input from which the fault is propagated and also the time profile of the fault in the input.

Fig. 6 is a block diagram depicting the inspection entity 110 according to embodiments herein for handling an output fault in a first output of a system 100 which is operable with multiple inputs and multiple outputs.

The inspection entity 110 may comprise processing circuitry 601, e.g. one or more processors, configured to perform the methods herein.

The inspection entity 110 may comprise an obtaining module 610, e.g. a receiver or transceiver. The inspection entity 110, the processing circuitry 601, and/or the obtaining module 610 is configured to obtain the multiple inputs and multiple outputs of the system 100.

The inspection entity 110 may comprise a fault detection and attribution module 611.

The inspection entity 110, the processing circuitry 601, and/or the fault detection and attribution module 611 may be configured to identify the output fault; and to attribute the output fault to the first input.

The inspection entity 110 may comprise a first determining module 612. The inspection entity 110, the processing circuitry 601, and/or the first determining module 612 is configured to determine, based on a dictionary, a time profile of an input fault associated with a time profile of an output fault. The input fault is in a first input. The respective time profile of the input and output fault indicates how the respective input and output fault changes over time. The dictionary comprises a mapping between the time profile of the input fault and the time profile of the output fault.

The inspection entity 110 may comprise a second determining module 613. The inspection entity 110, the processing circuitry 601, and/or the second determining module 613 is configured to determine, based on the determined time profile of the input fault, a time instant indicating when the input fault impacting the output fault occurred.

The inspection entity 110, the processing circuitry 601, and/or the obtaining module 610 may be configured to further obtain multiple inputs and multiple outputs of the system 100 under a test.

The inspection entity 110 may comprise a first constructing module 614. The inspection entity 110, the processing circuitry 601, and/or the first constructing module 614 may be configured to construct the plurality of adjacency matrices at different time instants based on the multiple inputs and multiple outputs under the test. At least one fault is in the multiple inputs under the test and at least one fault is in the multiple outputs under the test.

The inspection entity 110 may comprise a second constructing module 615. The inspection entity 110, the processing circuitry 601, and/or the second constructing module 615 may be configured to construct the dictionary based on a plurality of adjacency matrices, wherein each adjacency matrix specifies metrics between the multiple inputs and multiple outputs at a specific time instant. In an embodiment, the inspection entity 110 may comprise a second constructing module 615. The inspection entity 110, the processing circuitry 601, and/or the second constructing module 615 may be configured to determine the metrics in each adjacency matrix, and to derive, based on the metrics in each adjacency matrix, an algebraic expression defining the mapping between a time profile of each input fault and a time profile of each output fault.

The inspection entity 110 may further comprise a memory 604. The memory comprises one or more units to be used to store data on, such as the multiple inputs and multiple outputs, the adjacency matrices, the dictionary, the algebraic expression and/or the time instant indicating when the input fault impacting the output fault occurred. Thus, the inspection entity 110 may comprise the processing circuitry and the memory, said memory comprising instructions executable by said processing circuitry whereby said inspection entity 110 is operative to perform the methods herein.

The methods according to the embodiments described herein for the inspection entity 110 are respectively implemented by means of e.g. a computer program or a computer program product 605, comprising instructions, i.e., software code portions, which, when executed on at least one processor, cause the at least one processor to carry out the actions described herein, as performed by the inspection entity 110. The computer program product 605 may be stored on a computer-readable storage medium 606, e.g. a disc, universal serial bus (USB) stick or similar. The computer-readable storage medium 606, having stored thereon the computer program product 605, may comprise the instructions which, when executed on at least one processor, cause the at least one processor to carry out the actions described herein, as performed by the inspection entity 110. In some embodiments, the computer-readable storage medium may be a non-transitory computer-readable storage medium.

As will be readily understood by those familiar with communications design, that functions means or modules may be implemented using digital logic and/or one or more microcontrollers, microprocessors, or other digital hardware. In some embodiments, several or all of the various functions may be implemented together, such as in a single application-specific integrated circuit (ASIC), or in two or more separate devices with appropriate hardware and/or software interfaces between them. Several of the functions may be implemented on a processor shared with other functional components of a radio network node, for example.

Alternatively, several of the functional elements of the processing means discussed may be provided through the use of dedicated hardware, while others are provided with hardware for executing software, in association with the appropriate software or firmware. Thus, the term“processor” or“controller” as used herein does not exclusively refer to hardware capable of executing software and may implicitly include, without limitation, digital signal processor (DSP) hardware, read-only memory (ROM) for storing software, random-access memory for storing software and/or program or application data, and non volatile memory. Other hardware, conventional and/or custom, may also be included. Designers of radio network nodes will appreciate the cost, performance, and maintenance trade-offs inherent in these design choices.

Claims

CLAIMS:

1. A method performed by an inspection entity (110) for handling an output fault in a first output of a system (100) which is operable with multiple inputs and multiple outputs, the method comprising:

- obtaining (S210) the multiple inputs and multiple outputs of the system (100);

- determining (S240), based on a dictionary, a time profile of an input fault associated with a time profile of the output fault, wherein the input fault is in a first input, the respective time profile of the input and output fault indicates how the respective input and output fault changes over time, wherein the dictionary comprises a mapping between the time profile of the input fault and the time profile of the output fault; and

- determining (S250), based on the determined time profile of the input fault, a time instant indicating when the input fault impacting the output fault occurred.

2. The method according to claim 1, the method further comprising:

- constructing (S330) the dictionary based on a plurality of adjacency matrices, wherein each adjacency matrix specifies metrics between the multiple inputs and multiple outputs under a test at a specific time instant, at least one fault is in the multiple inputs under the test and at least one fault is in the multiple outputs under the test.

3. The method according to claim 2, the constructing (S330) the dictionary based on a plurality of adjacency matrices comprising:

- determining the metrics in each adjacency matrix; and

- deriving, based on the metrics in each adjacency matrix, an algebraic expression defining the mapping between a time profile of each input fault and a time profile of each output fault.

4. The method according to any one of claims 1-3, the method further comprising:

- obtaining (S310) multiple inputs and multiple outputs of the system (100) under a test; and

- constructing (S320) the plurality of adjacency matrices at different time instants based on the multiple inputs and multiple outputs under the test.

5. The method according to any one of claims 1-4, the method further comprising:

- identifying (S220) the output fault; and - attributing (S230) the output fault to the first input.

6. The method according to any one of claims 2-5, wherein each metric is associated with a measure of either correlation or mutual information between one input and one output under the test at that specific time instant.

7. An inspection entity (110) for handling an output fault in a first output of a system (100) which is operable with multiple inputs and multiple outputs, the inspection entity (110) is configured to:

- obtain the multiple inputs and multiple outputs of the system (100);

- determine, based on a dictionary, a time profile of an input fault associated with a time profile of an output fault, wherein the input fault is in a first input, the respective time profile of the input and output fault indicates how the respective input and output fault changes over time, wherein the dictionary comprises a mapping between the time profile of the input fault and the time profile of the output fault; and

- determine, based on the determined time profile of the input fault, a time instant indicating when the input fault impacting the output fault occurred.

8. The inspection entity (110) according to claim 7, the inspection entity (110) further configured to:

- construct the dictionary based on a plurality of adjacency matrices, wherein each adjacency matrix specifies metrics between the multiple inputs and multiple outputs under a test at a specific time instant, at least one fault is in the multiple inputs under the test and at least one fault is in the multiple outputs under the test.

9. The inspection entity (110) according to claim 8, the inspection entity (110) further configured to:

- determine the metrics in each adjacency matrix; and

- derive, based on the metrics in each adjacency matrix, an algebraic expression defining the mapping between a time profile of each input fault and a time profile of each output fault.

10. The inspection entity (110) according to any one of claims 8-9, wherein each metric is associated with a measure of either correlation or mutual information between one input and one output under the test at that specific time instant.

11. The inspection entity (110) according to any one of claim 7-10, the inspection entity (110) further configured to:

- obtain multiple inputs and multiple outputs of the system (100) under a test; and

- construct the plurality of adjacency matrices at different time instants based on the multiple inputs and multiple outputs under the test.

12. The inspection entity (110) according to any one of claims 7-11, the inspection entity (110) further configured to:

- identify the output fault; and

- attribute the output fault to the first input.

13. An inspection entity (110) comprising processing circuitry configured to:

- obtain multiple inputs and multiple outputs of a system (100) which is operable with multiple inputs and multiple outputs;

- determine, based on a dictionary, a time profile of an input fault associated with a time profile of an output fault, wherein the input fault is in a first input of the system (100), the output fault in a first output of the system (100), the respective time profile of the input and output fault indicates how the respective input and output fault changes over time, wherein the dictionary comprises a mapping between the time profile of the input fault and the time profile of the output fault; and

14. A system (100) operable with multiple inputs and multiple outputs, comprising an inspection entity (110) according to any one of claims 7-13 for handling an output fault in a first output of the system (100).

15. A computer program product comprising instructions, which, when executed on at least one processor, cause the at least one processor to carry the method according to any of the claims 1-6, as performed by the inspection entity (110).

16. A computer-readable storage medium, having stored thereon a computer program product comprising instructions which, when executed on at least one processor, cause the at least one processor to carry out the method according to any of the claims 1-6, as performed by the inspection entity (110).