Driving assistance device, driving assistance method, and program

Abstract

To provide a technique that obtains an optimum operation for returning from an abnormal state to a normal state in a short time.SOLUTION: A driving device according to one aspect of the present disclosure includes: an abnormality diagnosis unit that diagnoses whether or not an abnormality or a symptom of the abnormality has occurred in a control object by using operating data of the control object and an abnormality diagnosis model; a formulation unit that formulates an optimization problem for calculating an optimal operation for returning the control object to normal when it is diagnosed that an abnormality or a symptom of the abnormality has occurred in the control object; an operation amount calculation unit that calculates optimal operation, and an output unit that outputs the optimal operation to a predetermined output destination. The formulation unit formulates an optimization problem that minimizes a magnitude of the operation for returning the control object to a normal state and the degree of deviation from a distribution of normal data representing past operating data of the control object when the control object is normal, and that includes a first constraint condition indicating that a distance between the normal data and the operating data after the operation is greater than or equal to a distance between the normal data closest to the operating data after the operation and the operating data after the operation.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present disclosure relates to a driving assistance device, a driving assistance method, and a program. [Background technology]

[0002] A technology has been known for some time that calculates optimal operations to return plants to normal from an abnormality when an abnormality or its symptoms are diagnosed or detected in plants such as waste incineration plants, taking into account past operating data during normal times (Patent Document 1). This conventional technology uses a method called CE (DACE: Distribution-Aware Counterfactual Explanation), which takes data distribution into account (Non-Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 7103539 [Non-patent literature]

[0004] [Non-Patent Document 1] Kentaro Kanamori, Takuya Takagi, Ken Kobayashi, Hiroki Arimura, A counterfactual explanation method considering feasibility based on mixed integer linear programming, The 34th National Conference of the Japanese Society for Artificial Intelligence (2020) Summary of the Invention [Problem to be solved by the invention]

[0005] However, in Non-Patent Document 1, it may take a long time to formulate an optimization problem for finding an optimal operation, and as a result, it may take a long time to obtain the optimal operation.

[0006] The present disclosure has been made in consideration of the above points, and provides a technique that can obtain an optimal operation for returning from an abnormality to a normal state in a short period of time. [Means for solving the problem]

[0007] A driving device according to one aspect of the present disclosure includes an abnormality diagnosis unit that diagnoses whether an abnormality or a sign of the abnormality has occurred in the controlled object using driving data of the controlled object and an abnormality diagnosis model; a formulation unit that formulates an optimization problem for calculating an optimal operation for returning the controlled object to normal when it is diagnosed that an abnormality or a sign of the abnormality has occurred in the controlled object; an operation amount calculation unit that calculates the optimal operation by solving the optimization problem; and an output unit that outputs the optimal operation to a predetermined output destination, wherein the formulation unit formulates an optimization problem that minimizes the magnitude of the operation for returning the controlled object to normal and the degree of deviation from a distribution of normal data that represents past driving data of the controlled object when it is normal, and includes a first constraint condition that indicates that the distance between the normal data and the driving data after the operation is greater than or equal to the distance between the normal data closest to the driving data after the operation and the driving data after the operation. Effect of the Invention

[0008] To provide a technique for obtaining an optimal operation for returning from an abnormality to a normal state in a short time. [Brief description of the drawings]

[0009] [Figure 1] FIG. 10 is a diagram illustrating an example of calculation of an optimal operation. [Diagram 2] 1 is a diagram illustrating an example of an overall configuration of a plant control system according to an embodiment of the present invention. [Diagram 3] FIG. 2 is a diagram illustrating an example of a hardware configuration of the driving assistance device according to the present embodiment. [Figure 4] FIG. 2 is a diagram illustrating an example of a functional configuration of a driving assistance device according to the present embodiment. [Diagram 5] 5 is a flowchart illustrating an example of a driving assistance process according to the present embodiment. [Figure 6] FIG. 13 is a diagram illustrating a comparison of calculation times between the conventional method and the proposed method. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] An embodiment of the present invention will be described below. The following describes a plant control system 1 that can present to an operator in a short time the optimal operation for returning a plant to normal when an abnormality or a symptom thereof is diagnosed or detected in the plant, with the plant being controlled by the operation of an operator. In the following, abnormalities or symptoms thereof will be collectively referred to as "abnormalities, etc.", and diagnosing or detecting the presence or absence of an abnormality, etc. will be collectively referred to as "diagnosing the presence or absence of an abnormality, etc." or simply "diagnosing an abnormality, etc."

[0011] <Question setting> Let x be the plant operation data. Since plant operation data is generally multivariate data, the total number of variables (also called state variables, etc.) that represent the state of the process is D, so x = (x 1 ,···,x D ) is expressed as the operating data to be diagnosed for anomalies (hereinafter, also referred to as "diagnosis target data") as x (p) The process state may be, for example, temperature, pressure, flow rate, gas concentration, etc., and these are measured by various sensors installed in the plant. The state variable x 1 ,···,x D Among these variables, there may be not only variables that operators can manipulate, but also variables that cannot be manipulated (for example, in the case of a waste incineration plant, the amount of waste input and the air flow rate can be manipulated, but the CO concentration, etc. cannot be manipulated).

[0012] The anomaly diagnosis model is represented as H(·), and H(x (p) When H(x)=+1, the plant is normal (i.e., no abnormality occurs), and H(x (p))=-1, the plant is diagnosed as abnormal (i.e., an abnormality or the like is occurring). Here, the abnormality diagnosis model is a model for diagnosing the presence or absence of an abnormality or the like in the plant, and is created in advance, for example, by a machine learning method or the like using past operation data. Note that the abnormality diagnosis model H may be a linear model (i.e., a model that can linearly separate the region where H(x)=+1 and the region where H(x)=-1 are satisfied) or may be a nonlinear model.

[0013] In this case, H(x (p) When it is diagnosed that H(x )=-1, the distribution of the past operation data of the plant under normal conditions (hereinafter also referred to as "past normal data" or simply "normal data") is taken into consideration. (p) The objective is to calculate the optimal operation amount a that satisfies a = (a +a) = +1. 1 ,···,a D ) below, each a d (d=1, ,D) will also be called the instrumental variable.

[0014] <Formulation of conventional methods> In the conventional method described in Non-Patent Document 1 and Reference 1, each operation variable a is set to a discrete value. d For (d=1, ,D), the instrumental variable a d (i.e., the state variable x d A set of operation variables a d Let I be the total amount of change that can be set to d , the i-th change amount is Δa d,i In addition, each change amount Δa d,i For example, {(Δa 1,i , ,Δa D,i )|i=1,···,I d} is a set of lattice points at a given interval in the D-dimensional space. However, for example, if a certain operational variable a d If the variable a cannot be manipulated, d The change ind For Δa d,i =0.

[0015] Note that the instrumental variable a d If there is a continuous value, it is sufficient to discretize it and then formulate it. However, the instrumental variable a d It is not necessary that discrete values ​​are set for each manipulated variable a. For example, d A continuous value may be set for all or part of the above.

[0016] In this case, the objective function in the conventional method is d(a)+λq k (x (p) +a|X), and the constraints are expressed as the following equations (1) to (4). The optimization problem is formulated as the objective function d(a)+λq k (x (p) +a|X) is the optimal operation amount a. a d(a)+λq k (x (p) +a|X) st The optimal operation amount a is calculated by solving the optimization problems represented by equations (1) to (4).

[0017]

number

[0018] Outlier degree q k We can use the local outlier factor (LOF) as the density of neighboring points. k (x|X) can be calculated as follows:

[0019]

number

[0020]

number

[0021]

number

[0022] Also, π d,i is the instrumental variable a d The i-th change amount Δa d,i A binary variable, v, which indicates whether to set n The nth normal data is the operation data after operation Q:=x (p) +a is a binary variable that indicates whether it is the nearest point of a or not. d,i n is the instrumental variable a d The i-th change amount Δa d,i Operation data after operation when setting Q=x (p) +a's state variable x d and the state variable x of the nth normal data d It also represents the distance between the value of C n represents a sufficiently large value. However, π d,i is π d,i When =1, the instrumental variable a d The i-th change amount Δa d,i represents the setting of d,i When =0, the instrumental variable a d The i-th change amount Δad,i Similarly, v n v n When v = 1, it means that the nth normal data is the nearest point of the driving data Q after the operation, and v n When n=0, it indicates that the nth normal data is not the nearest point of the driving data Q after the operation.

[0023] As an example, D=2, R={x=(x 1 ,x 2 )|H(x)=+1}, H(·) is a linear model, and the set of driving data E 1 ⊂X is dense, and the set of driving data E 2 An example of calculating the optimal operation amount a when the density is low in ⊂X is shown in Figure 1. In the example shown in Figure 1, the distance of the operation amount (in other words, the magnitude of the operation amount) is as small as possible, taking into account the density of past normal data, and the operation data after operation Q = x (p) +a's neighborhood is as dense as possible (i.e., the outlier degree q k (x (p) The optimal manipulated variable a is calculated so that +a|X) is as small as possible.

[0024] However, in conventional methods, it can take a long time to formulate the optimization problem. Specifically, because the time required to formulate the constraint conditions shown in the above formula (2) is proportional to the square of N, when N (i.e., the number of past normal data) is large, it takes a long time to formulate the optimization problem. For this reason, when N is large, in conventional methods, after an abnormality is diagnosed by the abnormality diagnosis model H(·), it takes a long time for the optimal operation amount a to be presented to the operator, making it impossible to appropriately avoid or recover from the abnormality.

[0025] Therefore, in the following, an optimization problem is formulated using equivalent constraint conditions instead of the constraint conditions shown in the above equations (2) to (3).

[0026] <Formulation of the proposed method> In the proposed method, equations (2) to (3) are modified to constraint conditions that can be formulated in a shorter time.

[0027] Specifically, first, the driving data after the operation Q = x (p) + Variable d representing the distance between a and its nearest normal data point min is introduced and defined as the following equation (5).

[0028]

number

[0029] In addition, the instrumental variable a d The i-th change amount Δa d,i Operation data after operation when setting Q=x (p) +a's state variable x d and the state variable x of the nth normal data d The variable d represents the distance between n is introduced and defined as the following equation (6).

[0030]

number

[0031] Furthermore, we require that the following formula (7) be satisfied.

[0032]

number

[0033] Then, instead of the above equations (2) to (3), the above equations (5) to (7) are formulated as constraints for the optimization planning problem. a d(a)+λq k (x (p) +a|X) st The optimization problem represented by equations (1), (4), (5) to (7) is formulated, and the optimal operation amount a is calculated by solving this.

[0034] Unlike conventional methods, this proposed method can formulate the constraint conditions for the optimization problem to find the optimal operation amount a in a time proportional to N. Therefore, even when N is large, it is possible to formulate the constraint conditions in a shorter time than with conventional methods, and as a result, the optimal operation amount a can be presented to the operator in a short time, enabling them to quickly avoid or recover from an abnormality.

[0035] In addition, Min a d(a)+λq k (x (p) +a|X) st The optimization problems represented by equations (1), (4), and (5) to (7) become mixed integer programming problems when H(·) is a linear model, and can be solved quickly using known optimization methods. On the other hand, when H(·) is a nonlinear model, the problem can be solved using known metaheuristics or the like.

[0036] In addition, by using the Mahalanobis distance instead of the Euclidean distance as the distance function d, it becomes possible to take into account the correlation between driving data. Therefore, for example, it becomes possible to calculate the optimal operation amount while taking into account the actual driving in the past and the difficulty of the operation during that driving.

[0037] In addition, the outlier degree q k When calculating, the distance function Δ(·,·) is set to l 1By using the Mahalanobis distance, the number of constraints and auxiliary variables can be reduced when formulating a mixed integer programming problem, and therefore, in this case, the mixed integer programming problem is simplified and can be solved more quickly.

[0038] <Overall configuration example of plant control system 1> An example of the overall configuration of a plant control system 1 according to this embodiment is shown in Fig. 2. As shown in Fig. 2, the plant control system 1 according to this embodiment includes an operation support device 10, a control device 20, and a plant 30. Here, the operation support device 10 and the control device 20 are communicatively connected via an arbitrary communication network, and similarly, the control device 20 and the plant 30 are communicatively connected via an arbitrary communication network.

[0039] The operation support device 10 acquires operation data of the plant 30 from the control device 20, and diagnoses the occurrence of an abnormality or the like in the plant 30 from the operation data. Furthermore, when an abnormality or the like in the plant 30 is diagnosed, the operation support device 10 presents the operator with the optimal operation for returning the plant 30 to normal. In this way, the operation of the plant 30 by the operator is supported.

[0040] The control device 20 collects operation data from the plant 30 and controls the plant 30 in accordance with an operation set by an operator. Here, an example of the control device 20 is a programmable logic controller (PLC). The operator may set an operation directly for the control device 20, or may set an operation for the control device 20 using a terminal or device that can communicate with the control device 20 via a communication network. In particular, the operator may set an operation for the control device 20 using the operation assistance device 10.

[0041] The plant 30 is equipment, facilities, etc. that execute various processes under the control of the control device 20. Specific examples of the plant 30 include a waste incineration plant, a petrochemical plant, a food plant, an iron and steel plant, etc.

[0042] 2 is merely an example, and other configurations may be used. For example, the operation support device 10 and the control device 20 may be integrated, or a terminal used by an operator may be included in the plant control system 1.

[0043] <Hardware configuration example of the driving support device 10> An example of a hardware configuration of the driving support device 10 according to this embodiment is shown in Fig. 3. As shown in Fig. 3, the driving support device 10 according to this embodiment has an input device 101, a display device 102, an external I / F 103, a communication I / F 104, a RAM (Random Access Memory) 105, a ROM (Read Only Memory) 106, an auxiliary storage device 107, and a processor 108. Each of these pieces of hardware is connected to each other via a bus 109 so as to be able to communicate with each other.

[0044] The input device 101 is, for example, a keyboard, a mouse, a touch panel, a physical button, etc. The display device 102 is, for example, a display, a display panel, etc. Note that the driving assistance device 10 does not necessarily have to have at least one of the input device 101 and the display device 102.

[0045] The external I / F 103 is an interface with an external device such as a recording medium 103a. Examples of the recording medium 103a include a CD (Compact Disc), a DVD (Digital Versatile Disk), an SD memory card (Secure Digital memory card), and a USB (Universal Serial Bus) memory card.

[0046] The communication I / F 104 is an interface for connecting the driving assistance device 10 to a communication network. The RAM 105 is a volatile semiconductor memory (storage device) that temporarily stores programs and data. The ROM 106 is a non-volatile semiconductor memory (storage device) that can store programs and data even when the power is turned off. The auxiliary storage device 107 is a non-volatile storage device such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive) in which programs and data are stored. The processor 108 is various types of arithmetic devices such as a CPU (Central Processing Unit).

[0047] 3 is an example, and the driving support device 10 may have other hardware configurations. For example, the driving support device 10 may have multiple auxiliary storage devices 107 and multiple processors 108, or may have various hardware other than the hardware shown in the figure.

[0048] <Example of functional configuration of driving assistance device 10> FIG. 4 shows an example of a functional configuration of the driving support device 10 according to this embodiment. As shown in FIG. 4, the driving support device 10 according to this embodiment has an input unit 201, an abnormality diagnosis unit 202, a formulation unit 203, an operation amount calculation unit 204, and an output unit 205. These units are realized, for example, by a process in which one or more programs installed in the driving support device 10 are executed by the processor 108 or the like. The driving support device 10 according to this embodiment also has a database 206. The database 206 is realized, for example, by the auxiliary storage device 107 or the like. Note that the database 206 may be realized, for example, by a storage device or the like communicably connected to the driving support device 10 via a communication network.

[0049] The input unit 201 inputs the operation data of the abnormality diagnosis target (i.e., the diagnosis target data x (p) ) is acquired (input) from the control device 20.

[0050] The abnormality diagnosis unit 202 uses the abnormality diagnosis model H stored in the database 206 and the diagnosis target data x acquired by the input unit 201. (p) and is used to diagnose whether an abnormality or the like has occurred in the plant 30. Generally, an abnormality diagnosis model that takes into account a time delay (that is, a time delay from when some operation is performed until a response to the operation is measured) is created in the plant 30, and therefore, hereinafter, it is assumed that the abnormality diagnosis model H is a machine learning model that takes into account the time delay of the plant 30.

[0051] When the abnormality diagnosis unit 202 diagnoses that the plant 30 is abnormal, the formulation unit 203 calculates the Min a d(a)+λq k (x (p) +a|X) st The optimization problems are formulated as equations (1), (4), (5) to (7).

[0052] The manipulated variable calculation unit 204 calculates the optimal manipulated variable a for returning the plant 30 from an abnormality to a normal state by solving the optimization problem formulated by the formulation unit 203. a d(a)+λq k (x (p) +a|X) st The optimal operation amount a is calculated by solving the optimization problems of equations (1), (4), and (5) to (7).

[0053] The output unit 205 outputs the optimal manipulated variable a calculated by the manipulated variable calculation unit 204 to a predetermined output destination and presents it to an operator of the plant 30. As a result, the optimal manipulated variable a is set in the control device 20 by the operator, and the plant 30 is controlled according to the manipulated variable a. Note that the output destination of the optimal manipulated variable a may be, for example, the display device 102, or a terminal or device communicably connected via a communication network.

[0054] The database 206 stores various data including an abnormality diagnosis model H and a set X of past normal data.

[0055] <Driving assistance processing> The operation support process according to this embodiment will be described below with reference to Fig. 5. Note that steps S101 to S106 in Fig. 5 are repeatedly executed at predetermined time intervals (for example, at each control period of the plant 30).

[0056] Step S101: The input unit 201 receives diagnostic object data x (p) is obtained from the control device 20.

[0057] Step S102: Next, the abnormality diagnosis unit 202 compares the abnormality diagnosis model H stored in the database 206 with the diagnosis target data x acquired in the above step S101. (p) In other words, the abnormality diagnosis unit 202 diagnoses whether or not an abnormality has occurred in the plant 30 by using H(x (p) ) is calculated to be +1 or -1, and the presence or absence of an abnormality or the like occurring in the plant 30 is diagnosed based on the value.

[0058] Step S103: Next, the abnormality diagnosis unit 202 judges whether or not it has been diagnosed in the above step S102 that an abnormality or the like has occurred in the plant 30. If it has been diagnosed that an abnormality or the like has occurred in the plant 30, the process proceeds to step S104, and if not, the process in this iteration is terminated.

[0059] Step S104: When it is diagnosed in the above step S103 that an abnormality or the like has occurred in the plant 30, the formulation unit 203 a d(a)+λq k (x (p) +a|X) st The optimization problems are formulated as equations (1), (4), (5) to (7).

[0060] Step S105: Next, the operation amount calculation unit 204 calculates the optimal operation amount a for returning the plant 30 from the abnormality to the normal state by solving the optimization problem formulated in the above step S104. a d(a)+λq k (x(p) +a|X) st The optimal operation amount a is calculated by solving the optimization problems of equations (1), (4), and (5) to (7).

[0061] Step S106: The output unit 205 outputs the optimal manipulated variable a calculated in the above step S105 to a predetermined output destination and presents it to an operator or the like of the plant 30. This allows the operator or the like to set the manipulated variable a in the control device 20, and the plant 30 to be controlled in accordance with this manipulated variable a.

[0062] <Comparison of calculation time between conventional and proposed methods> Here, the calculation times of the conventional method and the proposed method will be explained with reference to Fig. 6. Fig. 6 is a graph plotting the relationship between the number of normal data N and the calculation time for the conventional method and the proposed method, with the horizontal axis representing the number of normal data N and the vertical axis representing the calculation time (s) required to solve the optimization problem.

[0063] In this case, if the horizontal axis is x and the vertical axis is y, the relationship between the number of normal data N and the calculation time in the conventional method is y = 0.0068x 2 The relationship between the number of normal data N and the calculation time was approximated by a quadratic equation, -0.256x+8.8299. On the other hand, in the proposed method, the relationship between the number of normal data N and the calculation time was approximated by a linear equation, y=0.3054x-16.032. In addition, when comparing when N=500, the proposed method was able to speed up the calculation time by 14.2 times compared to the conventional method.

[0064] This shows that the proposed method can reduce calculation time compared to conventional methods (especially when the number of normal data N is large).

[0065] <Summary> As described above, when an abnormality or the like occurs in the plant 30, the driving support device 10 according to this embodiment can calculate the optimal manipulated variable a for returning the plant 30 from the abnormality to normal. Moreover, at this time, the driving support device 10 according to this embodiment can formulate the optimization problem for determining the optimal manipulated variable a in a shorter time than the conventional method. Therefore, the driving support device 10 according to this embodiment can present the optimal manipulated variable a to the operator in a short time, making it possible to quickly avoid or recover from the abnormality.

[0066] In the above embodiment, the optimal operation amount a is presented to the operator, etc., but this is not limited to this. For example, the operation assistance device 10 may set the operation amount a in the control device 20, so that the plant 30 automatically returns to normal.

[0067] <Modification> In the above embodiment, when an abnormality or the like is diagnosed in the plant 30, the optimal manipulated variable a for returning the plant 30 from the abnormality to normal is calculated. In addition, each element a included in the manipulated variable a is d The operation sequence for (d=1, , D) may also be calculated. Here, the operation sequence is the sequence of operations for the operation amount a = (a 1 ,···,a D ) is the order of the nonzero elements of a = (a 1 ,a 2 ), a 1 →a 2 If you operate in the order of a 2 →a 1 This is because the state of the plant 30 may be different if the operation is performed in the order of a = (a 1 ,···,a D ) represents state variables that do not require or cannot be manipulated.

[0068] In the following, we will define V as a vector whose elements are the subscripts of state variables that can be manipulated by operators, σ as a mapping that rearranges the elements of V, and Σ as the set of all such mappings. For example, x=(x 1 ,x 2 ,x 3 ,x 4 ,x 5 ) if the operational state variable is x 1 , x 3 Then, V=(1,3) and Σ={σ 1 ,σ 2}, σ 1 ((1,3))=(1,3), σ 2 ((1,3))=(3,1).

[0069] In order to calculate the operation amount for the optimal operation amount a, in this modified example, a causal model that represents the causal relationship between state variables is created, and then a method called OrderedCE disclosed in Reference 1 is used.

[0070] The causal model can be created from past driving data using a method called LiNGAM (linear non-Gaussian acyclic model), which is one of the statistical causal search methods. Hereinafter, this causal model is represented as D, and is expressed in a form such as an adjacency matrix.

[0071] In this case, when d(·) is a distance function (e.g., Euclidean distance or Mahalanobis distance), the objective function is d(a)+λq k (x (p) +a|X)+γC ord (a, σ|D), and the constraints are given by equations (1), (4), and (5) to (7), and the operation amount a and the mapping σ that minimize the objective function are calculated. a,σ d(a)+λq k (x (p) +a|X)+γC ord(a,σ|D) st Formulate the optimization problem represented by equations (1), (4), (5) to (7), and solve it to calculate the optimal operation amount a and mapping σ. This mapping σ is the rearrangement of the non-zero elements of the optimal operation amount a, and σ(V) represents the operation order of the non-zero elements of the optimal operation amount a. Here, γ is a weight parameter (hyperparameter). Also, C ord (a,σ|D) is a function that represents the cost due to differences in the order of operations and is defined in advance.

[0072] In addition, Min a,σ d(a)+λq k (x (p) +a|X)+γC ord (a,σ|D) st Equations (1), (4), and (5) to (7) are mixed integer programming problems when H(·) is a linear model, and can be solved quickly using known optimization methods. On the other hand, when H(·) is a nonlinear model, they can be solved using known metaheuristics or the like.

[0073] When this modified example is used, in step S106 in FIG. 5, in addition to the optimal manipulated variable a, σ(V) representing the manipulation order of the non-zero elements is also output.

[0074] The present invention is not limited to the above-described specifically disclosed embodiments, and various modifications, changes, combinations with known technologies, etc. are possible without departing from the scope of the claims.

[0075] [References] Reference 1: Kentaro Kanamori, Takuya Takagi, Ken Kobayashi, Yuichi Ike, Kento Uemura, Hiroki Arimura, "Ordered Counterfactual Explanation by Mixed-Integer Linear Optimization", 35th AAAI Conference on Artificial Intelligence (AAAI 2021) [Explanation of symbols]

[0076] 1 Plant control system 10 Driving assistance devices 20 Control device 30 Plant 101 Input Device 102 Display device 103 External I / F 103a Recording media 104 Communication I / F 105 RAM 106 ROM 107 Auxiliary storage 108 processors 109 Bus 201 Input section 202 Abnormality Diagnosis Unit 203 Formulation part 204 Operation amount calculation section 205 Output section 206 Database

Claims

1. an abnormality diagnosis unit that uses operation data of a controlled object and an abnormality diagnosis model to diagnose whether or not an abnormality or a symptom of the abnormality has occurred in the controlled object; a formulation unit that formulates an optimization problem for calculating an optimal operation for returning the control target to a normal state when it is diagnosed that an abnormality or a symptom of the abnormality has occurred in the control target; an operation amount calculation unit that calculates the optimal operation by solving the optimization problem; an output unit that outputs the optimal operation to a predetermined output destination; having The formulation unit: a driving assistance device that formulates an optimization problem for minimizing the magnitude of the operation for returning the controlled object to normal and the degree of deviation from a distribution of normal data representing past driving data of the controlled object when the controlled object is normal, the optimization problem including a first constraint condition indicating that a distance between the normal data and the driving data after the operation is equal to or greater than a distance between a closest normal data to the driving data after the operation and the driving data after the operation.

2. The amount of operation of the operation is a=(a １ , ..., a Ｄ ), the number of normal data is N, π ｄ，ｉ A ｄ A binary variable indicates whether to set the i-th change amount to v ｎ is a binary variable representing whether the n-th normal data is the nearest neighbor of the driving data after the operation, c ｄ，ｉ ｎ A ｄ is a distance between the value of the d-th state variable of the operation data after the operation when the i-th change amount is set to be i and the value of the d-th state variable of the n-th normal data, The first constraint condition is: The distance d between the nearest normal data and the post-operation operation data ｍｉｎ o v ｎ c ｄ，ｉ ｎ π ｄ，ｉ i=1,...,I ｄ (However, I ｄ Is ｄ , the number of change amounts that can be set for d = 1, ..., D, and n = 1, ..., N, the sum for d = 1, ..., D, and n = 1, ..., N, the distance d ｎ A ｄ，ｉ ｎ π ｄ，ｉ i=1,...,I ｄ and d = 1, ..., D as a sum, d ｍｉｎ ≦d ｎ The driving assistance device according to claim 1 , wherein n=1, . . . , N.

3. The optimization problem includes: a second constraint condition indicating that the driving data after the operation is input to the abnormality diagnosis model and diagnosed as normal; The binary variable v １ , ..., v Ｎ and a third constraint indicating that there is only one binary variable representing the closest neighbor of the driving data after the operation among the above.

4. The formulation unit: Under the first constraint, the second constraint, and the third constraint, d(a)+λq ｋ (x （ｐ） + a | X) (where d(·) is the distance function, x （ｐ） is the operating data, q ｋ (x （ｐ） + a | X) is x when the normal data set X is given （ｐ） 4. The driving support device according to claim 3, wherein a problem of determining an operation amount a that minimizes (a+a is an outlier, and λ is a weighting parameter) is formulated as the optimization problem.

5. The formulation unit: Each element a of the operation amount a ｄ The driving support device according to claim 2 , further comprising: formulating, as the optimization problem, a problem of determining an operation sequence of operations represented by:

6. an abnormality diagnosis procedure for diagnosing whether or not an abnormality or a symptom of the abnormality has occurred in the controlled object by using operation data of the controlled object and an abnormality diagnosis model; a formulation procedure for formulating an optimization problem for calculating an optimal operation for returning the control target to a normal state when an abnormality or a symptom of the abnormality is diagnosed in the control target; a manipulation amount calculation step for calculating the optimal manipulation by solving the optimization problem; an output step of outputting the optimal operation to a predetermined output destination; The computer executes The formulation procedure includes: A driving assistance method comprising: formulating an optimization problem for minimizing the magnitude of an operation for returning the controlled object to normal and the degree of deviation from a distribution of normal data representing past driving data of the controlled object when the controlled object is normal, the optimization problem including a first constraint condition indicating that a distance between the normal data and the driving data after the operation is equal to or greater than a distance between a closest normal data to the driving data after the operation and the driving data after the operation.

7. an abnormality diagnosis procedure for diagnosing whether or not an abnormality or a symptom of the abnormality has occurred in the controlled object by using operation data of the controlled object and an abnormality diagnosis model; a formulation procedure for formulating an optimization problem for calculating an optimal operation for returning the control target to a normal state when an abnormality or a symptom of the abnormality is diagnosed in the control target; a manipulation amount calculation step for calculating the optimal manipulation by solving the optimization problem; an output step of outputting the optimal operation to a predetermined output destination; Run the following on your computer: The formulation procedure includes: a program that formulates an optimization problem for minimizing the magnitude of the operation for returning the controlled object to normal and the degree of deviation from a distribution of normal data representing past operating data of the controlled object when the controlled object is normal, the optimization problem including a first constraint condition that indicates that the distance between the normal data and the operating data after the operation is greater than or equal to the distance between the normal data closest to the operating data after the operation and the operating data after the operation.

Images