A method and apparatus for checking a control instruction execution process

Abstract

The application provides a check method and device for a control instruction execution process, which comprises the following steps: step 1: obtaining multi-element data cached locally; step 2: when a control instruction is issued, determining an extraction strategy corresponding to the control instruction based on a preset control instruction-extraction strategy library; step 3: extracting target data from the multi-element data based on the extraction strategy; and step 4: checking the execution process of the control instruction based on the target data. The check method and device for the control instruction execution process extract target data based on the extraction strategy corresponding to the first control instruction, check the execution process of the first control instruction based on the target data, and check whether the executor reliably and accurately executes the control command.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a method and apparatus for verifying the execution process of control instructions. Background Technology

[0002] Currently, with the gradual advancement of the Internet of Things (IoT), the upgrading of information technology and intelligence is gradually deepening from small-scale business pilots to core businesses. In the control of core equipment in many key industries (such as the power industry), errors can cause serious accidents resulting in loss of life and property, placing extremely high demands on the reliability of IoT control. Therefore, after a control command is issued, verifying whether the actuator has reliably and accurately executed the control command is crucial. Summary of the Invention

[0003] One of the objectives of this invention is to provide a method and apparatus for verifying the execution process of a control command. The method extracts target data based on an extraction strategy corresponding to a first control command, and verifies the execution process of the first control command based on the target data, thereby verifying whether the actuator has reliably and accurately executed the control command.

[0004] An embodiment of the present invention provides a method for verifying the execution process of control instructions, comprising:

[0005] Step 1: Retrieve local cached metadata;

[0006] Step 2: When the first control command is issued, the extraction strategy corresponding to the first control command is determined based on the preset control command extraction strategy library;

[0007] Step 3: Extract target data from multi-source data based on the extraction strategy;

[0008] Step 4: Verify the execution process of the first control command based on the target data.

[0009] Preferably, step 3: Based on the extraction strategy, extract the target data from the multi-source data, including:

[0010] The extraction strategy is analyzed to obtain at least one target requirement;

[0011] Determine the issuance time of the first control command, extract the first data corresponding to the demand target before the issuance time from the multi-source data, and at the same time, extract the second data corresponding to the demand target after the issuance time from the multi-source data.

[0012] The first data and the corresponding second data are combined to obtain the first data to be verified;

[0013] Obtain the type of requirement for the target, which includes: environmental parameter data, video stream data, and audio stream data;

[0014] associating the first to-be-verified data with the corresponding demand type;

[0015] integrating the first to-be-verified data to obtain target data, and completing extraction.

[0016] Preferably, step 4: verifying the executed process of the first control instruction based on the target data, comprising:

[0017] when the demand type is environmental parameter data, extracting the first to-be-verified data of the demand type in the target data as second to-be-verified data;

[0018] extracting the first data in the second to-be-verified data as third data, and extracting the second data in the second to-be-verified data as fourth data;

[0019] determining a first expected value corresponding to the first control instruction and the demand type based on a preset control instruction-demand type-expected value library;

[0020] verifying the executed process of the first control instruction through the following formula:

[0021]

[0022]

[0023] wherein Y1 is the first verification result, a value of 1 indicates that the verification passes, and a value of 0 indicates that the verification fails, is an intermediate variable, C1 is the first expected value, T1 is a preset first error value, is the third data, is the fourth data, and Mean is an average value function;

[0024] when the demand type is video stream data, extracting the first to-be-verified data of the demand type in the target data as third to-be-verified data;

[0025] extracting the first data in the third to-be-verified data as fifth data, and extracting the second data in the third to-be-verified data as sixth data;

[0026] determining a second expected value corresponding to the first control instruction and the demand type based on a preset control instruction-demand type-expected value library;

[0027] verifying the executed process of the first control instruction through the following formula:

[0028]

[0029] Where Y2 is the second verification result, a value of 1 indicates verification passed, and a value of 0 indicates verification failed, F v This is a pre-defined image analysis deep learning algorithm function. This is the fifth data point. C2 is the sixth data point, C2 is the second expected value, and T2 is the preset second error value.

[0030] When the demand type is audio stream data, extract the first data to be verified associated with the demand type from the target data and use it as the fourth data to be verified.

[0031] Extract the first data from the fourth data to be verified and use it as the seventh data. At the same time, extract the second data from the fourth data to be verified and use it as the eighth data.

[0032] Based on a preset control instruction-demand type-expected value library, determine the third expected value that corresponds to both the first control instruction and the demand type.

[0033] The execution process of the first control instruction is verified using the following formula:

[0034]

[0035] Where Y3 is the third verification result, a value of 1 indicates verification passed, and a value of 0 indicates verification failed. A This is a function for a deep learning algorithm for time-series audio information. This is the seventh data point. C3 is the eighth data point, C3 is the third expected value, and T3 is the preset third error value.

[0036] Preferably, step 1: Obtain multi-data elements cached locally, including:

[0037] Obtain a preset cache node set, which includes: multiple first cache nodes;

[0038] Based on a preset cache node-cache object library, at least one cache object of the first cache node is determined. The cache object includes: the uploader and the upload device.

[0039] When the cached object is the uploader, retrieve the uploader's first upload record. The first upload record includes multiple first record items.

[0040] Extract the operation objective, operation time point, and at least one first operator from the first record item;

[0041] Based on a preset operation target-3D space library, determine the 3D space corresponding to the operation target;

[0042] determine, based on a preset operator-face information library, face information corresponding to the first operator within a preset first time period before and / or after the operation time node, the face information including: a face position and a plurality of face orientations;

[0043] map the face position to a three-dimensional space based on a preset mapping rule to obtain a first position, and map the face orientation to the three-dimensional space to obtain a first direction;

[0044] obtain a second position of the operation target in the three-dimensional space and a second direction of the operation target display;

[0045] obtain a preset correct operation determination model, input the first position, the first direction, the second position and the second direction into the correct operation determination model to obtain a first determination result;

[0046] when the first determination result is correct operation, the corresponding first operator is taken as a second operator;

[0047] determine a first experience value required by the operation target based on a preset operation target-experience value library;

[0048] determine a second experience value of the second operator based on a preset operator-experience value library;

[0049] if the second experience value of the second operator is less than the first experience value, determine the difference between the second experience value and the first experience value;

[0050] query a preset operation target-difference-severity value library to determine a first severity value corresponding to the operation target and the difference;

[0051] sum the first severity value to obtain a first severity value sum;

[0052] if the first severity value sum is greater than or equal to a preset first severity value threshold, the corresponding first cache node is removed;

[0053] when the cache object is an upload device, obtain a second upload record of the upload device, the second upload record including: a plurality of second record items;

[0054] extract a target device, a record type and a record time node from the second record item;

[0055] determine state information corresponding to the target device within a preset second time period before and / or after the record time node based on a preset target device-state information library;

[0056] determine at least one verification target corresponding to the record type based on a preset record type-verification target library;

[0057] extract target data corresponding to the verification target from the state information;

[0058] performing defect scanning on the target data to obtain a defect value;

[0059] determining a second severity value corresponding to the target device and the defect value based on a preset target device-defect value-severity value library;

[0060] summing the second severity values to obtain a second severity value sum;

[0061] if the second severity value sum is greater than or equal to a preset second severity value sum threshold, removing the first cache node corresponding thereto;

[0062] after all the first cache nodes that need to be removed in the first cache node are removed, taking the remaining first cache nodes as second cache nodes;

[0063] obtaining cache data through the second cache nodes;

[0064] integrating the obtained cache data to obtain multi-element data, and completing the obtaining.

[0065] Preferably, the verification method of the execution process of the control instruction further comprises:

[0066] constructing a control instruction-misjudgment probability library, when the verification result is a verification failure, determining a misjudgment probability corresponding to the first control instruction based on the control instruction-misjudgment probability library, and if the misjudgment probability is greater than or equal to a preset probability threshold, re-verifying the execution process of the first control instruction based on the target data;

[0067] wherein, constructing the control instruction-misjudgment probability library comprises:

[0068] obtaining a preset control instruction set, the control instruction set comprising: a plurality of second control instructions;

[0069] obtaining a preset active acquisition mode set, the active acquisition mode set comprising: a plurality of active acquisition modes;

[0070] based on the active acquisition mode, actively acquiring at least one first misjudgment event corresponding to the second control instruction;

[0071] obtaining a time node of generation of the first misjudgment event, and simultaneously, obtaining operation records corresponding to a first source of the first misjudgment event;

[0072] sorting a plurality of third record items in the operation records in time sequence to obtain a record sequence;

[0073] determining a first position in the record sequence corresponding to the time node of generation;

[0074] determining a verification strategy corresponding to the second control instruction based on a preset control instruction-verification strategy library;

[0075] parsing the verification policy to obtain at least one verification direction and a verification range;

[0076] extracting a third record item in the verification range in the verification direction of the first position in the record sequence as a fourth record item;

[0077] obtaining a preset misoperation judgment model, inputting the second control instruction and the fourth record item into the misoperation judgment model, and obtaining a second judgment result;

[0078] if the second judgment result is that there is a misoperation, eliminating the corresponding first misjudgment event;

[0079] after the first misjudgment event that needs to be eliminated in the first misjudgment event is eliminated, taking the remaining first misjudgment event as a second misjudgment event;

[0080] when at least one third misjudgment event corresponding to the second control instruction is passively obtained, obtaining a second source of the third misjudgment event;

[0081] obtaining a plurality of third sources guaranteeing the second source, and determining a guarantee relationship between the second source and the third source, the guarantee relationship including direct guarantee and indirect guarantee;

[0082] if the guarantee relationship is direct guarantee, obtaining a first guarantee circle of the third source;

[0083] determining a plurality of fourth sources existing in the first guarantee circle;

[0084] determining at least one first malicious record corresponding to the fourth source based on a preset source-malicious record library;

[0085] obtaining a preset risk assessment model, inputting the first malicious record into the risk assessment model, and obtaining a first risk value;

[0086] summing up the first risk value to obtain a first risk value sum;

[0087] if the guarantee relationship is indirect guarantee, obtaining at least one fifth source guaranteeing the second source and the third source at the same time;

[0088] obtaining a second guarantee circle of the fifth source;

[0089] determining a plurality of sixth sources existing in the second guarantee circle;

[0090] determining at least one second malicious record corresponding to the sixth source based on the source-malicious record library;

[0091] inputting the second malicious record into the risk assessment model to obtain a second risk value;

[0092] aggregate the second risk value to obtain a second risk value sum;

[0093] if the first risk value sum is less than or equal to a preset first risk value sum threshold and / or the second risk value sum is less than or equal to a preset second risk value sum threshold, the third misjudgment event corresponding to the third misjudgment event to be removed is removed;

[0094] after the third misjudgment event to be removed in the third misjudgment event is removed, the remaining third misjudgment event is taken as a fourth misjudgment event;

[0095] obtain a preset probability determination model, input the second misjudgment event and the fourth misjudgment event into the probability determination model, and obtain a probability;

[0096] combine the probability with a corresponding control instruction to obtain a control group;

[0097] obtain a preset blank database, and store the control group into the blank database;

[0098] after the control group to be stored in the blank database is stored, the blank database is taken as a control instruction-misjudgment probability library, and the construction is completed.

[0099] The control instruction execution process verification system provided by the embodiment of the application comprises:

[0100] an obtaining module, configured to obtain multi-element data cached locally;

[0101] a determining module, configured to determine an extraction strategy corresponding to the first control instruction based on a preset control instruction-extraction strategy library when the first control instruction is issued;

[0102] an extraction module, configured to extract target data from the multi-element data based on the extraction strategy;

[0103] a verification module, configured to verify the execution process of the first control instruction based on the target data.

[0104] Preferably, the extraction module performs the following operations:

[0105] analyze the extraction strategy to obtain at least one demand target;

[0106] determine a time of issuing the first control instruction, extract first data corresponding to the demand target before the time of issuing from the multi-element data, and simultaneously extract second data corresponding to the demand target after the time of issuing from the multi-element data;

[0107] combine the first data and the corresponding second data to obtain first verification data;

[0108] obtain a demand type of the demand target, and the demand type comprises environmental parameter data, video stream data and audio stream data;

[0109] associate the first to-be-verified data with the corresponding demand type;

[0110] integrate the first to-be-verified data to obtain target data, and complete the extraction.

[0111] Preferably, the verification module performs the following operations:

[0112] When the demand type is environmental parameter data, the first to-be-verified data associated with the demand type in the target data is extracted and used as second to-be-verified data;

[0113] The first data in the second to-be-verified data is extracted and used as third data, and the second data in the second to-be-verified data is extracted and used as fourth data;

[0114] Based on the preset control instruction-demand type-expected value library, a first expected value corresponding to the first control instruction and the demand type is determined;

[0115] The executed process of the first control instruction is verified by the following formula:

[0116]

[0117]

[0118] wherein Y1 is the first verification result, a value of 1 indicates that the verification passes, and a value of 0 indicates that the verification fails, is an intermediate variable, C1 is the first expected value, and T1 is a preset first error value, is the third data, is the fourth data, and Mean is an average value function;

[0119] When the demand type is video stream data, the first to-be-verified data associated with the demand type in the target data is extracted and used as third to-be-verified data;

[0120] The first data in the third to-be-verified data is extracted and used as fifth data, and the second data in the third to-be-verified data is extracted and used as sixth data;

[0121] Based on the preset control instruction-demand type-expected value library, a second expected value corresponding to the first control instruction and the demand type is determined;

[0122] The executed process of the first control instruction is verified by the following formula:

[0123]

[0124] wherein Y2 is the second verification result, a value of 1 indicates that the verification passes, and a value of 0 indicates that the verification fails, vThe preset image analysis deep learning algorithm function, The fifth data, The sixth data, C2 is a second expected value, and T2 is a preset second error value.

[0125] When the demand type is an audio stream data, the first to-be-verified data associated with the demand type in the target data is extracted and taken as the fourth to-be-verified data.

[0126] The first data in the fourth to-be-verified data is extracted and taken as the seventh data, and the second data in the fourth to-be-verified data is extracted and taken as the eighth data.

[0127] Based on the preset control instruction-demand type-expected value library, a third expected value corresponding to the first control instruction and the demand type is determined.

[0128] The executed process of the first control instruction is verified through the following formula:

[0129]

[0130] Wherein, Y3 is a third verification result, a value of 1 indicates that the verification passes, and a value of 0 indicates that the verification fails, F A The time series audio information deep learning algorithm function, The seventh data, The eighth data, C3 is a third expected value, and T3 is a preset third error value.

[0131] Preferably, the acquisition module performs the following operations:

[0132] A preset cache node set is acquired, and the cache node set includes a plurality of first cache nodes.

[0133] Based on a preset cache node-cache object library, at least one cache object of the first cache node is determined, and the cache object includes an upload person and an upload device.

[0134] When the cache object is the upload person, a first upload record of the upload person is acquired, and the first upload record includes a plurality of first record items.

[0135] An operation target, an operation time node, and at least one first operator are extracted from the first record item.

[0136] Based on a preset operation target-three-dimensional space library, a three-dimensional space corresponding to the operation target is determined.

[0137] Based on a preset operator-information-oriented library, information oriented to the first operator within a preset first time period before and / or after the operation time node is determined, and the information oriented includes a face position and a plurality of orientation directions.

[0138] mapping the face position in the three-dimensional space according to the preset mapping rule to obtain a first position, and mapping the facing direction in the three-dimensional space to obtain a first direction;

[0139] obtaining a second position of the operation target in the three-dimensional space and a second direction of the operation target display;

[0140] obtaining a preset correct operation determination model, inputting the first position, the first direction, the second position and the second direction into the correct operation determination model to obtain a first determination result;

[0141] when the first determination result is correct operation, corresponding to the first operator as the second operator;

[0142] determining the first experience value required by the operation target based on the preset operation target-experience value library;

[0143] determining the second experience value of the second operator based on the preset operator-experience value library;

[0144] if the second experience value of the second operator is less than the first experience value, determining the difference between the corresponding second experience value and the first experience value;

[0145] querying the preset operation target-difference-severity value library to determine the first severity value corresponding to the operation target and the difference;

[0146] summing up the first severity value to obtain the first severity value sum;

[0147] if the first severity value sum is greater than or equal to a preset first severity value threshold, the corresponding first cache node is removed;

[0148] when the cache object is an upload device, obtaining a second upload record of the upload device, the second upload record comprising: a plurality of second record items;

[0149] extracting the target device, the record type and the record time node from the second record item;

[0150] determining the state information of the target device corresponding to the preset second time period before and / or after the record time node based on the preset target device-state information library;

[0151] determining at least one verification target corresponding to the record type based on the preset record type-verification target library;

[0152] extracting target data corresponding to the verification target from the state information;

[0153] defect scanning on the target data to obtain a defect value;

[0154] determine a second severity value corresponding to the target device and the defect value based on a preset target device-defect value-severity value library;

[0155] aggregate the second severity values to obtain a second severity value sum;

[0156] if the second severity value sum is greater than or equal to a preset second severity value sum threshold, eliminate the first cache node corresponding thereto;

[0157] after all the first cache nodes that need to be eliminated in the first cache node are eliminated, take the remaining first cache nodes as second cache nodes;

[0158] obtain cache data through the second cache nodes;

[0159] integrate the obtained cache data to obtain multi-element data, and complete the obtaining.

[0160] Preferably, the verification system for the execution process of the control instruction further comprises:

[0161] a re-verification module configured to construct a control instruction-misjudgment probability library, when the verification result is a verification failure, determine a misjudgment probability corresponding to the first control instruction based on the control instruction-misjudgment probability library, and if the misjudgment probability is greater than or equal to a preset probability threshold, re-verify the execution process of the first control instruction based on the target data;

[0162] The re-verification module performs the following operations:

[0163] obtain a preset control instruction set, the control instruction set comprising a plurality of second control instructions;

[0164] obtain a preset active acquisition mode set, the active acquisition mode set comprising a plurality of active acquisition modes;

[0165] based on the active acquisition mode, actively acquire at least one first misjudgment event corresponding to the second control instruction;

[0166] obtain a time node at which the first misjudgment event occurs, and simultaneously, obtain operation records corresponding to a first source of the first misjudgment event;

[0167] sort a plurality of third record items in the operation records in a time sequence to obtain a record sequence;

[0168] determine a first position in the record sequence corresponding to the time node at which the first misjudgment event occurs;

[0169] based on a preset control instruction-verification strategy library, determine a verification strategy corresponding to the second control instruction;

[0170] analyze the verification strategy to obtain at least one verification direction and verification range;

[0171] extract a third record item in a verification range in a verification direction of a first position in the record sequence as a fourth record item;

[0172] obtain a preset misoperation judgment model, input the second control instruction and the fourth record item into the misoperation judgment model, and obtain a second judgment result;

[0173] if the second judgment result is that there is misoperation, eliminate the corresponding first misjudgment event;

[0174] after the first misjudgment event that needs to be eliminated in the first misjudgment event is eliminated, take the remaining first misjudgment event as a second misjudgment event;

[0175] when at least one third misjudgment event corresponding to the second control instruction is passively obtained, obtain a second source of the third misjudgment event;

[0176] obtain a plurality of third sources that guarantee the second source, and determine a guarantee relationship between the second source and the third sources, the guarantee relationship including direct guarantee and indirect guarantee;

[0177] if the guarantee relationship is direct guarantee, obtain a first guarantee circle of the third source;

[0178] determine a plurality of fourth sources existing in the first guarantee circle;

[0179] determine at least one first malicious record corresponding to the fourth source based on a preset source-malicious record library;

[0180] obtain a preset risk assessment model, input the first malicious record into the risk assessment model, and obtain a first risk value;

[0181] summarize the first risk value to obtain a first risk value sum;

[0182] if the guarantee relationship is indirect guarantee, obtain at least one fifth source that guarantees the second source and the third source at the same time;

[0183] obtain a second guarantee circle of the fifth source;

[0184] determine a plurality of sixth sources existing in the second guarantee circle;

[0185] determine at least one second malicious record corresponding to the sixth source based on the source-malicious record library;

[0186] input the second malicious record into the risk assessment model to obtain a second risk value;

[0187] summarize the second risk value to obtain a second risk value sum;

[0188] If the first risk value is less than or equal to a preset first risk value threshold and / or the second risk value is less than or equal to a preset second risk value threshold, the corresponding third misjudgment event is eliminated;

[0189] After the third misjudgment events that need to be eliminated in the third misjudgment events are eliminated, the remaining third misjudgment events are taken as fourth misjudgment events;

[0190] A preset probability determination model is obtained, the second misjudgment events and the fourth misjudgment events are input into the probability determination model, and a probability is obtained;

[0191] The probability is combined with a corresponding control instruction to obtain a control group;

[0192] A preset blank database is obtained, and the control group is stored in the blank database;

[0193] After the control groups that need to be stored in the blank database are stored, the blank database is taken as a control instruction-misjudgment probability library, and the construction is completed.

[0194] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application will be realized and achieved by means of the structures particularly pointed out in the written description and claims hereof as well as in the appended drawings.

[0195] The technical solutions of the present application will be further described in detail below with the help of the drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0196] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, illustrate the present application and explain the present application together with the embodiments of the present application, and do not constitute a limitation on the present application. In the drawings:

[0197] Figure 1 A flow chart of a verification method for a control instruction execution process in an embodiment of the present application;

[0198] Figure 2 A schematic diagram of a verification application for a control instruction execution process in an embodiment of the present application;

[0199] Figure 3 A schematic diagram of a verification system for a control instruction execution process in an embodiment of the present application. DETAILED DESCRIPTION

[0200] The preferred embodiments of the present application will be described below with the help of the drawings, and it should be understood that the preferred embodiments described herein are only used to illustrate and explain the present application, and do not constitute a limitation on the present application.

[0201] The embodiment of the present application provides a verification method for a control instruction execution process, as shown in the following Figure 1 , which comprises the following steps:

[0202] Step 1: obtaining local cached multi-element data;

[0203] Step 2: when a first control instruction is issued, determining an extraction strategy corresponding to the first control instruction based on a preset control instruction-extraction strategy library;

[0204] Step 3: extracting target data from the multi-element data based on the extraction strategy;

[0205] Step 4: verifying the execution process of the first control instruction based on the target data.

[0206] The working principle and beneficial effects of the above technical solution are as follows:

[0207] The preset control instruction-extraction strategy library is specifically a database containing extraction strategies corresponding to different control instructions, for example, the control instruction is to close a valve, and the extraction strategy is to extract the monitoring data (video and audio, etc.) corresponding to the valve;

[0208] As shown in the following Figure 2 , each device (for example, a video monitoring device and an audio monitoring device, etc.) collects multi-element data (for example, video stream data and audio stream data), which is cached locally when uploaded to a server; when a first control instruction (for example, closing a valve) is issued, target data (for example, video data and audio data of the valve before and after the instruction is issued) is extracted based on the corresponding extraction strategy; and the execution process of the first control instruction is verified based on the target data (for example, whether the valve is represented by opening to closing in the video data before and after the instruction is issued, whether the valve closing sound appears before and after the instruction is issued, etc.).

[0209] The embodiment of the present application extracts target data based on the extraction strategy corresponding to the first control instruction, verifies the execution process of the first control instruction based on the target data, and realizes the verification of whether the executor reliably and accurately executes the control command.

[0210] The embodiment of the present application provides a verification method for a control instruction execution process, and step 3: extracting target data from multi-element data based on an extraction strategy, comprises the following steps:

[0211] Analyzing the extraction strategy to obtain at least one demand target;

[0212] Determining an issuing time of the first control instruction, extracting first data corresponding to the demand target before the issuing time from the multi-element data, and simultaneously extracting second data corresponding to the demand target after the issuing time from the multi-element data;

[0213] combining the first data and the corresponding second data to obtain first to-be-verified data;

[0214] obtaining a demand type of the demand target, the demand type including: environmental parameter data, video stream data and audio stream data;

[0215] associating the first to-be-verified data with the corresponding demand type;

[0216] integrating the first to-be-verified data to obtain target data, and completing extraction.

[0217] The working principle and beneficial effects of the above technical solution are as follows:

[0218] analyzing the demand target in the extraction strategy (for example, the first control instruction is to close a certain valve, and the demand target is the video stream data of the valve); extracting the first data and the second data before and after the first control instruction is issued from the multi-element data; combining the first data and the second data to obtain the first to-be-verified data, and associating the first to-be-verified data with the corresponding demand type; and integrating the first to-be-verified data to obtain the target data.

[0219] The embodiment of the application provides a verification method for a control instruction execution process, and step 4 is based on the target data to verify the executed process of the first control instruction, including:

[0220] When the demand type is environmental parameter data, the first to-be-verified data associated with the demand type in the target data is extracted and used as second to-be-verified data;

[0221] The first data in the second to-be-verified data is extracted and used as third data, and the second data in the second to-be-verified data is extracted and used as fourth data;

[0222] Based on a preset control instruction-demand type-expected value library, a first expected value corresponding to the first control instruction and the demand type is determined;

[0223] The executed process of the first control instruction is verified through the following formula:

[0224]

[0225]

[0226] Y1 is the first verification result, the value 1 indicates that the verification passes, and the value 0 indicates that the verification fails, is an intermediate variable, C1 is the first expected value, T1 is a preset first error value, is the third data, is the fourth data, and Mean is an average value function;

[0227] When the demand type is video stream data, first to-be-verified data associated with the demand type in the target data is extracted and taken as third to-be-verified data;

[0228] First data in the third to-be-verified data is extracted and taken as fifth data, and second data in the third to-be-verified data is extracted and taken as sixth data;

[0229] Based on the preset control instruction-demand type-expected value library, a second expected value corresponding to the first control instruction and the demand type is determined;

[0230] The executed process of the first control instruction is verified through the following formula:

[0231]

[0232] Wherein, Y2 is the second verification result, the value of 1 indicates that the verification passes, and the value of 0 indicates that the verification fails, F v is a preset image analysis deep learning algorithm function, is the fifth data, is the sixth data, C2 is the second expected value, and T2 is a preset second error value;

[0233] When the demand type is audio stream data, first to-be-verified data associated with the demand type in the target data is extracted and taken as fourth to-be-verified data;

[0234] First data in the fourth to-be-verified data is extracted and taken as seventh data, and second data in the fourth to-be-verified data is extracted and taken as eighth data;

[0235] Based on the preset control instruction-demand type-expected value library, a third expected value corresponding to the first control instruction and the demand type is determined;

[0236] The executed process of the first control instruction is verified through the following formula:

[0237]

[0238] Wherein, Y3 is the third verification result, the value of 1 indicates that the verification passes, and the value of 0 indicates that the verification fails, F A is a time series audio information deep learning algorithm function, is the seventh data, is the eighth data, C3 is the third expected value, and T3 is a preset third error value.

[0239] The working principle and beneficial effects of the above technical solution are as follows:

[0240] The control instruction-demand type-expected value library specifically refers to a database containing expected values corresponding to different control instructions and different demand types, for example, the control instruction is to close a certain valve, the demand type is video stream data, and the expected value is 6.

[0241] The data (environment parameter data) of the digital / analog quantity acquisition unit is verified by the following function:

[0242]

[0243] Wherein, is a two-dimensional matrix composed of (the number of digital / analog sensor data groups, the length of the time series); first, the data groups of the time series t0 and t1 before and after the control command is issued are averaged, then the data change difference before and after the control command is calculated, and the result is compared with the expected value C1. If the absolute value difference is within the error threshold range, the verification is successful, and the next step of verifying the video stream data is entered; otherwise, the verification fails, and an exception is reported to the center;

[0244] The video stream data is calculated using an artificial intelligence deep learning algorithm. is a multi-frame image matrix extracted before and after the control instruction is issued; represents an image analysis deep learning algorithm, and according to the actual accuracy, computing resources, power consumption and other requirements in different application scenarios, a suitable deep learning model is selected (for example, the image analysis deep learning algorithm function can use the CNN algorithm), and the output result is compared with the expected value C2. If the absolute value is within the error threshold range, the verification is passed, and the next step of verifying the audio stream data is entered; otherwise, the verification fails.

[0245] The audio stream data is calculated using a time series analysis related artificial intelligence deep learning algorithm. is an audio sequence matrix before and after the control instruction is issued; represents a time series information related deep learning algorithm, and according to the actual accuracy, computing resources, power consumption and other requirements in different application scenarios, a suitable deep learning model is selected (for example, the time series audio information deep learning algorithm function can use the BERT algorithm to analyze the audio time series information in a typical scenario), and the output result is compared with the expected value C3. If the absolute value is within the error threshold range, the verification is passed; otherwise, the verification fails.

[0246] The embodiment of the application provides a verification method for the execution process of a control instruction, step 1: acquiring local cached multi-element data, including:

[0247] A preset cache node set is acquired, and the cache node set includes: a plurality of first cache nodes;

[0248] determine at least one cache object of the first cache node based on a preset cache node-cache object library, the cache object comprising: an uploading person and an uploading device;

[0249] when the cache object is the uploading person, obtain a first uploading record of the uploading person, the first uploading record comprising: a plurality of first record items;

[0250] extract an operation target, an operation time node and at least one first operation person from the first record items;

[0251] determine a three-dimensional space corresponding to the operation target based on a preset operation target-three-dimensional space library;

[0252] determine facing information corresponding to the first operation person within a preset first time period before and / or after the operation time node based on a preset operation person-facing information library, the facing information comprising: a face position and a plurality of facing directions;

[0253] map the face position in the three-dimensional space to obtain a first position, and map the facing directions in the three-dimensional space to obtain a first direction based on a preset mapping rule;

[0254] obtain a second position of the operation target in the three-dimensional space and a second direction displayed by the operation target;

[0255] obtain a preset correct operation judgment model, input the first position, the first direction, the second position and the second direction into the correct operation judgment model to obtain a first judgment result;

[0256] when the first judgment result is correct operation, take the corresponding first operation person as a second operation person;

[0257] determine a first experience value required by the operation target based on a preset operation target-experience value library;

[0258] determine a second experience value of the second operation person based on a preset operation person-experience value library;

[0259] if the second experience value of the second operation person is less than the first experience value, determine a difference value between the corresponding second experience value and the first experience value;

[0260] query a preset operation target-difference value-severity value library to determine a first severity value corresponding to the operation target and the difference value;

[0261] sum the first severity value to obtain a first severity value sum;

[0262] if the first severity value sum is greater than or equal to a preset first severity value sum threshold, eliminate the corresponding first cache node;

[0263] When the cache object is the uploading device, a second uploading record of the uploading device is acquired, and the second uploading record includes a plurality of second record items;

[0264] The target device, the record type, and the record time node are extracted from the second record item;

[0265] Based on a preset target device-state information library, it is determined that the target device corresponds to state information in a preset second time period before and / or after the record time node;

[0266] Based on a preset record type-verification target library, it is determined that the record type corresponds to at least one verification target;

[0267] Target data corresponding to the verification target is extracted from the state information;

[0268] Defect scanning is performed on the target data to obtain a defect value;

[0269] Based on a preset target device-defect value-severity value library, it is determined that the target device and the defect value correspond to a second severity value together;

[0270] The second severity values are summarized to obtain a second severity value sum;

[0271] If the second severity value sum is greater than or equal to a preset second severity value sum threshold, the corresponding first cache node is removed;

[0272] After the first cache nodes that need to be removed in the first cache nodes are removed, the remaining first cache nodes are taken as second cache nodes;

[0273] Cache data is acquired through the second cache nodes;

[0274] The acquired cache data is integrated to obtain multi-element data, and the acquisition is completed.

[0275] The working principle and beneficial effects of the above technical solution are as follows:

[0276] The preset cache node-cache object library specifically refers to a database containing cache objects corresponding to different cache nodes, and the cache objects are divided into uploaders and uploading devices, that is, when the uploaders upload data to the server, the cache nodes corresponding to the uploaders cache the data uploaded by the uploaders, and the uploading devices are the same; the preset operation target-three-dimensional space library specifically refers to a database containing three-dimensional spaces corresponding to different operation targets, for example: the operation target is a certain valve, and the three-dimensional space is a one-to-one virtual space in the range near the valve; the preset operator-oriented information library specifically refers to a database containing different operator-oriented information libraries corresponding to different operators, and the oriented information specifically refers to the face position (which can be collected by a positioning sensor) and the face orientation (which can be collected by an orientation sensor) of the operator; the preset first time period specifically refers to, for example, 5 seconds; the preset mapping rule specifically refers to, for example, the operation target is a certain valve, the positional relationship between the face and the valve is determined, the face position is mapped to the three-dimensional space, and the face orientation is also mapped to the three-dimensional space; the preset correct operation judgment model specifically refers to a model generated by learning a large number of records of whether the operation is correct determined by artificial judgment, for example: the model is used to determine whether the face of the operator is close to the valve (less than the normal visible distance) and to determine whether the face orientation of the operator meets the visible relationship with the orientation of the valve (generally, the angle between the vectors of the two is obtuse); the preset operation target-experience value library specifically refers to a database containing experience values required by different operation targets, for example: the operation target is a certain valve, and the experience value is 60; the preset operation target-difference-severity value library specifically refers to a database containing severity values corresponding to different operation targets and different difference values, the greater the difference value, the more verified the event, and the greater the severity value; the preset first severity value and threshold value specifically refer to, for example, 388; the preset target device-state information library specifically refers to a database containing state information corresponding to different target devices, for example: the target device is a monitoring device for collecting images of a certain valve, and the state information is temperature, load, etc.; the preset second time period specifically refers to, for example, 3 seconds; the preset record type-verification target library specifically refers to a database containing verification targets corresponding to different record types, for example: the record type is a valve video record, and the corresponding verification target is to verify whether the temperature is normal and the direction is aligned, etc.; the preset target device-defect value-severity value library specifically refers to a database containing severity values corresponding to different target devices and different defect values, the greater the defect value, the greater the severity value; the preset second severity value and threshold value specifically refer to 425; the aggregation specifically refers to summation calculation;

[0277] In actual use, some devices cannot be collected by sensors and the like, for example: the state parameters of high-temperature operation devices, the sensor will be affected by the temperature, and manual inspection is needed for collection, therefore, the cache objects of the first cache node are divided into two types of upload people and upload devices, when the cache object is an upload person, it is necessary to verify whether the upload person has a record of irregular operation, that is, to verify whether the first record item (including record data, operation target, operation time node and operation person) is in compliance, to verify whether the first operation person correctly performs the operation (for example: facing the valve and keeping a small distance can be seen) for collection, if so, the corresponding first operation person is taken as a second operation person, however, since the inspection and collection of some devices are relatively complex (for example: numerical fluctuation, experienced inspection personnel can decisively obtain the intermediate value or mode around it), it is determined whether the second experience value of the second operation person is higher than the first experience value required by the operation target, if both are not, the upload person produces irregular operation, the first severity value and are determined, when the cache object is an upload device, it is determined whether the state of the target device is normal before and after the record time node of the record, if not, the second severity value and are determined, based on the first severity value and and the second severity value and, the corresponding first cache node is removed, and the second cache node is obtained; the cache data is obtained through the second cache node, integrated, and multi-element data is obtained; the accuracy of the multi-element data acquisition is greatly ensured, the data quality of the multi-element data is improved, at the same time, the operation person's mistake cost is improved, and the mistake probability is reduced to a certain extent.

[0278] The embodiment of the application provides a verification method for a control instruction execution process, and also comprises:

[0279] A control instruction-misjudgment probability library is constructed, when the verification result is not passed, the misjudgment probability corresponding to the first control instruction is determined based on the control instruction-misjudgment probability library, if the misjudgment probability is greater than or equal to a preset probability threshold, the executed process of the first control instruction is re-verified based on the target data;

[0280] The control instruction-misjudgment probability library is constructed, comprising:

[0281] A preset control instruction set is obtained, and the control instruction set comprises: a plurality of second control instructions;

[0282] A preset active acquisition mode set is obtained, and the active acquisition mode set comprises: a plurality of active acquisition modes;

[0283] At least one first misjudgment event corresponding to the second control instruction is actively acquired based on the active acquisition mode;

[0284] The generation time node of the first misjudgment event is obtained, and at the same time, the operation record corresponding to the first source of the first misjudgment event is obtained;

[0285] Sort the plurality of third record items in the operation record in time sequence to obtain a record sequence;

[0286] Determine a first position in the record sequence corresponding to a time node;

[0287] Determine a verification strategy corresponding to the second control instruction based on a preset control instruction-verification strategy library;

[0288] Parse the verification strategy to obtain at least one verification direction and a verification range;

[0289] Extract a third record item in the verification range in the verification direction of the first position in the record sequence as a fourth record item;

[0290] Obtain a preset misoperation judgment model, input the second control instruction and the fourth record item into the misoperation judgment model, and obtain a second judgment result;

[0291] If the second judgment result is that there is a misoperation, eliminate the corresponding first misjudgment event;

[0292] When all the first misjudgment events that need to be eliminated in the first misjudgment event are eliminated, the remaining first misjudgment events are taken as second misjudgment events;

[0293] When at least one third misjudgment event corresponding to the second control instruction is passively obtained, obtain a second source of the third misjudgment event;

[0294] Obtain a plurality of third sources that guarantee the second source, and determine a guarantee relationship between the second source and the third source, the guarantee relationship including direct guarantee and indirect guarantee;

[0295] If the guarantee relationship is direct guarantee, obtain a first guarantee circle of the third source;

[0296] Determine a plurality of fourth sources existing in the first guarantee circle;

[0297] Determine at least one first malicious record corresponding to the fourth source based on a preset source-malicious record library;

[0298] Obtain a preset risk assessment model, input the first malicious record into the risk value assessment model, and obtain a first risk value;

[0299] Summarize the first risk value to obtain a first risk value sum;

[0300] If the guarantee relationship is indirect guarantee, obtain at least one fifth source that guarantees the second source and the third source at the same time;

[0301] Obtain a second guarantee circle of the fifth source;

[0302] determining a plurality of sixth sources existing in the second guarantee circle;

[0303] determining at least one second malicious record corresponding to the sixth source based on the source-malicious record library;

[0304] inputting the second malicious record into the risk assessment model to obtain a second risk value;

[0305] summing the second risk value to obtain a second risk value sum;

[0306] if the first risk value sum is less than or equal to a preset first risk value sum threshold and / or the second risk value sum is less than or equal to a preset second risk value sum threshold, eliminating the third misjudgment event corresponding thereto;

[0307] after the third misjudgment event that needs to be eliminated in the third misjudgment event is eliminated, taking the remaining third misjudgment event as a fourth misjudgment event;

[0308] obtaining a preset probability determination model, inputting the second misjudgment event and the fourth misjudgment event into the probability determination model to obtain a probability;

[0309] combining the probability with the corresponding control instruction to obtain a control group;

[0310] obtaining a preset blank database, and storing the control group into the blank database;

[0311] after the control group that needs to be stored into the blank database is stored, taking the blank database as a control instruction-misjudgment probability library to complete the construction.

[0312] The working principle and beneficial effects of the above technical solution are as follows:

[0313] The preset probability threshold is, for example, 0.2; the active acquisition mode is, for example, actively entering the database of other institutions that also use the above-mentioned instruction execution process verification method to perform retrieval; the preset control instruction-verification strategy library is, for example, a database containing verification strategies corresponding to different control instructions, such as: the control instruction is to close a certain valve, and the verification strategy is to verify whether there has been an abnormal operation (such as adding unqualified lubricant) on the valve within 3 days (verification range) before the time node of the false positive event (verification direction); the preset misoperation determination model is, for example, a model generated by using a machine learning algorithm to learn a large number of manually determined misoperation records, which can determine whether there is a misoperation corresponding to the control instruction in the fourth record item; the preset source-malicious record library is, for example, a database containing malicious records (data falsification) corresponding to different sources; the preset risk assessment model is, for example, a model generated by using a machine learning algorithm to learn a large number of records manually assessing the risk level of malicious records, which can output a risk value, and the larger the risk value, the higher the risk level; the preset first risk value and threshold are, for example, 855; the preset second risk value and threshold are, for example, 750; the preset probability determination model is, for example, a model generated by using a machine learning algorithm to learn a large number of records manually determining the probability of false positive events; the preset blank database is, for example, a database with no content; the aggregation is, for example, summation calculation;

[0314] Due to the immaturity of time series audio information deep learning algorithms and other issues, verifying the first control instruction execution process will inevitably result in false positives, which is a probability problem; therefore, the false positive probability corresponding to different first control instructions can be determined, and when the probability is high, re-verification is needed;

[0315] In constructing the control instruction-misjudgment probability library, first, the misjudgment events are acquired, the acquisition modes are divided into active acquisition (active acquisition by itself, and the other party allows itself to enter the database of the other party) and passive acquisition (sent by the other party, and the other party does not allow itself to enter the database of the other party), when the acquisition mode is active acquisition, it is necessary to verify whether there is a misoperation in the verification range before and / or after the time node generated by the first misjudgment event acquired by the operation record of the corresponding first source, if yes, the corresponding first misjudgment event is removed, for example: the first control instruction is to close the valve, then the valve cannot be closed due to that the user does not add lubricant in time or adds unqualified lubricant, causing the valve to be blocked; when the acquisition mode is passive acquisition, first, the guarantee relationship between the third source guaranteeing the second source of the acquired third misjudgment event and the second source is determined, the guarantee relationship is divided into two kinds, which are direct guarantee (the third source guarantees the second source) and indirect guarantee (the third source and the second source are both guaranteed by the same source, and a guarantee relationship is formed between them); when the guarantee relationship is direct guarantee, the first guarantee circle of the third source (the second source is also in the guarantee circle) is acquired, based on the malicious record of the fourth source in the first guarantee circle, the first risk value and are determined, the greater the first risk value and, the faster the progress of the invalidation of the first guarantee circle (when the sources in the guarantee circle produce malicious records, the life of the guarantee circle will be reduced, when the guarantee circle is invalidated, the guarantee relationship does not exist either); when the guarantee relationship is indirect guarantee, the fifth source guaranteeing the second source and the third source at the same time is determined; the second guarantee circle of the fifth source is acquired, the second risk value and are determined; the greater the second risk value and, the faster the progress of the invalidation of the second guarantee circle; the corresponding third misjudgment event is removed, the probability is determined based on the remaining fourth misjudgment event, which greatly improves the accuracy of the probability determination and guarantees the accuracy of the data acquisition.

[0316] The embodiment of the application provides a verification method for a control instruction execution process, and also comprises:

[0317] Randomly acquiring a verification flow for verifying the execution process of the first control instruction;

[0318] Splitting the verification flow into a plurality of first flows;

[0319] Acquiring a preset specification verification model, inputting the first flow into the specification verification model, and acquiring a specification value;

[0320] If the specification value is less than or equal to a preset specification threshold value, the first flow corresponding to the first flow is taken as a second flow, and the remaining first flows are taken as third flows;

[0321] Acquiring a preset influence analysis model, inputting the second flow and the third flow into the influence analysis model, and acquiring a plurality of influence values;

[0322] a decision index is calculated based on the specification value, the specification threshold value and the influence value, and the calculation formula is as follows:

[0323]

[0324] wherein γ is the decision index, β i is the i th influence value, n is the total number of the influence values, α is the specification value, α 0 is the specification threshold value, and σ 1 and σ 2 are preset weight values;

[0325] If the decision index is greater than or equal to a preset decision index threshold value, the executed process of the corresponding first control instruction is re-verified.

[0326] The working principle and beneficial effects of the above technical solution are as follows:

[0327] The preset specification verification model is specifically a model generated by learning records of a large number of manual verification process specifications using a machine learning algorithm, which can verify process specification and output a specification value; the greater the specification value, the more standardized the process; the preset specification threshold value is specifically, for example, 95; the preset influence analysis model is specifically a model generated by learning records of a large number of manual process non-standardization impact on the rest of the process using a machine learning algorithm, which can analyze the influence between processes and output multiple influence values; the greater the influence value, the greater the influence; and the preset decision index threshold value is specifically, for example, 75;

[0328] The embodiment of the application randomly acquires a verification process for verifying the executed process of the first control instruction, calculates a decision index, and further verifies the executed process of the first control instruction when the decision index is greater than or equal to the decision index threshold value, thereby improving the standardization of the system; in the formula, the influence value and the decision index are positively correlated, and the specification value and the decision index are negatively correlated.

[0329] The embodiment of the application provides a verification system for a control instruction executed process, as shown in Figure 3 , which comprises:

[0330] An acquisition module 1 is configured to acquire multi-element data cached locally;

[0331] A determination module 2 is configured to determine an extraction strategy corresponding to a first control instruction based on a preset control instruction-extraction strategy library when the first control instruction is issued.

[0332] An extraction module 3 is configured to extract target data from the multi-element data based on the extraction strategy.

[0333] A verification module 4 is configured to verify the executed process of the first control instruction based on the target data.

[0334] The working principle and beneficial effects of the above technical solution are:

[0335] The preset control instruction-extraction strategy library specifically is a database containing extraction strategies corresponding to different control instructions, for example, the control instruction is to close a certain valve, and the extraction strategy is to extract the monitoring data (video and audio, etc.) corresponding to the valve.

[0336] Each device (for example, video monitoring device and audio monitoring device, etc.) collects multi-element data (for example, video stream data, audio stream data), which is cached locally when uploaded to the server; when the first control instruction (for example, closing a certain valve) is issued, the target data (for example, video data and audio data of the valve before and after the instruction is issued) is extracted based on the corresponding extraction strategy; based on the target data, the execution process of the first control instruction is verified (for example, whether the valve is represented by opening to closing in the video data before and after the instruction is issued, whether the valve closing sound appears before and after the instruction is issued, etc.).

[0337] The embodiment of the application extracts target data based on the extraction strategy corresponding to the first control instruction, and verifies the execution process of the first control instruction based on the target data, thereby realizing the verification of whether the executor reliably and accurately executes the control command.

[0338] The embodiment of the application provides a control instruction execution process verification system, and the extraction module 3 performs the following operations:

[0339] The extraction strategy is analyzed to obtain at least one demand target;

[0340] The issuance time of the first control instruction is determined, the first data corresponding to the demand target before the issuance time is extracted from the multi-element data, and the second data corresponding to the demand target after the issuance time is extracted from the multi-element data;

[0341] The first data and the corresponding second data are combined to obtain first verification data;

[0342] The demand type of the demand target is obtained, and the demand type includes environmental parameter data, video stream data and audio stream data;

[0343] The first verification data is associated with the corresponding demand type;

[0344] The first verification data is integrated to obtain target data, and the extraction is completed.

[0345] The working principle and beneficial effects of the above technical solution are:

[0346] The demand target in the extraction strategy is analyzed (for example, the first control instruction is to close a valve, and the demand target is the video stream data of the valve); the first data and the second data before and after the first control instruction is issued are extracted from the multi-element data; the first data and the second data are combined to obtain the first to-be-verified data, and are associated with the corresponding demand type; the target data is obtained by integrating the first to-be-verified data.

[0347] The embodiment of the application provides a verification system for a control instruction execution process, and the verification module 4 performs the following operations:

[0348] When the demand type is environmental parameter data, the first to-be-verified data associated with the demand type in the target data is extracted and used as second to-be-verified data;

[0349] The first data in the second to-be-verified data is extracted and used as third data, and the second data in the second to-be-verified data is extracted and used as fourth data;

[0350] Based on the preset control instruction-demand type-expected value library, the first expected value corresponding to the first control instruction and the demand type is determined;

[0351] The execution process of the first control instruction is verified through the following formula:

[0352]

[0353]

[0354] Wherein Y1 is the first verification result, the value 1 represents that the verification passes, and the value 0 represents that the verification fails, is an intermediate variable, C1 is the first expected value, T1 is the preset first error value, is the third data, is the fourth data, and Mean is the mean function;

[0355] When the demand type is video stream data, the first to-be-verified data associated with the demand type in the target data is extracted and used as third to-be-verified data;

[0356] The first data in the third to-be-verified data is extracted and used as fifth data, and the second data in the third to-be-verified data is extracted and used as sixth data;

[0357] Based on the preset control instruction-demand type-expected value library, the second expected value corresponding to the first control instruction and the demand type is determined;

[0358] The execution process of the first control instruction is verified through the following formula:

[0359]

[0360] wherein Y2 is a second check result, a value of 1 represents that the check passes, and a value of 0 represents that the check fails, F v is a preset image analysis deep learning algorithm function, is fifth data, is sixth data, C2 is a second expected value, and T2 is a preset second error value;

[0361] When the demand type is audio stream data, the first to-be-checked data associated with the demand type in the target data is extracted and taken as fourth to-be-checked data;

[0362] The first data in the fourth to-be-checked data is extracted and taken as seventh data, and the second data in the fourth to-be-checked data is extracted and taken as eighth data;

[0363] Based on a preset control instruction-demand type-expected value library, a third expected value corresponding to the first control instruction and the demand type is determined;

[0364] The executed process of the first control instruction is checked through the following formula:

[0365]

[0366] wherein Y3 is a third check result, a value of 1 represents that the check passes, and a value of 0 represents that the check fails, F A is a time series audio information deep learning algorithm function, is seventh data, is eighth data, C3 is a third expected value, and T3 is a preset third error value.

[0367] The working principle and beneficial effects of the above technical solution are as follows:

[0368] The control instruction-demand type-expected value library specifically refers to a database containing expected values corresponding to different control instructions and different demand types, for example, the control instruction is to close a certain valve, the demand type is video stream data, and the expected value is 6.

[0369] The data (environmental parameter data) of the digital / analog quantity acquisition unit is checked, and the following function is used for calculation:

[0370]

[0371] wherein, A two-dimensional matrix composed of (the number of digital / analog sensor data groups, the length of the time series); first, the data groups of the time series t0, t1 before and after the control command is issued are averaged, then the data change difference before and after the control command is calculated, and the result is compared with the expected value C1. If the absolute value difference is within the error threshold range, the step of verification is successful, and the next step of verifying and calculating the video stream data is entered; otherwise, the verification fails, and the center is reported to be abnormal;

[0372] The video stream data is calculated using an artificial intelligence deep learning algorithm. A plurality of image matrices extracted from the video before and after the control command is issued; An image analysis deep learning algorithm is used to calculate the video stream data. According to the actual accuracy, computing resources, power consumption and other requirements in different application scenarios, a suitable deep learning model is selected (for example, the image analysis deep learning algorithm function can use the CNN algorithm), and the output result is compared with the expected value C2. If the absolute value is within the error threshold range, the step of verification is passed, and the next step of verifying the audio stream data is entered; otherwise, the verification fails.

[0373] The audio stream data is calculated using a time series analysis related artificial intelligence deep learning algorithm. A plurality of audio sequence matrices before and after the control command is issued; A time series information related deep learning algorithm is used to calculate the audio stream data. According to the actual accuracy, computing resources, power consumption and other requirements in different application scenarios, a suitable deep learning model is selected (for example, the time series audio information deep learning algorithm function can use the BERT algorithm to analyze the audio time series information in a typical scenario), and the output result is compared with the expected value C3. If the absolute value is within the error threshold range, the verification is passed; otherwise, the verification fails.

[0374] The embodiment of the present application provides a verification system for a control instruction execution process, and the obtaining module 1 performs the following operations:

[0375] Obtain a preset cache node set, and the cache node set includes a plurality of first cache nodes;

[0376] Based on the preset cache node-cache object library, at least one cache object of the first cache node is determined, and the cache object includes an upload person and an upload device;

[0377] When the cache object is the upload person, the first upload record of the upload person is obtained, and the first upload record includes a plurality of first record items;

[0378] The operation target, operation time node and at least one first operator are extracted from the first record item;

[0379] determine the three-dimensional space corresponding to the operation target based on the preset operation target-three-dimensional space library;

[0380] determine the facing information corresponding to the first operator within a preset first time period before and / or after the operation time node based on the preset operator-facing information library, the facing information including: a face position and a plurality of facing directions;

[0381] map the face position in the three-dimensional space to obtain a first position, and map the facing direction in the three-dimensional space to obtain a first direction based on the preset mapping rule;

[0382] obtain a second position of the operation target in the three-dimensional space and a second direction displayed by the operation target;

[0383] obtain a preset correct operation judgment model, input the first position, the first direction, the second position and the second direction into the correct operation judgment model to obtain a first judgment result;

[0384] when the first judgment result is correct operation, the corresponding first operator is taken as a second operator;

[0385] determine the first experience value required by the operation target based on the preset operation target-experience value library;

[0386] determine the second experience value of the second operator based on the preset operator-experience value library;

[0387] if the second experience value of the second operator is less than the first experience value, determine the difference between the corresponding second experience value and the first experience value;

[0388] query the preset operation target-difference-severity value library to determine the first severity value corresponding to the operation target and the difference;

[0389] sum the first severity value to obtain a first severity value sum;

[0390] if the first severity value sum is greater than or equal to a preset first severity value threshold, the corresponding first cache node is removed;

[0391] when the cache object is an upload device, obtain a second upload record of the upload device, the second upload record including a plurality of second record items;

[0392] extract the target device, the record type and the record time node from the second record item;

[0393] determine the state information corresponding to the target device within a preset second time period before and / or after the record time node based on the preset target device-state information library;

[0394] determine at least one verification target corresponding to the record type based on a preset record type-verification target library;

[0395] extract target data corresponding to the verification target from the state information;

[0396] perform defect scanning on the target data to obtain a defect value;

[0397] determine a second severity value corresponding to the target device and the defect value based on a preset target device-defect value-severity value library;

[0398] sum up the second severity values to obtain a second severity value sum;

[0399] if the second severity value sum is greater than or equal to a preset second severity value sum threshold, remove the corresponding first cache node;

[0400] after all the first cache nodes that need to be removed in the first cache nodes are removed, take the remaining first cache nodes as second cache nodes;

[0401] obtain cache data through the second cache nodes;

[0402] integrate the obtained cache data to obtain multi-element data, and complete the obtaining.

[0403] The working principle and beneficial effects of the above technical solution are as follows:

[0404] The preset cache node-cache object library specifically refers to a database containing cache objects corresponding to different cache nodes, and the cache objects are divided into uploaders and uploading devices, that is, when the uploaders upload data to the server, the cache nodes corresponding to the uploaders cache the data uploaded by the uploaders, and the uploading devices are the same; the preset operation target-three-dimensional space library specifically refers to a database containing three-dimensional spaces corresponding to different operation targets, for example: the operation target is a certain valve, and the three-dimensional space is a one-to-one virtual space in the range near the valve; the preset operator-oriented information library specifically refers to a database containing different operator-oriented information libraries corresponding to different operators, and the oriented information specifically refers to the face position (which can be collected by a positioning sensor) and the face orientation (which can be collected by an orientation sensor) of the operator; the preset first time period specifically refers to, for example, 5 seconds; the preset mapping rule specifically refers to, for example, the operation target is a certain valve, the positional relationship between the face and the valve is determined, the face position is mapped to the three-dimensional space, and the face orientation is also mapped to the three-dimensional space; the preset correct operation judgment model specifically refers to a model generated by learning a large number of records of whether the operation is correct determined by artificial judgment, for example: the model is used to determine whether the face of the operator is close to the valve (less than the normal visible distance) and to determine whether the face orientation of the operator meets the visible relationship with the orientation of the valve (generally, the angle between the vectors of the two is obtuse); the preset operation target-experience value library specifically refers to a database containing experience values required by different operation targets, for example: the operation target is a certain valve, and the experience value is 60; the preset operation target-difference-severity value library specifically refers to a database containing severity values corresponding to different operation targets and different difference values, the greater the difference value, the more verified the event, and the greater the severity value; the preset first severity value and threshold value specifically refer to, for example, 388; the preset target device-state information library specifically refers to a database containing state information corresponding to different target devices, for example: the target device is a monitoring device for collecting images of a certain valve, and the state information is temperature, load, etc.; the preset second time period specifically refers to, for example, 3 seconds; the preset record type-verification target library specifically refers to a database containing verification targets corresponding to different record types, for example: the record type is a valve video record, and the corresponding verification target is to verify whether the temperature is normal and the direction is aligned, etc.; the preset target device-defect value-severity value library specifically refers to a database containing severity values corresponding to different target devices and different defect values, the greater the defect value, the greater the severity value; the preset second severity value and threshold value specifically refer to 425; the aggregation specifically refers to summation calculation;

[0405] In actual use, some devices cannot be collected by sensors and the like, for example: the state parameters of high-temperature operation devices, the sensor will be affected by the temperature, and manual inspection is needed for collection, therefore, the cache objects of the first cache node are divided into two types of upload people and upload devices, when the cache object is an upload person, it is necessary to verify whether the upload person has a record of irregular operation, that is, to verify whether the first record item (including record data, operation target, operation time node and operation person) is in compliance, to verify whether the first operation person correctly performs the operation (for example: facing the valve and keeping a small distance can be seen) for collection, if so, the corresponding first operation person is taken as a second operation person, however, since the inspection and collection of some devices are relatively complex (for example: numerical fluctuation, experienced inspection personnel can decisively obtain the median value or mode around it), it is determined whether the second experience value of the second operation person is higher than the first experience value required by the operation target, if both are not, the upload person produces irregular operation, and the first severity value and are determined; when the cache object is an upload device, it is determined whether the state of the target device is normal before and after the record time node of the record, if not, the second severity value and are determined; based on the first severity value and and the second severity value and, the corresponding first cache node is removed, and the second cache node is obtained; the cache data is obtained through the second cache node, integrated, and the multivariate data is obtained; the accuracy of the multivariate data acquisition is greatly ensured, the data quality of the multivariate data is improved, at the same time, the operation person's mistake cost is improved, and the mistake probability is reduced to a certain extent.

[0406] The embodiment of the application provides a verification system for controlling the execution process of an instruction, and also comprises:

[0407] The re-verification module 4 is configured to construct a control instruction-misjudgment probability library, and when the verification result is a verification failure, determine a misjudgment probability corresponding to the first control instruction based on the control instruction-misjudgment probability library, and if the misjudgment probability is greater than or equal to a preset probability threshold, re-verify the executed process of the first control instruction based on the target data.

[0408] The re-verification module 4 performs the following operations:

[0409] Obtain a preset control instruction set, and the control instruction set comprises a plurality of second control instructions.

[0410] Obtain a preset active acquisition mode set, and the active acquisition mode set comprises a plurality of active acquisition modes.

[0411] Based on the active acquisition mode, at least one first misjudgment event corresponding to the second control instruction is actively acquired.

[0412] Obtain a time node at which the first misjudgment event occurs, and simultaneously, obtain operation records corresponding to a first source of the first misjudgment event.

[0413] Sort the plurality of third record items in the operation record in time sequence to obtain a record sequence;

[0414] Determine a first position in the record sequence corresponding to a time node of generation;

[0415] Based on the preset control instruction-verification strategy library, determine the verification strategy corresponding to the second control instruction;

[0416] Parse the verification strategy to obtain at least one verification direction and verification range;

[0417] Extract the third record item in the verification range in the verification direction of the first position in the record sequence, and take it as a fourth record item;

[0418] Obtain a preset misoperation judgment model, input the second control instruction and the fourth record item into the misoperation judgment model, and obtain a second judgment result;

[0419] If the second judgment result is that there is a misoperation, eliminate the corresponding first misjudgment event;

[0420] When all the first misjudgment events that need to be eliminated in the first misjudgment event are eliminated, the remaining first misjudgment events are taken as second misjudgment events;

[0421] When at least one third misjudgment event corresponding to the second control instruction is passively obtained, obtain a second source of the third misjudgment event;

[0422] Obtain a plurality of third sources that guarantee the second source, and determine a guarantee relationship between the second source and the third source, the guarantee relationship including direct guarantee and indirect guarantee;

[0423] If the guarantee relationship is direct guarantee, obtain a first guarantee circle of the third source;

[0424] Determine a plurality of fourth sources existing in the first guarantee circle;

[0425] Based on a preset source-malicious record library, determine at least one first malicious record corresponding to the fourth source;

[0426] Obtain a preset risk assessment model, input the first malicious record into the risk value assessment model, and obtain a first risk value;

[0427] Summarize the first risk value to obtain a first risk value sum;

[0428] If the guarantee relationship is indirect guarantee, obtain at least one fifth source that guarantees the second source and the third source at the same time;

[0429] Obtain a second guarantee circle of the fifth source;

[0430] determining a plurality of sixth sources existing in the second guarantee circle;

[0431] determining at least one second malicious record corresponding to the sixth source based on the source-malicious record library;

[0432] inputting the second malicious record into the risk assessment model to obtain a second risk value;

[0433] summing the second risk values to obtain a second risk value sum;

[0434] if the first risk value sum is less than or equal to a preset first risk value sum threshold and / or the second risk value sum is less than or equal to a preset second risk value sum threshold, eliminating the third misjudgment events corresponding thereto;

[0435] after the third misjudgment events that need to be eliminated in the third misjudgment events are eliminated, taking the remaining third misjudgment events as fourth misjudgment events;

[0436] obtaining a preset probability determination model, inputting the second misjudgment events and the fourth misjudgment events into the probability determination model to obtain a probability;

[0437] combining the probability with the corresponding control instructions to obtain a control group;

[0438] obtaining a preset blank database, and storing the control group into the blank database;

[0439] after the control groups that need to be stored in the blank database are all stored, taking the blank database as a control instruction-misjudgment probability library to complete the construction.

[0440] The working principle and beneficial effects of the above technical solutions are as follows:

[0441] The preset probability threshold is, for example, 0.2; the active acquisition mode is, for example, actively entering the database of other institutions that also use the above-mentioned instruction execution process verification method to perform retrieval; the preset control instruction-verification strategy library is, for example, a database containing verification strategies corresponding to different control instructions, such as: the control instruction is to close a certain valve, and the verification strategy is to verify whether there has been an abnormal operation (such as adding unqualified lubricant) on the valve within 3 days (verification range) before the time node of the false positive event (verification direction); the preset misoperation determination model is, for example, a model generated by using a machine learning algorithm to learn a large number of manually determined misoperation records, which can determine whether there is a misoperation corresponding to the control instruction in the fourth record item; the preset source-malicious record library is, for example, a database containing malicious records (data falsification) corresponding to different sources; the preset risk assessment model is, for example, a model generated by using a machine learning algorithm to learn a large number of records manually assessing the risk level of malicious records, which can output a risk value, and the larger the risk value, the higher the risk level; the preset first risk value and threshold are, for example, 855; the preset second risk value and threshold are, for example, 750; the preset probability determination model is, for example, a model generated by using a machine learning algorithm to learn a large number of records manually determining the probability of false positive events; the preset blank database is, for example, a database with no content; the aggregation is, for example, summation calculation;

[0442] Due to the immaturity of time series audio information deep learning algorithms and other issues, verifying the first control instruction execution process will inevitably result in false positives, which is a probability problem; therefore, the false positive probability corresponding to different first control instructions can be determined, and when the probability is high, re-verification is needed;

[0443] In constructing the control instruction-misjudgment probability library, first, the misjudgment events are acquired, the acquisition modes are divided into active acquisition (active acquisition by itself, and the other party allows itself to enter the database of the other party) and passive acquisition (sent by the other party, and the other party does not allow itself to enter the database of the other party), when the acquisition mode is active acquisition, it is necessary to verify whether there is a misoperation in the verification range before and / or after the time node generated by the first misjudgment event acquired by the operation record of the corresponding first source, if yes, the corresponding first misjudgment event is removed, for example: the first control instruction is to close the valve, then the valve cannot be closed due to that the user does not add lubricant in time or adds unqualified lubricant, causing the valve to be blocked; when the acquisition mode is passive acquisition, first, the guarantee relationship between the third source guaranteeing the second source of the acquired third misjudgment event and the second source is determined, the guarantee relationship is divided into two kinds, which are direct guarantee (the third source guarantees the second source) and indirect guarantee (the third source and the second source are both guaranteed by the same source, and a guarantee relationship is formed between them); when the guarantee relationship is direct guarantee, the first guarantee circle of the third source (the second source is also in the guarantee circle) is acquired, based on the malicious record of the fourth source in the first guarantee circle, the first risk value and are determined, the greater the first risk value and, the faster the progress of the invalidation of the first guarantee circle (when the sources in the guarantee circle produce malicious records, the life of the guarantee circle will be reduced, when the guarantee circle is invalidated, the guarantee relationship does not exist either); when the guarantee relationship is indirect guarantee, the fifth source guaranteeing the second source and the third source at the same time is determined; the second guarantee circle of the fifth source is acquired, the second risk value and are determined; the greater the second risk value and, the faster the progress of the invalidation of the second guarantee circle; the corresponding third misjudgment event is removed, the probability is determined based on the remaining fourth misjudgment event, which greatly improves the accuracy of the probability determination and guarantees the accuracy of the data acquisition.

[0444] The embodiment of the application provides a verification system for a control instruction execution process, and also comprises:

[0445] A flow verification module, which performs the following operations:

[0446] Randomly acquiring a verification flow for verifying the execution process of the first control instruction;

[0447] Splitting the verification flow into a plurality of first flows;

[0448] Acquiring a preset specification verification model, inputting the first flow into the specification verification model, and acquiring a specification value;

[0449] If the specification value is less than or equal to a preset specification threshold, the first flow is taken as a second flow, and the remaining first flows are taken as third flows;

[0450] obtain a preset influence analysis model, input the second process and the third process into the influence analysis model, and obtain a plurality of influence values;

[0451] calculate a decision index based on the standard value, a standard threshold value, and the influence values, and the calculation formula is as follows:

[0452]

[0453] wherein γ is the decision index, β i is the i-th influence value, n is the total number of the influence values, α is the standard value, α0 is the standard value threshold, and σ1 and σ2 are preset weight values;

[0454] If the decision index is greater than or equal to a preset decision index threshold value, the executed process of the corresponding first control instruction is rechecked.

[0455] The working principle and beneficial effects of the above technical solution are as follows:

[0456] The preset standard verification model is specifically a model generated by learning records of a large number of artificial verification process specifications by using a machine learning algorithm, which can verify process specifications and output a standard value; the greater the standard value, the more standardized the process; the preset standard threshold value is specifically, for example, 95; the preset influence analysis model is specifically a model generated by learning records of a large number of influence degrees of artificial process non-standardization on other processes by using a machine learning algorithm, which can analyze the influence between processes and output a plurality of influence values; the greater the influence value, the greater the influence; and the preset decision index threshold value is specifically, for example, 75.

[0457] The embodiment of the application randomly obtains a checking process for checking the executed process of the first control instruction, calculates a decision index, and further checks; when the decision index is greater than or equal to a decision index threshold value, the executed process of the first control instruction is rechecked, thereby improving the standardization of the system; in the formula, the influence value is positively correlated with the decision index, and the standard value is negatively correlated with the decision index.

[0458] Obviously, those skilled in the art can make various modifications and variations to the application without departing from the spirit and scope of the application. Thus, if these modifications and variations of the application belong to the scope of the claims of the application and their equivalent technologies, the application also intends to include these modifications and variations.

Claims

1. A checking method of a process in which an instruction is executed, characterized by, The method comprises the following steps: Step 1: obtaining local cached multi-element data; Step 2: when a first control instruction is issued, determining an extraction strategy corresponding to the first control instruction based on a preset control instruction-extraction strategy library; Step 3: extracting target data from the multi-element data based on the extraction strategy; Step 4: verifying an executed process of the first control instruction based on the target data; When the verification result is a verification failure, determining a misjudgment probability corresponding to the first control instruction based on a control instruction-misjudgment probability library, and if the misjudgment probability is greater than or equal to a preset probability threshold, re-verifying the executed process of the first control instruction based on the target data; The construction of the control instruction-misjudgment probability library comprises the following steps: actively acquiring at least one first misjudgment event corresponding to a second control instruction based on an active acquisition mode; acquiring a time node at which the first misjudgment event occurs, and simultaneously acquiring operation records corresponding to a first source of the first misjudgment event; sorting a plurality of third record items in the operation records in a time sequence to obtain a record sequence; determining a first position in the record sequence corresponding to the time node; determining a verification strategy corresponding to the second control instruction based on a preset control instruction-verification strategy library; analyzing the verification strategy to obtain at least one verification direction and a verification range; extracting third record items in the verification range in the verification direction of the first position in the record sequence as fourth record items; inputting the second control instruction and the fourth record items into a misoperation judgment model to obtain a second judgment result; if the second judgment result is that there is misoperation, eliminating the corresponding first misjudgment event; after the first misjudgment events that need to be eliminated in the first misjudgment events are all eliminated, taking the remaining first misjudgment events as second misjudgment events; when at least one third misjudgment event corresponding to the second control instruction is passively acquired, acquiring a second source of the third misjudgment event; acquiring a plurality of third sources that guarantee the second source, and determining a guarantee relationship between the second source and the third sources; if the guarantee relationship is direct guarantee, acquiring a first guarantee circle of the third source; determining a plurality of fourth sources existing in the first guarantee circle; determining at least one first malicious record corresponding to the fourth source based on a preset source-malicious record library; inputting the first malicious record into a risk value evaluation model to obtain a first risk value; summing the first risk values to obtain a first risk value sum; if the guarantee relationship is indirect guarantee, acquiring at least one fifth source that guarantees the second source and the third sources at the same time; acquiring a second guarantee circle of the fifth source; determining a plurality of sixth sources existing in the second guarantee circle; determining at least one second malicious record corresponding to the sixth source based on the source-malicious record library; inputting the second malicious record into the risk evaluation model to obtain a second risk value; summing the second risk values to obtain a second risk value sum; if the first risk value sum and / or the second risk value sum are less than or equal to preset first and second risk value sum thresholds, eliminating the corresponding third misjudgment event; after the third misjudgment events that need to be eliminated in the third misjudgment events are all eliminated, taking the remaining third misjudgment events as fourth misjudgment events; Input the second misjudgment event and the fourth misjudgment event into the probability determination model to obtain a probability; Combine the probability with the corresponding control instruction to obtain a control group; Obtain a preset blank database and store the control group into the blank database; After the control group to be stored in the blank database is stored, the blank database is used as a control instruction-misjudgment probability library to complete the construction.

2. The method of claim 1, wherein the control instruction is executed by a process, and the process is a process of a computer program. Step 3: Based on the extraction strategy, target data is extracted from the multi-element data, including: Analyzing the extraction strategy to obtain at least one demand target; Determining the issuing time of the first control instruction, extracting the first data corresponding to the demand target before the issuing time from the multi-element data, and simultaneously extracting the second data corresponding to the demand target after the issuing time from the multi-element data; Combining the first data and the corresponding second data to obtain first to-be-verified data; Obtaining the demand type of the demand target, the demand type including: environmental parameter data, video stream data and audio stream data; Associating the first to-be-verified data with the corresponding demand type; Integrating the first to-be-verified data to obtain target data, and completing the extraction.

3. The method of claim 2, wherein the control instruction is executed by a process, and the process is a process of a computer program. Step 4: Based on the target data, the executed process of the first control instruction is verified, including: When the demand type is environmental parameter data, the first to-be-verified data associated with the demand type in the target data is extracted and used as second to-be-verified data; Extracting the first data in the second to-be-verified data as third data, and simultaneously extracting the second data in the second to-be-verified data as fourth data; Based on a preset control instruction-demand type-expected value library, a first expected value corresponding to the first control instruction and the demand type is determined; The executed process of the first control instruction is verified through the following formula: wherein, is a first check result, a value of 1 indicates that the check passes, and a value of 0 indicates that the check fails, is an intermediate variable, is the first expected value, is a preset first error value, is the third data, is the fourth data, is an averaging function; When the demand type is video stream data, the first to-be-verified data associated with the demand type in the target data is extracted and used as third to-be-verified data; Extracting the first data in the third to-be-verified data as fifth data, and simultaneously extracting the second data in the third to-be-verified data as sixth data; Based on a preset control instruction-demand type-expected value library, a second expected value corresponding to the first control instruction and the demand type is determined; The executed process of the first control instruction is verified through the following formula: wherein, is a second check result, a value of 1 represents that the check passes, and a value of 0 represents that the check fails, is a preset image analysis deep learning algorithm function, is the fifth data, is the sixth data, is the second expected value, is a preset second error value; When the demand type is audio stream data, the first to-be-verified data associated with the demand type in the target data is extracted and used as fourth to-be-verified data; Extracting the first data in the fourth to-be-verified data as seventh data, and simultaneously extracting the second data in the fourth to-be-verified data as eighth data; Based on a preset control instruction-demand type-expected value library, a third expected value corresponding to the first control instruction and the demand type is determined; The executed process of the first control instruction is verified through the following formula: wherein, is a third check result, a value of 1 represents that the check passes, and a value of 0 represents that the check fails, is a time series audio information deep learning algorithm function, is the seventh data, is the eighth data, is the third expected value, is a preset third error value.

4. The method of claim 1, wherein the control instruction is executed by a process, and the process is a process of a computer program. Step 1: Obtain the multi-element data cached locally, including: Obtaining a preset cache node set, the cache node set comprising: a plurality of first cache nodes; Determining at least one cache object of the first cache node based on a preset cache node-cache object library, the cache object comprising: an upload person and an upload device; When the cache object is an upload person, obtaining a first upload record of the upload person, the first upload record comprising: a plurality of first record items; Extracting an operation target, an operation time node, and at least one first operation person from the first record items; Determining a three-dimensional space corresponding to the operation target based on a preset operation target-three-dimensional space library; Determining the facing information corresponding to the first operation person within a preset first time period before and / or after the operation time node based on a preset operation person-facing information library, the facing information comprising: a face position and a plurality of facing directions; Mapping the face position in the three-dimensional space to obtain a first position, and mapping the facing direction in the three-dimensional space to obtain a first direction based on a preset mapping rule; Obtaining a second position of the operation target in the three-dimensional space and a second direction displayed by the operation target; Obtaining a preset correct operation judgment model, and inputting the first position, the first direction, the second position, and the second direction into the correct operation judgment model to obtain a first judgment result; When the first judgment result is correct operation, corresponding to the first operation person is taken as a second operation person; Determining a first experience value required by the operation target based on a preset operation target-experience value library; Determining a second experience value of the second operation person based on a preset operation person-experience value library; If the second experience value of the second operation person is less than the first experience value, determining the difference value between the second experience value and the first experience value; Querying a preset operation target-difference value-severity value library to determine a first severity value corresponding to the operation target and the difference value; Summarizing the first severity value to obtain a first severity value sum; If the first severity value sum is greater than or equal to a preset first severity value threshold, the corresponding first cache node is removed; When the cache object is an upload device, obtaining a second upload record of the upload device, the second upload record comprising: a plurality of second record items; Extracting a target device, a record type, and a record time node from the second record items; Determining state information corresponding to the target device within a preset second time period before and / or after the record time node based on a preset target device-state information library; Determining at least one verification target corresponding to the record type based on a preset record type-verification target library; Extracting target data corresponding to the verification target from the state information; Performing defect scanning on the target data to obtain a defect value; Determining a second severity value corresponding to the target device and the defect value based on a preset target device-defect value-severity value library; Summarizing the second severity value to obtain a second severity value sum; If the second severity value sum is greater than or equal to a preset second severity value threshold, the corresponding first cache node is removed; After the first cache nodes that need to be eliminated in the first cache nodes are all eliminated, the remaining first cache nodes are taken as second cache nodes; Obtaining cache data through the second cache nodes; Integrating the obtained cache data to obtain multi-element data, and completing the obtaining.

5. A verification system for controlling the execution of instructions, characterized in that Comprise: An obtaining module, configured to obtain multi-element data cached locally; A determining module, configured to, when a first control instruction is issued, determine an extraction strategy corresponding to the first control instruction based on a preset control instruction-extraction strategy library; An extraction module, configured to extract target data from the multi-element data based on the extraction strategy; A verification module, configured to verify an executed process of the first control instruction based on the target data; When the verification result is a verification failure, determine a misjudgment probability corresponding to the first control instruction based on a control instruction-misjudgment probability library, and if the misjudgment probability is greater than or equal to a preset probability threshold, re-verify the executed process of the first control instruction based on the target data; The construction of the control instruction-misjudgment probability library comprises: Based on an active obtaining manner, at least one first misjudgment event corresponding to a second control instruction is actively obtained; The time node at which the first misjudgment event occurs is obtained, and at the same time, operation records corresponding to a first source of the first misjudgment event are obtained; A plurality of third record items in the operation records are sorted in time sequence to obtain a record sequence; A first position corresponding to the time node in the record sequence is determined; Based on a preset control instruction-verification strategy library, a verification strategy corresponding to the second control instruction is determined; The verification strategy is analyzed to obtain at least one verification direction and a verification range; The third record items in the verification range in the verification direction of the first position in the record sequence are extracted and taken as fourth record items; The second control instruction and the fourth record items are input into a misoperation judgment model to obtain a second judgment result; If the second judgment result is that there is misoperation, the corresponding first misjudgment event is eliminated; After the first misjudgment events that need to be eliminated in the first misjudgment events are all eliminated, the remaining first misjudgment events are taken as second misjudgment events; When at least one third misjudgment event corresponding to the second control instruction is passively obtained, a second source of the third misjudgment event is obtained; A plurality of third sources that guarantee the second source are obtained, and a guarantee relationship between the second source and the third sources is determined; If the guarantee relationship is direct guarantee, a first guarantee circle of the third source is obtained; A plurality of fourth sources existing in the first guarantee circle are determined; Based on a preset source-malicious record library, at least one first malicious record corresponding to the fourth source is determined; The first malicious record is input into a risk value evaluation model to obtain a first risk value; The first risk values are summarized to obtain a first risk value sum; If the guarantee relationship is indirect guarantee, at least one fifth source that guarantees the second source and the third sources at the same time is obtained; A second guarantee circle of the fifth source is obtained; A plurality of sixth sources existing in the second guarantee circle are determined; Based on the source-malicious record library, at least one second malicious record corresponding to the sixth source is determined; The second malicious record is input into a risk evaluation model to obtain a second risk value; aggregate the second risk values to obtain a second risk value sum; if the first risk value sum is less than or equal to a preset first risk value sum threshold and / or the second risk value sum is less than or equal to a preset second risk value sum threshold, eliminate the third misjudgment event corresponding thereto; after all the third misjudgment events that need to be eliminated are eliminated, take the remaining third misjudgment events as fourth misjudgment events; input the second misjudgment events and the fourth misjudgment events into a probability determination model to obtain a probability; combine the probability with a corresponding control instruction to obtain a control group; obtain a preset blank database and store the control group in the blank database; after all the control groups that need to be stored in the blank database are stored, take the blank database as a control instruction-misjudgment probability library to complete the construction.

6. A verification system for controlling the execution of a process according to claim 5, characterized in that, The extraction module performs the following operations: analyze the extraction strategy to obtain at least one demand target; determine a first control instruction issuing time, extract first data corresponding to the demand target before the issuing time from the multi-element data, and simultaneously extract second data corresponding to the demand target after the issuing time from the multi-element data; combine the first data and the corresponding second data to obtain first to-be-verified data; obtain a demand type of the demand target, the demand type including: environmental parameter data, video stream data, and audio stream data; associate the first to-be-verified data with the corresponding demand type; integrate the first to-be-verified data to obtain target data, and complete the extraction.

7. A system for checking the execution of control instructions according to claim 6, characterized in that, The verification module performs the following operations: when the demand type is environmental parameter data, extract the first to-be-verified data associated with the demand type in the target data as second to-be-verified data; extract the first data in the second to-be-verified data as third data, and simultaneously extract the second data in the second to-be-verified data as fourth data; determine a first expected value corresponding to the first control instruction and the demand type based on a preset control instruction-demand type-expected value library; verify the executed process of the first control instruction by the following formula: wherein, is a first check result, a value of 1 indicating that the check passes, and a value of 0 indicating that the check fails, is an intermediate variable, is the first expected value, is a preset first error value, is the third data, is the fourth data, is an averaging function; when the demand type is video stream data, extract the first to-be-verified data associated with the demand type in the target data as third to-be-verified data; extract the first data in the third to-be-verified data as fifth data, and simultaneously extract the second data in the third to-be-verified data as sixth data; determine a second expected value corresponding to the first control instruction and the demand type based on a preset control instruction-demand type-expected value library; verify the executed process of the first control instruction by the following formula: wherein, is a second check result, a value of 1 indicates that the check passes, and a value of 0 indicates that the check fails, is a preset image analysis deep learning algorithm function, is the fifth data, is the sixth data, is the second expected value, is a preset second error value; when the demand type is audio stream data, extract the first to-be-verified data associated with the demand type in the target data as fourth to-be-verified data; extract the first data in the fourth to-be-verified data as seventh data, and simultaneously extract the second data in the fourth to-be-verified data as eighth data; determine a third expected value corresponding to the first control instruction and the demand type based on a preset control instruction-demand type-expected value library; verify the executed process of the first control instruction by the following formula: wherein, is a third check result, a value of 1 represents that the check passes, and a value of 0 represents that the check fails, is a time series audio information deep learning algorithm function, is the seventh data, is the eighth data, is the third expected value, is a preset third error value.

8. The system of claim 5, wherein the control instruction is executed by a process, and the process is a process of a computer program. The acquisition module performs the following operations: acquire a preset cache node set, the cache node set comprising: a plurality of first cache nodes; determine at least one cache object of the first cache node based on a preset cache node-cache object library, the cache object comprising: an upload person and an upload device; when the cache object is an upload person, acquire a first upload record of the upload person, the first upload record comprising: a plurality of first record items; extract an operation target, an operation time node, and at least one first operator from the first record item; determine a three-dimensional space corresponding to the operation target based on a preset operation target-three-dimensional space library; determine the facing information corresponding to the first operator within a preset first time period before and / or after the operation time node based on a preset operator-facing information library, the facing information comprising: a face position and a plurality of facing directions; map the face position in the three-dimensional space based on a preset mapping rule to obtain a first position, and map the facing direction in the three-dimensional space to obtain a first direction; acquire a second position of the operation target in the three-dimensional space and a second direction displayed by the operation target; acquire a preset correct operation judgment model, input the first position, the first direction, the second position, and the second direction into the correct operation judgment model to obtain a first judgment result; when the first judgment result is correct operation, the corresponding first operator is taken as a second operator; determine a first experience value required by the operation target based on a preset operation target-experience value library; determine a second experience value of the second operator based on a preset operator-experience value library; if the second experience value of the second operator is less than the first experience value, determine the difference value between the second experience value and the first experience value; query a preset operation target-difference value-severity value library to determine a first severity value corresponding to the operation target and the difference value; sum the first severity value to obtain a first severity sum; if the first severity sum is greater than or equal to a preset first severity sum threshold, eliminate the corresponding first cache node; when the cache object is an upload device, acquire a second upload record of the upload device, the second upload record comprising: a plurality of second record items; extract a target device, a record type, and a record time node from the second record item; determine state information corresponding to the target device within a preset second time period before and / or after the record time node based on a preset target device-state information library; determine at least one verification target corresponding to the record type based on a preset record type-verification target library; extract target data corresponding to the verification target from the state information; perform defect scanning on the target data to obtain a defect value; determining a second severity value corresponding to the target device and the defect value based on a preset target device-defect value-severity value library; summing the second severity values to obtain a second severity value sum; if the second severity value sum is greater than or equal to a preset second severity value sum threshold, eliminating the first cache node corresponding to the first cache node; after the first cache nodes that need to be eliminated are eliminated, taking the remaining first cache nodes as second cache nodes; obtaining cache data through the second cache nodes; integrating the obtained cache data to obtain multi-element data, and completing the obtaining.

Images