CN103761172B - Hardware fault diagnosis system based on neutral net - Google Patents
Hardware fault diagnosis system based on neutral net Download PDFInfo
- Publication number
- CN103761172B CN103761172B CN201310738215.9A CN201310738215A CN103761172B CN 103761172 B CN103761172 B CN 103761172B CN 201310738215 A CN201310738215 A CN 201310738215A CN 103761172 B CN103761172 B CN 103761172B
- Authority
- CN
- China
- Prior art keywords
- unit
- fault
- symptom
- diagnosis
- diagnostic result
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 87
- 230000007935 neutral effect Effects 0.000 title claims abstract description 21
- 208000024891 symptom Diseases 0.000 claims abstract description 61
- 238000011084 recovery Methods 0.000 claims abstract description 18
- 238000013528 artificial neural network Methods 0.000 claims abstract description 13
- 238000000034 method Methods 0.000 claims abstract description 12
- 230000008569 process Effects 0.000 claims abstract description 8
- 230000001052 transient effect Effects 0.000 claims description 29
- 210000002569 neuron Anatomy 0.000 claims description 25
- 238000004364 calculation method Methods 0.000 claims description 5
- 235000013399 edible fruits Nutrition 0.000 claims 1
- 210000005036 nerve Anatomy 0.000 claims 1
- 230000001537 neural effect Effects 0.000 claims 1
- 238000012360 testing method Methods 0.000 description 15
- 238000002474 experimental method Methods 0.000 description 6
- 239000000243 solution Substances 0.000 description 6
- 238000001514 detection method Methods 0.000 description 5
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 5
- 230000032683 aging Effects 0.000 description 4
- 208000013404 behavioral symptom Diseases 0.000 description 4
- 230000007246 mechanism Effects 0.000 description 4
- 230000000737 periodic effect Effects 0.000 description 4
- 238000013461 design Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000002405 diagnostic procedure Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005530 etching Methods 0.000 description 2
- 238000002347 injection Methods 0.000 description 2
- 239000007924 injection Substances 0.000 description 2
- 230000007257 malfunction Effects 0.000 description 2
- 238000013508 migration Methods 0.000 description 2
- 230000005012 migration Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000007420 reactivation Effects 0.000 description 2
- 239000004065 semiconductor Substances 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 230000001351 cycling effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000002427 irreversible effect Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
Landscapes
- Test And Diagnosis Of Digital Computers (AREA)
Abstract
Hardware fault diagnosis system based on neutral net, belongs to hardware fault diagnosis field.Solve the problem that the accuracy rate of diagnosis of existing online hardware fault diagnosis system is low, it is provided with enumerator in symptom collector unit of the present invention, this symptom collector unit is for collecting the high-level symptom that fault propagation process China and foreign countries are aobvious, by enumerator during symptom thread re-executes, persistently symptom triggering times is added up, and calculate arrival rate, afterwards, arrival rate being delivered to Neural Network Diagnosis unit diagnose, arrival rate is symptom information；Neural Network Diagnosis unit is classified for the symptom information sending symptom collector unit here, and export diagnostic result to arbitration unit, arbitration unit is for collecting diagnostic result, and arbitration unit is for carrying out examination to invalid result, fault recovery unit is for after receiving diagnostic result, according to the process to trouble unit of the described diagnostic result, it is achieved fault recovery.The present invention applies in hardware fault diagnosis field.
Description
Technical field
The invention belongs to hardware fault diagnosis field.
Background technology
Compared with transient fault and permanent fault, the research and development of processor intermittent fault dependent diagnostic method is the slowest.Main reason is that, when semiconductor fabrication process rises, characteristic size is relatively big, and processor device (such as transistor) life-span is affected the most not notable by characteristic size, and therefore intermittent fault correlational study is still in high-level or Theoretical Proof stage.1975, document [89] demonstrated the system architecture impact on intermittent fault diagnosticability.1992, intermittent fault diagnostic method etc. under a kind of multiple processor structure of document [90] proposition.But, along with scrupulously abiding by Moore's Law, manufacturing process is radical, and characteristic size steps into nanometer era, the approaching atomic limit of gate oxide thickness.Within 2002, Intel 90 nanometer technology gate oxide thickness is 1.2 nanometer thickness, just corresponds to five stacked thickness of atom.In this context, semiconductor equipment problem of aging is on the rise, and intermittent fault takes place frequently, and has threatened the processor vital stage, and intermittent fault diagnosis mechanism demand is the most urgent.
M.L.Li etc., based on retrying capable thought, propose processor permanent fault diagnosis mechanism SWAT(SoftWare Anomely Treatment in high-level).This mechanism hypotheses application scenarios is multinuclear, and at least a core is reliable.When in processor, certain core breaks down, SWAT thread on which retries row, failure vanishes, then explanation core generation transient fault；Otherwise, thread context migrating to fault-free core and re-executes, if it is correct to perform result, then this core is diagnosed as permanent fault；In addition to above-mentioned two situations, then it is assumed that be software fault.It is considered that SWAT diagnosis policy can get a promotion from following several respects.First, SWAT diagnosis granularity is core level (Core level).Although there being multiple hardware core available in multinuclear, but for more fine-grained parts (such as functional unit), core remains more valuable calculating resource.Component-level diagnosis (Unit level) at least bring of both advantage:
(1) for multicore architecture, if component-level diagnostic result can be given, it is possible to temporary close trouble unit, to keep the availability of other resources in thread.Such as, for UltraSPARC T2, in each core, there are 8 rigid line journeys and 4 streamlines.Wherein there are 2 integer flowing water, 1 floating-point flowing water, 1 memory access flowing water.If the ALU diagnosed in 1 integer flowing water breaks down, can only close this flowing water, this guarantees in fault keranel 8 threads and continue correctly to perform integer, floating-point and accessing operation.
(2) for promote polycaryon processor flow success rate, design time usually to core in critical component implement redundancy strategy.During flow, once certain parts occurs that processing was lost efficacy (Manufacturing defect), these parts can close or use the technology such as E-Fuse to ensure the availability of monoblock chip.This provides the resource pool of " natural " the most just to component-level diagnosis, and redundant component is used the ideal effect reaching not reduce hardware performance when there is component-level fault.
Summary of the invention
The present invention is the problem low in order to solve the accuracy rate of diagnosis of existing online hardware fault diagnosis system, the invention provides a kind of hardware fault diagnosis system based on neutral net.
Hardware fault diagnosis system based on neutral net, it includes symptom collector unit, Neural Network Diagnosis unit, arbitration unit and fault recovery unit；
It is provided with enumerator in described symptom collector unit, this symptom collector unit is for collecting the high-level symptom that fault propagation process China and foreign countries are aobvious, by enumerator during symptom thread re-executes, persistently symptom triggering times is added up, and calculate arrival rate, afterwards, arrival rate being delivered to Neural Network Diagnosis unit and diagnoses, described arrival rate is symptom information；
Neural Network Diagnosis unit is classified for the symptom information of sending symptom collector unit here, and exports diagnostic result to arbitration unit,
Arbitration unit is for collecting diagnostic result, the diagnostic result collected includes invalid result and valid result, and arbitration unit is for carrying out examination to invalid result, if examination goes out invalid result, then thread symptom occur is re-executed, until there is valid result by notice symptom collector unit, and this valid result is sent to fault recovery unit, find invalid result if being unscreened, then directly the diagnostic result collected is sent to fault recovery unit
Fault recovery unit is for after receiving diagnostic result, according to the process to trouble unit of the described diagnostic result, it is achieved fault recovery.
Diagnostic result includes that fault model and trouble unit two parts, only two parts are all correct, is only correct diagnosis, is otherwise diagnostic error.Experiment obtains some groups of locally optimal solutions by off-line training；Afterwards, utilize locally optimal solution to carry out global diagnosis, ranked draw globally optimal solution；Finally, accuracy rate of diagnosis is drawn.It is assumed that each time etching system in only one of which component malfunction, and fault model determines.In experimentation, altogether 3 trouble units, 3 class fault models are completed the direct fault location of 10 Mibench test benchmarks, each fault model, trouble unit and test benchmark, carries out 300 fault injection experiment；Every test benchmark carries out altogether 43200 direct fault location, wherein, intermittent fault 36000 times (300 * 3 functional modules * 4 break out length * 10 test benchmark), transient fault 3600 times, permanent fault 3600 times；Experimental result and data referring specifically to Fig. 2 to shown in 7,
Experiment shows: the accuracy rate of diagnosis of transient fault, intermittent fault and permanent fault has respectively reached 84.4%, 95.7% and 96.6%, this demonstrate that the effectiveness of Neural Network Diagnosis Method based on high-level symptom；
As Fig. 2 to 7 gives hardware fault diagnosis system based on neutral net of the present invention to fault model and the diagnostic result of trouble location, and Fig. 2 to Fig. 7 all uses the typical test benchmark in mibench, including: basicmath, dijkstra, FFT, qsort and stringsearch, in addition to transient fault, in addition to transient fault, the accuracy rate of diagnosis of permanent fault and intermittent fault has exceeded 95%, and accuracy rate of diagnosis Fast Convergent, most of diagnosis examples are restrained within 10 times.This explanation: for just entering the intermittent fault parts of aging period and just having shown the parts of permanent fault characteristic to carry out fault diagnosis largely effective.For calculating, to patrol ponent design accuracy rate be 0 to Dijkstra and FFT, and through being analyzed data, reason is by intermittent fault that transient fault mistaken diagnosis is outburst a length of 2.It is the highest and cause that this results in the discrimination of behavioral symptoms mainly due to both outburst length is sufficiently close to (transient fault outburst a length of 1).
Under normal circumstances, after it is desirable to obtain capturing symptom, the accuracy rate of system the first diagnosis.Table 5-4 give accuracy rate of diagnosis.
Table 5-4 accuracy rate of diagnosis
For three kinds of fault models, the diagnosis of intermittent fault and permanent fault is higher than transient fault.The above two arrive more than 99% for the accuracy rate of parts, for the accuracy rate of diagnosis of model also more than 97.86%.The first cycle accuracy rate of transient fault is compared relatively low, especially fault model, and only 84.24%, main reason is that the fault model diagnosis that calculation is patrolled parts only has 56.97%.This parts transient fault is similar with the behavioral symptoms of intermittent fault outburst a length of 2, causes both feature samples to mingle in sample space, it is difficult to distinguish.It will be noted that the outburst a length of 1 of transient fault, from the beginning of second interval between diagnosis, fault not reactivation (for single bit upset fault), symptom will be wholly absent, and for interval and permanent fault, symptom will produce repeatedly or persistently.Therefore, transient fault symptom can be significantly hotter than first periodic diagnostics accuracy rate through the result that the multicycle diagnoses.In a word, hardware fault diagnosis system head periodic diagnostics accuracy rate based on neutral net of the present invention is 99.11% (trouble unit) and 95.75% (fault model), it was demonstrated that the effectiveness of this system diagnostics.
Delay in Diagnosis, within 1K bar instructs, postpones to allow lightweight hardware check point to recover fault.
The hardware fault diagnosis system based on neutral net that the present invention provides is possible not only to realize three kinds of fault model transient faults, permanent fault and the Precise Diagnosis of intermittent fault, and typical random logic device in processor can be realized, including: the component-level diagnosis that unit is carried out is patrolled in scalar/vector, decoding unit and calculation.
Present system uses the mode of capture symptom to diagnose, say, that only in system, faulty symptom produces, and hardware fault diagnosis system based on neutral net of the present invention just can start.And after interval between diagnosis starts, this system carries out the diagnosis of (one or multicycle until transference cure) concentration.Therefore, it can be said that be the mode of a kind of passive detection, active diagnosing.In this, otherwise varied with periodicity test class detection technique.Periodically test carries out fault detect by appointed interval time testing results code, say, that in the case of not having fault to occur, and test code is still to cycling service, is the mode of a kind of active detecting, active diagnosing.
From the point of view of in terms of three below, passive detection, the mode of active diagnosing have the advantage that
1. intermittent fault has aperiodicity and non-repeated.Fault propagation causes symptom, the mode more specific aim of detected method capture.
2. intermittent fault is caused by device aging, generally causes after system is on active service certain time.Fault unactivated very long waiting period, periodically test can increase system burden.Therefore, passive detection mode should be used.
3. intermittent fault and permanent fault belong to irreversible physical damnification, are intolerable, once occur should diagnosing as early as possible, replacing defective device.Accordingly, it would be desirable to use the mode of active diagnosing.
To sum up, on the premise of ensureing not reduce fault coverage, the mode of passive detection can effectively reduce systematic cost.
Accompanying drawing explanation
Fig. 1 is the principle schematic of hardware fault diagnosis system based on neutral net of the present invention；
Fig. 2 is in detailed description of the invention one, and when transient fault, diagnostic result is the accuracy rate of trouble unit；
Fig. 3 is in detailed description of the invention one, and when transient fault, diagnostic result is the accuracy rate of fault model；
Fig. 4 is in detailed description of the invention one, and when volatile fault, diagnostic result is the accuracy rate of trouble unit；
Fig. 5 is in detailed description of the invention one, and when volatile fault, diagnostic result is the accuracy rate of fault model；
Fig. 6 is in detailed description of the invention one, and when permanent fault, diagnostic result is the accuracy rate of trouble unit；
Fig. 7 is in detailed description of the invention one, and when permanent fault, diagnostic result is the accuracy rate of fault model.
Detailed description of the invention
Detailed description of the invention one: see Fig. 1 and present embodiment is described, the hardware fault diagnosis system based on neutral net described in present embodiment, it includes symptom collector unit 1, Neural Network Diagnosis unit 2, arbitration unit 3 and fault recovery unit 4；
It is provided with enumerator in described symptom collector unit 1, this symptom collector unit 1 is for collecting the high-level symptom that fault propagation process China and foreign countries are aobvious, by enumerator during symptom thread re-executes, persistently symptom triggering times is added up, and calculate arrival rate, afterwards, arrival rate being delivered to Neural Network Diagnosis unit 2 and diagnoses, described arrival rate is symptom information；
Neural Network Diagnosis unit 2 is classified for the symptom information sending symptom collector unit 1 here, and exports diagnostic result to arbitration unit 3,
Arbitration unit 3 is for collecting diagnostic result, the diagnostic result collected includes invalid result and valid result, and arbitration unit 3 is for carrying out examination to invalid result, if examination goes out invalid result, then thread symptom occur is re-executed, until there is valid result by notice symptom collector unit 1, and this valid result is sent to fault recovery unit 4, find invalid result if being unscreened, then directly the diagnostic result collected is sent to fault recovery unit 4
Fault recovery unit 4 is for after receiving diagnostic result, according to the process to trouble unit of the described diagnostic result, it is achieved fault recovery.
In present embodiment, diagnostic result includes fault model and trouble unit two parts.Only two parts are all correct, are only correct diagnosis, are otherwise diagnostic error.Experiment obtains some groups of locally optimal solutions by off-line training；Afterwards, utilize locally optimal solution to carry out global diagnosis, ranked draw globally optimal solution；Finally, accuracy rate of diagnosis is drawn.It is assumed that each time etching system in only one of which component malfunction, and fault model determines.In experimentation, altogether 3 trouble units, 3 class fault models are completed the direct fault location of 10 Mibench test benchmarks, each fault model, trouble unit and test benchmark, carries out 300 fault injection experiment；Every test benchmark carries out altogether 43200 direct fault location, wherein, intermittent fault 36000 times (300 * 3 functional modules * 4 break out length * 10 test benchmark), transient fault 3600 times, permanent fault 3600 times；Experimental result and data referring specifically to shown in Fig. 2-7,
Experiment shows: the accuracy rate of diagnosis of transient fault, intermittent fault and permanent fault has respectively reached 84.4%, 95.7% and 96.6%, this demonstrate that the effectiveness of Neural Network Diagnosis Method based on high-level symptom；
As Fig. 2 to 7 gives hardware fault diagnosis system based on neutral net of the present invention to fault model and the diagnostic result of trouble location, and Fig. 2 to Fig. 7 all uses the typical test benchmark in mibench, including: basicmath, dijkstra, FFT, qsort and stringsearch, in addition to transient fault, the accuracy rate of diagnosis of permanent fault and intermittent fault has exceeded 95%, and accuracy rate of diagnosis Fast Convergent, and most of diagnosis examples are restrained within 10 times.This explanation: for just entering the intermittent fault parts of aging period and just having shown the parts of permanent fault characteristic to carry out fault diagnosis largely effective.For calculating, to patrol ponent design accuracy rate be 0 to Dijkstra and FFT, and through being analyzed data, reason is by intermittent fault that transient fault mistaken diagnosis is outburst a length of 2.It is the highest and cause that this results in the discrimination of behavioral symptoms mainly due to both outburst length is sufficiently close to (transient fault outburst a length of 1).
Under normal circumstances, after it is desirable to obtain capturing symptom, the accuracy rate of system the first diagnosis.Table 5-4 give accuracy rate of diagnosis.
Table 5-4 accuracy rate of diagnosis
For three kinds of fault models, the diagnosis of intermittent fault and permanent fault is higher than transient fault.The above two arrive more than 99% for the accuracy rate of parts, for the accuracy rate of diagnosis of model also more than 97.86%.The first cycle accuracy rate of transient fault is compared relatively low, especially fault model, and only 84.24%, main reason is that the fault model diagnosis that calculation is patrolled parts only has 56.97%.This parts transient fault is similar with the behavioral symptoms of intermittent fault outburst a length of 2, causes both feature samples to mingle in sample space, it is difficult to distinguish.It will be noted that the outburst a length of 1 of transient fault, from the beginning of second interval between diagnosis, fault not reactivation (for single bit upset fault), symptom will be wholly absent, and for interval and permanent fault, symptom will produce repeatedly or persistently.Therefore, transient fault symptom can be significantly hotter than first periodic diagnostics accuracy rate through the result that the multicycle diagnoses.In a word, hardware fault diagnosis system head periodic diagnostics accuracy rate based on neutral net of the present invention is 99.11% (trouble unit) and 95.75% (fault model), it was demonstrated that the effectiveness of this system diagnostics.
Detailed description of the invention two: see Fig. 1 and present embodiment is described, present embodiment is with the difference of the hardware fault diagnosis system based on neutral net described in detailed description of the invention one, described Neural Network Diagnosis unit 2 is made up of three layers of neuron, it is respectively input layer, hidden neuron and output layer neuron, symptom information is sent to input layer as input information, symptom is classified by the calculating through hidden neuron, after classification results is processed by output layer neuron, obtain diagnostic result, and this diagnostic result is sent to arbitration unit 3.
Detailed description of the invention three: see Fig. 1 and present embodiment is described, present embodiment is with the difference of the hardware fault diagnosis system based on neutral net described in detailed description of the invention two, described output layer neuron includes multiple neuron, each neuron produces Boolean True or False, True and represents that diagnostic result is the fault type or parts that this neuron is corresponding；False represents the fault type or parts that non-neuron is corresponding.
Detailed description of the invention four: see Fig. 1 and present embodiment is described, present embodiment is with the difference of the hardware fault diagnosis system based on neutral net described in detailed description of the invention one, described output layer neuron includes 6 neurons, wherein 3 is hardware fault model, other 3 is trouble unit, 3 hardware fault models are respectively transient fault, intermittent fault and permanent fault, and unit is patrolled in 3 trouble unit respectively decoding units, scalar/vector and calculation.
In present embodiment, the processing procedure of trouble unit is, for transient fault by fault recovery unit 4, then only symptom thread need to be re-executed, for intermittent fault, then can use core migration mechanism, by symptom thread migration to reliable core after fault occurs, after transference cure, migrate back fault keranel again perform, for permanent fault, by disabling trouble unit, if any redundant component, or by part replacement, processed.
Detailed description of the invention five: seeing Fig. 1 and illustrate that present embodiment, present embodiment are with the difference of the hardware fault diagnosis system based on neutral net described in detailed description of the invention two, described input layer includes 8 neurons.
In present embodiment, input layer is provided with the amount of reach of corresponding 7 symptoms (including that 6 class keys are absorbed in and activity occurred frequently) of 8 neurons, and key is absorbed in the most nested labelling.
Claims (4)
1. hardware fault diagnosis system based on neutral net, it is characterised in that it includes symptom collector unit (1), nerve
Network diagnosis unit (2), arbitration unit (3) and fault recovery unit (4)；
Being provided with enumerator in described symptom collector unit (1), this symptom collector unit (1) is used for collecting fault propagation mistake
The high-level symptom that journey China and foreign countries are aobvious, by enumerator during symptom thread re-executes, persistently to symptom triggering times
Add up, and calculate arrival rate, afterwards, arrival rate is delivered to Neural Network Diagnosis unit (2) and diagnoses, described
Arrival rate is symptom information；
Neural Network Diagnosis unit (2) is classified for the symptom information sending symptom collector unit (1) here, and exports
Diagnostic result to arbitration unit (3),
Arbitration unit (3) is for collecting diagnostic result, and the diagnostic result collected includes invalid result and valid result,
And arbitration unit (3) is for carrying out examination to invalid result, if examination goes out invalid result, then notice symptom collector unit (1)
Thread symptom occur is re-executed, until valid result occurs, and this valid result is sent to fault recovery list
Unit (4), finds invalid result if being unscreened, then directly send the diagnostic result collected to fault recovery unit (4),
Fault recovery unit (4) is for after receiving diagnostic result, according to the process to trouble unit of the described diagnostic result, it is achieved
Fault recovery.
Hardware fault diagnosis system based on neutral net the most according to claim 1, it is characterised in that described god
Being made up of three layers of neuron through network diagnosis unit (2), respectively input layer, hidden neuron and output layer is neural
Unit, symptom information is sent to input layer as input information, and symptom is classified by the calculating through hidden neuron,
After classification results is processed by output layer neuron, it is thus achieved that diagnostic result, and this diagnostic result is sent to arbitration unit (3).
Hardware fault diagnosis system based on neutral net the most according to claim 2, it is characterised in that described is defeated
Going out a layer neuron and include multiple neuron, each neuron produces Boolean True or False, True and represents diagnosis knot
Fruit is the fault type or parts that this neuron is corresponding；False represents the fault type or parts that non-neuron is corresponding.
Hardware fault diagnosis system based on neutral net the most according to claim 1, it is characterised in that described is defeated
Going out a layer neuron and include 6 neurons, wherein 3 is hardware fault model, and other 3 is trouble unit, 3 hardware
Fault model is respectively transient fault, intermittent fault and permanent fault, and 3 trouble units are respectively decoding unit, address life
Unit and calculation is become to patrol unit.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310738215.9A CN103761172B (en) | 2013-12-29 | 2013-12-29 | Hardware fault diagnosis system based on neutral net |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310738215.9A CN103761172B (en) | 2013-12-29 | 2013-12-29 | Hardware fault diagnosis system based on neutral net |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103761172A CN103761172A (en) | 2014-04-30 |
| CN103761172B true CN103761172B (en) | 2016-08-31 |
Family
ID=50528414
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310738215.9A Active CN103761172B (en) | 2013-12-29 | 2013-12-29 | Hardware fault diagnosis system based on neutral net |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103761172B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10380557B2 (en) * | 2015-07-31 | 2019-08-13 | Snap-On Incorporated | Methods and systems for clustering of repair orders based on alternative repair indicators |
| CN108875919B (en) * | 2017-07-21 | 2021-04-02 | 北京旷视科技有限公司 | Data processing apparatus, data processing method, and storage medium product |
| CN108109688A (en) * | 2017-12-18 | 2018-06-01 | 上海联影医疗科技有限公司 | A kind of imaging system and its scan method |
| CN108153618B (en) * | 2017-12-22 | 2021-12-03 | 国网浙江杭州市萧山区供电有限公司 | Hard disk data recovery method and device and hard disk data recovery equipment |
| KR20200138565A (en) | 2019-05-31 | 2020-12-10 | 삼성전자주식회사 | Method and apparatus for managing a plurality of remote radio heads in a communication network |
| CN112965854B (en) * | 2021-04-16 | 2022-04-29 | 吉林大学 | Method, system and equipment for improving reliability of convolutional neural network |
| CN113568816B (en) * | 2021-09-22 | 2022-01-25 | 国汽智控(北京)科技有限公司 | Process monitoring method, device and equipment |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5383192A (en) * | 1992-12-23 | 1995-01-17 | Intel Corporation | Minimizing the likelihood of slip between the instant a candidate for a break event is generated and the instant a microprocessor is instructed to perform a break, without missing breakpoints |
| CN1149735A (en) * | 1994-05-25 | 1997-05-14 | 西门子公司 | Service personal computer of modular structure |
-
2013
- 2013-12-29 CN CN201310738215.9A patent/CN103761172B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5383192A (en) * | 1992-12-23 | 1995-01-17 | Intel Corporation | Minimizing the likelihood of slip between the instant a candidate for a break event is generated and the instant a microprocessor is instructed to perform a break, without missing breakpoints |
| CN1149735A (en) * | 1994-05-25 | 1997-05-14 | 西门子公司 | Service personal computer of modular structure |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103761172A (en) | 2014-04-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103761172B (en) | Hardware fault diagnosis system based on neutral net | |
| Gizopoulos et al. | Architectures for online error detection and recovery in multicore processors | |
| CN110955571B (en) | Fault management system for functional safety of vehicle-specification-level chip | |
| CN103218272A (en) | Spaceborne digital signal processor turning reinforcing method | |
| Collet et al. | Chip self-organization and fault tolerance in massively defective multicore arrays | |
| CN102033789A (en) | Reliability analysis method for embedded safety-critical system | |
| CN1929034A (en) | Method and system for RAM fault testing | |
| CN105320579A (en) | Self-repairing dual-redundancy assembly line oriented to SPARC V8 processor and fault-tolerant method | |
| Hanif et al. | Dependable deep learning: Towards cost-efficient resilience of deep neural network accelerators against soft errors and permanent faults | |
| Dang et al. | TSV-OCT: A scalable online multiple-TSV defects localization for real-time 3-D-IC systems | |
| US10303566B2 (en) | Apparatus and method for checking output data during redundant execution of instructions | |
| CN102591763B (en) | System and method for detecting faults of integral processor on basis of determinacy replay | |
| Li et al. | Exploiting soft computing for increased fault tolerance | |
| CN105320575B (en) | A kind of self checking of duplication redundancy streamline and recovery device and method | |
| Dang et al. | TSV-IaS: Analytic analysis and low-cost non-preemptive on-line detection and correction method for TSV defects | |
| Vera et al. | Selective replication: A lightweight technique for soft errors | |
| Dadashi et al. | Hardware-software integrated diagnosis for intermittent hardware faults | |
| CN101706746A (en) | Device and method for carrying out online debugging on memory interface circuit | |
| CN105260272B (en) | A kind of synchronous error correction Pipeline control structure and its method | |
| CN103514071B (en) | Nondestructive internal storage online testing method | |
| CN102750194A (en) | Large-scale integrated circuit level error recording and responding method | |
| CN104050051A (en) | Fault diagnosis method for on-board computer | |
| CN108762999A (en) | A kind of kernel failure collection method and device | |
| Lee et al. | Hardware-based online self-diagnosis for faulty device identification in large-scale IoT systems | |
| CN206833419U (en) | Dual redundant streamline based on checkpoint technology |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20190613 Address after: 150000 Heilongjiang Harbin Dalian economic and Trade Zone, the North Road and Xingkai Road intersection Patentee after: HIT ROBOT GROUP Co.,Ltd. Address before: 150001 No. 92 West straight street, Nangang District, Heilongjiang, Harbin Patentee before: Harbin Institute of Technology |
|
| PP01 | Preservation of patent right | ||
| PP01 | Preservation of patent right |
Effective date of registration: 20240626 Granted publication date: 20160831 |