A functional safety control system and fault diagnosis decision system of an electric fork truck

Abstract

The application discloses a functional safety control system and a fault diagnosis decision system of an electric forklift, belongs to the technical field of safety control systems of electric forklifts, and comprises safety function modules of a driving control subsystem, an operating position monitoring subsystem, an unexpected acceleration subsystem during step starting, and an unexpected behavior subsystem of vehicle driving, wherein the safety function modules execute safety control logic of each subsystem.The application has the advantages that the functional safety control system and the fault diagnosis decision system applied to the electric forklift are built, and the problems of safety function protection, fault tracing after function protection, and fault decision processing and repair of the electric forklift in the monitoring of a driving control state, the monitoring of an operating position, the unexpected behavior monitoring of the electric forklift during step starting on a flat road, and the vehicle driving are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of functional safety control system of electric forklift, in particular to a functional safety control system and a fault diagnosis decision system of electric forklift. BACKGROUND

[0002] With the enhancement of global environmental protection consciousness and the requirement of reducing carbon emissions in various countries, electric forklifts have become an ideal substitute for traditional internal combustion forklifts due to their zero emission and low noise characteristics, and have lower maintenance costs and energy consumption. The application scenarios of electric forklifts have expanded from traditional warehouse handling to more industries, such as ports, automobiles, and steel industries. As the demand for overseas export business of electric forklifts gradually increases, the functional safety compliance requirements of overseas markets for electric forklifts are becoming increasingly high.

[0003] The technical progress in the aspects of electric control systems, fault diagnosis systems, and fault post-processing decision systems makes it urgent not only to develop functional safety of electric forklifts but also to meet the functional requirements of electric forklifts in terms of fault model building and fault post-processing decision. SUMMARY

[0004] The present application aims to solve the technical problems mentioned in the background. The first aspect provides a functional safety control system of electric forklift, which facilitates the detection of the safety of each system of the electric forklift.

[0005] The second aspect provides a fault diagnosis decision system, which is used in combination with the functional safety control system of electric forklift provided in the first aspect, uses the detection data to calculate the most possible fault cause of the electric forklift, and facilitates the fault recovery by authorized technical personnel.

[0006] To achieve the above purpose, the first aspect of the present application provides a functional safety control system of electric forklift, which includes safety function modules of a drive control subsystem, an operating position monitoring subsystem, a step-by-step accidental acceleration subsystem, and a vehicle driving accidental behavior subsystem. The safety function modules execute safety control logic of each subsystem.

[0007] In some embodiments, the safety control logic of each subsystem is as follows: after the whole vehicle is powered on and initialized, the safety function module monitors the signal state of each component of the drive control subsystem, the operating position monitoring subsystem, the step-by-step accidental acceleration subsystem, and the vehicle driving accidental behavior subsystem. The detection period of each component signal of the safety function module monitoring subsystem is less than or equal to 50 ms. If the defined abnormal trigger condition is triggered, the safety function protection state of the corresponding subsystem module is entered. At this time, the main positive contactor of the BMS is disconnected, and the fault recovery needs to be performed by authorized technical personnel.

[0008] If no abnormality is monitored or the technician has recovered the fault, the monitoring of the signals of the components of the drive control subsystem, the operating position monitoring subsystem, the start-up unexpected acceleration subsystem, and the vehicle travel unexpected behavior subsystem continues.

[0009] In some embodiments, the safety function module monitors the signals of the components of the drive control subsystem, the operating position monitoring subsystem, the start-up unexpected acceleration subsystem, and the vehicle travel unexpected behavior subsystem includes the combination handle, the accelerator pedal, the brake pedal, the seat switch, the hand brake switch, the key switch, and the emergency stop switch, and determines whether to enter the corresponding subsystem safety function protection state according to the set monitoring signal threshold.

[0010] In some embodiments, when the vehicle starts, the events that need to be detected by the drive control subsystem are:

[0011] Trigger event 1: the forward gear switch is TRUE;

[0012] Trigger event 2: the reverse gear switch is TRUE;

[0013] Trigger event 3: the accelerator pedal has an output;

[0014] Fault detection trigger event: sensor fault, controller fault, and actuator fault.

[0015] In some embodiments, the trigger events that need to be detected by the operating position monitoring subsystem are:

[0016] Trigger event 1: the accelerator pedal has an output all the time, and the drive motor does not slow down normally after the operator leaves the seat for 2s;

[0017] Trigger event 2: the accelerator pedal has no output, and the drive motor does not slow down normally after the operator leaves the seat for 0.2s;

[0018] Trigger event 3: there is no reset action during the deceleration process, and the drive motor deceleration state changes;

[0019] Trigger event 4: there is no reset action after being stationary, and the drive motor stationary state changes.

[0020] Trigger event 5: fault detection trigger event, including sensor fault, controller fault, and actuator fault.

[0021] In some embodiments, the trigger events that need to be detected by the start-up unexpected acceleration subsystem are:

[0022] Trigger event 1: the travel direction and the gear do not match;

[0023] Trigger event 2: there is no speed request, and the acceleration exceeds the limit;

[0024] Trigger event 3: fault detection trigger event, the fault detection trigger event includes sensor failure, controller failure, actuator failure.

[0025] In some embodiments, the trigger events that need to be detected by the vehicle driving unexpected behavior subsystem are:

[0026] Trigger event 1: driving direction response trigger condition:

[0027] When in forward gear, the motor speed is negative, and not decelerating;

[0028] When in reverse gear, the motor speed is positive, and not decelerating;

[0029] When in neutral gear, the motor speed is unstable, and not decelerating;

[0030] Trigger event 2: unexpected acceleration trigger event, i.e. the absolute value of motor acceleration is greater than 1.3 times the maximum acceleration, and not in deceleration state;

[0031] Trigger event 3: unexpected speed trigger event, i.e. the absolute value of motor speed is greater than 1.2 times the expected speed absolute value, and not in deceleration state;

[0032] Trigger event 4: fault detection trigger event, the fault detection trigger event includes sensor failure, controller failure, actuator failure.

[0033] The second aspect of the application provides a fault diagnosis decision system, which is used based on the functional safety control system of the electric fork truck provided in the first aspect, collects historical fault mode data samples of the drive control subsystem, the operating position monitoring subsystem, the unexpected acceleration during starting subsystem, and the vehicle driving unexpected behavior subsystem, uses a Bayesian greedy algorithm to train the model and establish a fault diagnosis backtracking reasoning model for each subsystem;

[0034] According to the potential causes corresponding to the fault probability set inferred from the fault diagnosis model, the maximum likelihood estimation algorithm is used to arrange the potential fault causes in probability, the causal relationship between the historical fault data is established, and the model is represented by a directed acyclic graph;

[0035] And according to the ordering value of the probability value, the potential causes triggering the safety state protection are recommended to the authorized technical personnel one by one, the authorized technical personnel handle and repair the fault causes according to the probability ordering of the diagnosis model after the recommendation, and after the repair, the normal operation is restored and the safety state protection is removed.

[0036] Further, the method for establishing a fault diagnosis backtracking reasoning model for each subsystem includes:

[0037] S1, construct a fault set for each functional safety subsystem, respectively use 、 、 、 represents:

[0038] ; ;

[0039] ; ;

[0040] in the set, 、 b…is the safety function trigger event of subsystem 1, …is the safety function trigger event of subsystem 2 …is the safety function trigger event of subsystem 3, …is the safety function trigger event of subsystem 4.

[0041] S2, collect and summarize the reasons for triggering each functional safety subsystem fault into a safe state protection, represented by the set :

[0042] 、 、 ;

[0043] in the set, 、 …is the potential reason for triggering the safety function trigger event of safety subsystem 1; 、 …is the potential reason for triggering the safety function trigger event of safety subsystem 2; …is the potential reason for triggering the safety function trigger event of safety subsystem 3; 、 …is the potential reason for triggering the safety function trigger event of safety subsystem 4. The set of reasons for safety state protection can be expanded at any time.

[0044] S3, use the Bayesian greedy learning algorithm to train the collected and quantified historical fault reason data sample set of each functional safety subsystem, and can respectively obtain the Bayesian fault diagnosis model:

[0045] ;

[0046] In the formula: represents the directed acyclic graph structure DAG of the Bayesian fault diagnosis model; represents the conditional probability table of the variable;

[0047] The trained Bayesian fault diagnosis model is evaluated by using BIC score function, the optimal Bayesian fault diagnosis model obtained by training is saved, and the BIC score function is as follows:

[0048] ;

[0049] ;

[0050] In the formula: is the maximum likelihood value of the model on the given data, that is, the likelihood function value after parameter estimation of the model, is the number of free parameters in the model, that is, the complexity of the model, represents the network structure dimension of the Bayesian diagnosis model;

[0051] The model with the minimum BIC value is selected as the final functional safety subsystem Bayesian fault diagnosis model.

[0052] Further, according to the fault probability set corresponding to the potential causes obtained by reasoning of the fault diagnosis model, the method for probabilistic arrangement of potential fault causes includes:

[0053] When the safety function of the subsystem is triggered, the current subsystem Bayesian fault diagnosis model is cut in The maximum likelihood estimation algorithm is used to diagnose and reason the potential causes of Each fault cause set enters the safety state protection;

[0054] S1, first, according to independent observation data, a likelihood function is constructed:

[0055] ;

[0056] In the formula: is the parameter to be estimated, is the observation sample, is the likelihood function;

[0057] S2, take the logarithm of the likelihood function to obtain the log-likelihood function:

[0058] ;

[0059] S3, according to the parameter value function target requirement of the maximum log-likelihood, the formula is as follows:

[0060] ;

[0061] Regarding , the derivative is taken and the derivative is set to 0:

[0062] ​ ;

[0063] S4, the parameter can be obtained by solving the equation The estimated value , thereby deducing the fault cause set The corresponding probability value is represented as ;

[0064] ;

[0065] ;

[0066] ;

[0067] ;

[0068] S5, after obtaining the probability of the potential fault cause by using the fault diagnosis model, enter the fault post-processing decision module, and the probability value of each fault cause triggering the operation position monitoring into the safety state protection is sorted from large to small;

[0069] And according to the sorting value of the probability value, the potential causes triggering the safety state protection are recommended to the authorized technical personnel one by one, and the authorized technical personnel processes and repairs the fault cause according to the probability sorting of the fault cause recommended by the diagnosis model.

[0070] Compared with the prior art, the application has the advantages that a functional safety control system and a fault diagnosis decision system applied to an electric forklift are built, and the problems of safety function protection, fault tracing after function protection, fault decision processing and repair of the electric forklift in the driving control state monitoring, operation position monitoring, accidental behavior monitoring of the vehicle during driving and the like are solved.

[0071] Additional aspects and advantages of the application will be in part apparent and in part pointed out hereinafter. BRIEF DESCRIPTION OF DRAWINGS

[0072] Figure 1 is a schematic diagram of the overall logic flow of the functional safety control system and the fault diagnosis decision system of the electric forklift according to the embodiment of the application;

[0073] Figure 2 is a schematic diagram of the overall flow framework of the fault diagnosis decision of the safety control module of each subsystem of the electric forklift according to the embodiment of the application;

[0074] Figure 3 is a schematic diagram of the fault diagnosis decision flow of the driving control subsystem shown in Figure 2 ;

[0075] Figure 4is Figure 2 a flowchart of a fault diagnosis decision-making process performed by the operation position monitoring subsystem shown in

[0076] Figure 5 is Figure 2 a flowchart of a fault diagnosis decision-making process performed by the unexpected acceleration during flat road start subsystem shown in

[0077] Figure 6 is Figure 2 a flowchart of a fault diagnosis decision-making process performed by the vehicle driving unexpected behavior subsystem shown in DETAILED DESCRIPTION

[0078] The present application will be further described in detail.

[0079] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments will be described clearly and completely below with reference to the drawings in the embodiments of the present application. The following embodiments are used to explain the present application but not to limit the scope of the present application.

[0080] In combination with Figure 1 , Figure 2 The present embodiment is described in combination with the use of the electric forklift functional safety control system and the fault diagnosis decision-making system. After the electric forklift functional safety control system triggers the safety function, the Bayesian fault diagnosis is performed, and after the fault tracing, the fault handling decision is made. The electric forklift functional safety control system is further divided into the driving control subsystem, the operation position monitoring subsystem, the unexpected acceleration during flat road start subsystem, and the vehicle driving unexpected behavior subsystem.

[0081] After the whole vehicle system is powered on and initialized, the key input signals such as the combination handle, the accelerator pedal, the brake pedal, the seat switch, the hand brake switch, the key switch, and the emergency stop switch are collected in real time, and the safety function modules of the subsystems built are started to monitor the safety state of the whole vehicle. If it is detected that one of the signals is abnormal and causes the safety function to be triggered, the safety state protection is performed, and after entering the state protection, the Bayesian fault diagnosis model of the subsystem built is entered to perform the reverse diagnosis reasoning on the possible fault causes of the triggered safety function. The Bayesian maximum likelihood estimation (MLE) is used to trace a set of potential fault causes of the fault caused by the triggered safety function, and the fault handling decision module is used to analyze the decision of the set of fault causes after the reverse reasoning, so as to finally obtain the most possible fault and recommend the most possible fault to the authorized technical personnel for the handling and repair of the fault and the reset of the safety function. Meanwhile, it is monitored whether the authorized technical personnel has completed the fault recovery, and if the authorized technical personnel has completed the fault recovery, the safety protection state is released and the signal monitoring module is entered.

[0082] In combination withFigure 3 The drive control subsystem safety function logic and implementation process are shown as follows:

[0083] (1) Safety function logic design:

[0084] The safety controller running cycle is set to be less than or equal to 50 ms. If the safety function trigger condition is detected when the vehicle starts, the controller will enter the safety state, and the main positive contactor of the BMS will be disconnected.

[0085] (2) Safety function trigger condition:

[0086] The events that need to be detected when the vehicle starts are shown as follows:

[0087] a: Trigger event 1, forward gear switch is TRUE;

[0088] b: Trigger event 2, reverse gear switch is TRUE;

[0089] c: Trigger event 3, accelerator pedal has output;

[0090] d: Fault detection trigger event: sensor fault, controller fault, actuator fault.

[0091] If one of the event triggers is detected when the vehicle starts, it is judged that the drive control system safety function is triggered. To prevent the forklift from suddenly moving and causing harm to the operator and surrounding personnel, leading to a dangerous situation, the safety state should be entered:

[0092] (3) Design of travel control subsystem fault diagnosis model:

[0093] If the safety function setting event is triggered, after entering the safety state, the safety control function needs to be enabled to perform action protection, and further diagnosis of the cause of the fault is required. In this embodiment, the historical fault data of each working device (i.e., each component) is trained using Bayesian to obtain a fault diagnosis model, and the cause of the fault is obtained through reverse reasoning to solve the fault.

[0094] S1, construct a travel control subsystem fault set, denoted as F:

[0095] ;

[0096] In the set F, is the safety function trigger event 1, is the safety function trigger event 2, is the safety function trigger event 3, is the fault detection trigger event.

[0097] ​S2, collect and aggregate the reasons for entering the safe state protection of the driving control subsystem failure, using a set represents:

[0098] ;

[0099] in the set, is the potential reason for triggering event 1, the forward gear switch is TRUE, is the potential reason for triggering event 2, the reverse gear switch is TRUE, is the potential reason for triggering event 3, the accelerator pedal has output, is the potential reason for related sensor failure, controller failure, actuator failure. The reason set for entering the safe state protection can be expanded at any time.

[0100] S3, train the collected and quantized historical fault reason data sample set using the Bayesian greedy learning algorithm to obtain a Bayesian fault diagnosis model:

[0101] ;

[0102] In the formula: represents the Bayesian fault diagnosis model DAG (Directed Acyclic Graph); represents the conditional probability table of the variable.

[0103] The trained Bayesian fault diagnosis model is evaluated using the BIC (Bayesian Information Criteria) score function, and the optimal Bayesian fault diagnosis model obtained by training is saved. The BIC score function is as follows:

[0104] ;

[0105] ;

[0106] In the formula: is the maximum likelihood value of the model on the given data, that is, the likelihood function value after the model parameter estimation, is the number of free parameters in the model, that is, the complexity of the model, represents the network structure dimension of the Bayesian diagnosis model.

[0107] Since measures the fitting degree of the model to the data, is the penalty term of model complexity, the larger this part indicates the more complex the model. Therefore, the model with the minimum BIC value is selected as the final driving control subsystem Bayesian fault diagnosis model.

[0108] (4) Fault post-processing decision and repair module design

[0109] When the safety function is triggered, the driving control subsystem Bayesian fault diagnosis model is cut in , and the maximum likelihood estimation algorithm is used to diagnose and reason the potential causes of entering the safety state protection;

[0110] S1, first, according to the independent observation data, the likelihood function is constructed:

[0111] ;

[0112] In the formula: is the parameter to be estimated, is the observation sample, is the likelihood function.

[0113] S2, in order to simplify the calculation, take the logarithm of the likelihood function to get the log-likelihood function:

[0114] ;

[0115] S3, according to the parameter value function target requirement of maximum log-likelihood, the formula is as follows:

[0116] ;

[0117] Take the derivative of with respect to , and set the derivative to 0:

[0118]

[0119] S4, by solving the equation can get the estimated value of the parameter , thus the corresponding probability value of the fault cause set is obtained by reverse reasoning, denoted as :

[0120] ;

[0121] S5, after getting the probability of the potential fault cause using the diagnostic model, enter the fault post-processing decision module, and the probability value of each fault cause Sort from large to small, and according to the probability value of the order value, the potential causes of triggering the safety state protection are recommended to the authorized technicians one by one, and the authorized technicians handle and repair the fault according to the probability of the fault reason recommended by the diagnostic model.

[0122] (5) Safety state reset condition

[0123] Fault recovery:

[0124] When the safety function is triggered after the controller loop enters the safety state, the fault should be repaired by authorized technicians, and the authorized technicians handle and repair the fault according to the probability of the fault reason recommended by the diagnostic model to restore normal operation.

[0125] Combined Figure 4 The operating position monitoring safety function

[0126] (1) Safety function logic:

[0127] The operating position monitoring safety function related components include: seat switch, combination handle, accelerator pedal, safety controller running period is set to ≤50ms, if the start detects the safety function trigger condition, the controller will enter the safety state, and the main positive contactor of the BMS is disconnected.

[0128] (2) Safety function trigger event:

[0129] a: Trigger event 1 The accelerator pedal has output all the time and the operator leaves the seat for 2s, and the drive motor does not slow down normally;

[0130] b: Trigger event 2 The accelerator pedal has no output and the operator leaves the seat for 0.2s, and the drive motor does not slow down normally;

[0131] c: Trigger event 3 No reset action during deceleration, the drive motor deceleration state changes;

[0132] d: Trigger event 4 No reset action after being stationary, the drive motor stationary state changes.

[0133] e: Trigger event 5 Fault detection trigger event:

[0134] Trigger event 1 sensor failure;

[0135] Trigger event 2 controller failure;

[0136] Trigger event 3 actuator failure.

[0137] Trigger one of them, that is, judge the safety function trigger of the drive control system, to prevent the forklift from suddenly walking and causing damage to the operator and surrounding personnel, leading to danger, and should enter the safety state:

[0138] (3) Operation position monitoring fault diagnosis model design:

[0139] If the operation position monitoring safety function setting event is triggered, after entering the safety state, while enabling the safety control function to act as protection, further diagnosis of the cause of the failure is needed. The present embodiment uses Bayesian to train the historical fault data of the operation position state protection to obtain a fault diagnosis model, and reversely reason the cause of the failure to solve the failure.

[0140] S1, construct the operation position subsystem fault set, use to represent:

[0141] ;

[0142] In the set, is the safety function trigger event 1, is the safety function trigger event 2, is the safety function trigger event 3, is the safety function trigger event 4, is the fault detection trigger event.

[0143] S2, collect and summarize the reasons for triggering the operation position subsystem fault to enter the safety state protection, use the set to represent:

[0144] ;

[0145] In the set, is the potential reason for triggering event 1 that the accelerator pedal has output all the time and the operator leaves the seat for 2s, and the drive motor does not normally decelerate, is the potential reason for triggering event 2 that the accelerator pedal has no output and the operator leaves the seat for 0.2s, and the drive motor does not normally decelerate, is the potential reason for triggering event 3 that there is no reset action during deceleration, and the drive motor deceleration state changes, is the potential reason for triggering event 4 that there is no reset action after being stationary, and the drive motor stationary state changes, is the potential reason for the fault detection trigger event. The reason set for safety state protection can be expanded at any time.

[0146] S3, use Bayesian greedy learning algorithm to train the collected and quantized historical fault reason data sample set to obtain a Bayesian fault diagnosis model:

[0147] ;

[0148] In the formula: Directed Acyclic Graph (DAG) representing the Bayesian fault diagnosis model; a conditional probability table representing a variable.

[0149] The trained Bayesian fault diagnosis model is evaluated by using a BIC (Bayesian Information Criteria) score function, and the optimal Bayesian fault diagnosis model obtained by training is saved. The BIC score function is as follows:

[0150]

[0151]

[0152] In the formula: is the maximum likelihood value of the model on the given data, that is, the likelihood function value after the model parameter estimation, is the number of free parameters in the model, that is, the complexity of the model, represents the network structure dimension of the Bayesian diagnosis model.

[0153] Since measures the fitting degree of the model to the data, is a penalty term for the complexity of the model, and the larger this part is, the more complex the model is. Therefore, the model with the minimum BIC value is selected as the final operating position monitoring subsystem Bayesian fault diagnosis model.

[0154] (4) Fault post-processing decision and repair module design

[0155] When the safety function is triggered, the operating position monitoring subsystem Bayesian fault diagnosis model is cut in , and the maximum likelihood estimation algorithm is used to diagnose and reason the potential causes of entering the safety state protection;

[0156] S1, first, according to the independent observation data, the likelihood function is constructed:

[0157]

[0158] In the formula: is the parameter to be estimated, is the observation sample, is the likelihood function.

[0159] S2, in order to simplify the calculation, the likelihood function is taken as the logarithm, and the logarithm likelihood function is obtained:

[0160]

[0161] ​S3, according to the maximum likelihood of the parameter value function target requirements, can be obtained as follows:

[0162]

[0163] To About Derivation, and set the derivative to 0:

[0164]

[0165] S4, by solving the equation can be obtained by the parameter Estimate , so that the reverse reasoning to obtain the fault cause set Corresponding probability value, denoted by :

[0166]

[0167] S5, after using the fault diagnosis model to get the probability of potential failure causes, enter the fault post-processing decision module, and the probability value of each fault cause , , , , From large to small, and according to the probability value of the sorting value of the potential cause of triggering the safety state protection is recommended to the authorized technical personnel one by one, and the authorized technical personnel according to the probability of the fault cause after the diagnosis model recommendation to carry out fault processing and repair.

[0168] (5) safety state reset condition

[0169] Fault recovery:

[0170] When the safety function of the operation position monitoring is triggered to enter the state protection, the controller loop enters the safety state, which should be repaired by the authorized technical personnel, and the authorized technical personnel according to the probability of the fault cause after the diagnosis model recommendation to carry out fault processing and repair to restore normal operation.

[0171] Combined Figure 5 It is shown that the safety function logic and implementation process of the flat road starting accidental acceleration subsystem;

[0172] (1) flat road starting accidental acceleration subsystem safety function logic:

[0173] The flat road start-up accidental acceleration subsystem safety function is composed of a combination handle and a coded safety-related component, a controller running cycle is set to be less than or equal to 50 ms, and if the running direction is inconsistent with the direction of the gear after starting, the safety function is triggered; if the acceleration exceeds the limit without a speed request, the safety function is also triggered, and after the safety function is triggered, the controller will enter a safety state, the main positive contactor of the BMS is disconnected, and safety protection is performed.

[0174] (2) Safety function triggering event:

[0175] a: Trigger event 1: running direction and gear do not match:

[0176] That is, the start-up accidental acceleration-starting direction is not controlled, and the safety function is triggered, in order to prevent the forklift from running out of control and causing damage to personnel and causing danger.

[0177] b: Trigger event 2: no speed request, acceleration exceeds limit:

[0178] That is, the start-up accidental acceleration-starting acceleration is not controlled, and the safety function is triggered, in order to prevent the forklift from running out of control and causing damage to personnel and causing danger.

[0179] c: Trigger event 3: fault detection triggering event:

[0180] Sensor failure;

[0181] Controller failure;

[0182] Actuator failure;

[0183] Triggering one of them, that is, the drive control system safety function is triggered, in order to prevent the forklift from suddenly running and causing damage to the operator and surrounding personnel, and to prevent danger from occurring, the safety state should be entered:

[0184] (3) Flat road start-up accidental acceleration subsystem fault diagnosis model design:

[0185] If the flat road start-up accidental acceleration subsystem safety function setting event is triggered, after entering the safety state, while enabling the safety control function to perform action protection, further diagnosis of the cause of the fault is required. In this embodiment, a fault diagnosis model is obtained by training historical fault data of the flat road start-up accidental acceleration subsystem using Bayes, and the cause of the fault is obtained by reverse reasoning to solve the fault.

[0186] S1, construct a flat road start-up accidental acceleration subsystem fault set, use to represent:

[0187]

[0188] In the set, Safety function trigger event 1, Safety function trigger event 2, Safety function trigger event 3.

[0189] S2, collect and aggregate the reasons for triggering the safety state protection of the accidental acceleration subsystem fault of the flat road start, and use set to represent:

[0190]

[0191] In the set, is the potential reason for triggering event 1 that the driving direction and the gear position do not match, is the potential reason for triggering event 2 that there is no speed request and the acceleration exceeds the limit, is the potential reason for triggering event 3 that the fault detection trigger event. The reason set for safety state protection can be extended at any time.

[0192] S3, train the collected and quantized historical fault reason data sample set using a Bayesian greedy learning algorithm to obtain a Bayesian fault diagnosis model:

[0193]

[0194] In the formula: represents the directed acyclic graph structure DAG (Directed Acyclic Graph) of the Bayesian fault diagnosis model; represents the conditional probability table of the variable.

[0195] The trained Bayesian fault diagnosis model is evaluated using the BIC (Bayesian Information Criteria) score function, and the optimal Bayesian fault diagnosis model obtained by training is saved. The BIC score function is as follows:

[0196]

[0197]

[0198] In the formula: is the maximum likelihood value of the model on the given data, that is, the likelihood function value after the model parameter estimation, is the number of free parameters in the model, that is, the complexity of the model, represents the network structure dimension of the Bayesian diagnosis model.

[0199] Since measures the fitting degree of the model to the data, is the penalty term of model complexity, the greater this part indicates the more complex model. Therefore, the model with the minimum BIC value is selected as the final operating position monitoring subsystem Bayesian fault diagnosis model.

[0200] (4) Fault post-processing decision and repair module design

[0201] When the safety function is triggered, the Bayesian fault diagnosis model of the accidental acceleration subsystem during the cut-in flat road start is used , the maximum likelihood estimation algorithm is used to diagnose the potential causes of entering the safety state protection

[0202] First, according to the independent observation data, the likelihood function is constructed:

[0203]

[0204] In the formula: is the parameter to be estimated, is the observation sample, is the likelihood function.

[0205] In order to simplify the calculation, take the logarithm of the likelihood function to get the log-likelihood function:

[0206]

[0207] According to the parameter value function target requirement of the maximum log-likelihood, the formula is as follows:

[0208]

[0209] Regarding Take the derivative and set the derivative to 0:

[0210]

[0211] By solving the equation, the estimated value of the parameter can be obtained , thereby inversely reasoning the fault cause set , such as the corresponding probability value, denoted by :

[0212]

[0213] After using the diagnostic model to obtain the probability of the potential fault cause, enter the fault post-processing decision module, and the probability value of each fault cause ​Sort from large to small, and according to the probability value of the sorting value, the potential causes of triggering the safety state protection are recommended to the authorized technicians one by one, and the authorized technicians handle and repair the fault according to the probability of the fault reason recommended by the diagnostic model.

[0214] (5) Safety state reset condition

[0215] Fault recovery:

[0216] When the accidental acceleration subsystem safety function enters the state protection, the controller loop enters the safety state, and the authorized technician should repair the fault, and the authorized technician handles and repairs the fault according to the probability of the fault reason recommended by the diagnostic model. After the normal operation is restored.

[0217] Combined Figure 6 As shown in the figure, the vehicle driving accidental behavior subsystem safety function

[0218] (1) Safety function logic of accidental behavior during vehicle driving:

[0219] The accidental behavior safety function during vehicle driving is composed of a combination handle and a coded safety-related component. The controller running period is set to ≤50ms. If accidental behaviors such as driving direction response, unexpected acceleration, unexpected speed, etc. are detected, the safety function is triggered, and the controller will enter the safety state. The main positive contactor of the BMS is disconnected, and safety protection is performed.

[0220] (2) Safety function triggering event:

[0221] a: Trigger event 1 driving direction response trigger condition:

[0222] When in forward gear, the motor speed is negative and not decelerating;

[0223] When in reverse gear, the motor speed is positive and not decelerating;

[0224] When in neutral gear, the motor speed is unstable and not decelerating;

[0225] b: Trigger event 2 unexpected acceleration trigger event:

[0226] The absolute value of motor acceleration is greater than 1.3 times the maximum acceleration, and is not in a deceleration state.

[0227] c: Trigger event 3 unexpected speed trigger event:

[0228] The absolute value of motor speed is greater than 1.2 times the expected speed absolute value, and is not in a deceleration state.

[0229] d: Trigger event 4 fault detection trigger event:

[0230] Sensor failure

[0231] Controller failure

[0232] Actuator failure

[0233] (3) Unexpected behavior failure diagnosis model design when the vehicle is running:

[0234] If the unexpected behavior safety function setting event when the vehicle is running is triggered, after entering the safety state, while enabling the safety control function to act as protection, further diagnosis of the cause of the failure is needed. In this embodiment, the Bayesian is used to train the historical failure data of the unexpected behavior of the vehicle running to obtain a failure diagnosis model, and the cause of the failure is obtained by reverse reasoning to solve the failure.

[0235] S1, the vehicle running unexpected behavior subsystem failure set, use to represent:

[0236]

[0237] In the set, is the safety function trigger event 1, is the safety function trigger event 2, is the safety function trigger event 3, is the safety function trigger event 4.

[0238] S2, collect and summarize the reasons for the unexpected behavior subsystem failure of the vehicle running into the safety state protection, use the set to represent:

[0239]

[0240] In the set, is the potential cause of the trigger event 1 running direction response trigger condition, is the potential cause of the unexpected acceleration trigger event, is the potential cause of the trigger event 3 unexpected speed trigger event, is the fault detection trigger event. The reason set for safety state protection can be expanded at any time.

[0241] S3, use the Bayesian greedy learning algorithm to train the collected and quantized historical failure cause data sample set to obtain a Bayesian failure diagnosis model:

[0242]

[0243] In the formula: A Bayesian fault diagnosis model is represented by a directed acyclic graph (DAG). A conditional probability table representing a variable.

[0244] The trained Bayesian fault diagnosis model is evaluated by using a BIC (Bayesian Information Criteria) score function, and the optimal Bayesian fault diagnosis model obtained by training is saved. The BIC score function is shown as follows:

[0245]

[0246]

[0247] In the formula: is the maximum likelihood value of the model on the given data, that is, the likelihood function value after the model parameter estimation, is the number of free parameters in the model, that is, the complexity of the model, represents the network structure dimension of the Bayesian diagnosis model.

[0248] Since measures the fitting degree of the model to the data, is a penalty term for the complexity of the model, and the greater this part is, the more complex the model is. Therefore, the model with the minimum BIC value is selected as the final operating position monitoring subsystem Bayesian fault diagnosis model.

[0249] (4) Fault post-processing decision and repair module design

[0250] When the safety function is triggered, the Bayesian fault diagnosis model of the vehicle driving unexpected behavior subsystem is cut in The maximum likelihood estimation method is used to diagnose and reason the potential causes of entering the safety state protection

[0251] First, the likelihood function is constructed according to the independent observation data:

[0252]

[0253] In the formula: is the parameter to be estimated, is the observation sample, is the likelihood function.

[0254] In order to simplify the calculation, the likelihood function is taken as the logarithm, and the logarithm likelihood function is obtained:

[0255]

[0256] According to the parameter value function target requirement of maximum log-likelihood, the following formula can be obtained:

[0257]

[0258] To Regarding Derivation, and set the derivative to 0:

[0259]

[0260] By solving the equation, the estimated value of the parameter can be obtained , and the fault cause set can be obtained by reverse reasoning. , which corresponds to the probability value:

[0261]

[0262] After obtaining the probability of potential fault causes using the diagnostic model, enter the fault post-processing decision module, sort the probability values of each fault cause from large to small, and recommend the potential causes that trigger the safety state protection to the authorized technical personnel one by one according to the sorting value of the probability value. The authorized technical personnel handle and repair the fault according to the probability of the fault cause recommended by the diagnostic model.

[0263] (5) Safety state reset condition

[0264] Fault recovery:

[0265] When the unexpected behavior safety function enters the state protection while the vehicle is driving, the controller loop enters the safety state, and the authorized technical personnel should repair the fault. After the authorized technical personnel handle and repair the fault according to the probability of the fault cause recommended by the diagnostic model, the normal operation is restored.

[0266] Although embodiments of the present application have been shown and described, those skilled in the art can understand that various changes, modifications, replacements and variations can be made to these embodiments without departing from the principles and purposes of the present application, and the scope of the present application is defined by the claims and their equivalents.

Claims

1. A fault diagnosis and decision-making system, used in conjunction with a functional safety control system for an electric forklift, wherein the functional safety control system for the electric forklift includes a safety function module, which includes a drive control subsystem, an operating position monitoring subsystem, a start-up unexpected acceleration subsystem, and a vehicle driving unexpected behavior subsystem. The safety function module monitors the signal status of each component in the drive control subsystem, operating position monitoring subsystem, start-up unexpected acceleration subsystem, and vehicle driving unexpected behavior subsystem. The detection period for each component signal of the safety function module is ≤50ms. If the abnormal trigger condition defined for the monitored subsystem is triggered, the corresponding subsystem module enters a safety function protection state. At this time, the main positive contactor of the BMS is disconnected, requiring authorized technicians to perform fault recovery. If no abnormality is detected or the technicians have already recovered the fault, monitoring of the component signals of the drive control subsystem, operating position monitoring subsystem, start-up unexpected acceleration subsystem, and vehicle driving unexpected behavior subsystem continues. Its features are: Historical fault model data samples of drive control subsystem, operation position monitoring subsystem, start-up unexpected acceleration subsystem, and vehicle driving unexpected behavior subsystem are collected. Bayesian greedy algorithm is used to train the model and establish fault diagnosis and reasoning models for each subsystem. Based on the set of fault probabilities corresponding to potential causes obtained by fault diagnosis model inference, the maximum likelihood estimation algorithm is used to arrange the probabilities of potential fault causes, establish the causal relationship between historical fault data, and represent the model through a directed acyclic graph. Based on the probability values, the potential causes that trigger the safety protection are recommended to the authorized technicians one by one. The authorized technicians then handle and repair the faults based on the probability-ranked fault causes recommended by the diagnostic model. After the repairs are completed, normal operation is restored and the safety protection is deactivated.

2. The fault diagnosis decision system according to claim 1, characterized in that: The method for establishing fault diagnosis back-inference models for each subsystem includes: S1. Construct the fault set for each functional safety subsystem, and use... , , , express; ； ； ； ； In the set, To drive the safety function trigger events of the control subsystem, Safety function trigger events for the operation location monitoring subsystem; This is to trigger a safety function event for the unexpected acceleration subsystem. Safety function triggering events for the vehicle driving accident behavior subsystem; S2. Collect and summarize the reasons that trigger the failure of each functional safety subsystem to enter the safety state protection state, and use setter... express: 、 、 ； In the set, Potential causes for triggering safety function events in the drive control subsystem; Potential causes for triggering safety function events in the operation position monitoring subsystem; …potential causes that could lead to the triggering of safety function events in the Acceleration Subsystem at Startup; Potential causes that could trigger safety functions in the vehicle driving accident behavior subsystem; the set of reasons for implementing safety state protection. It can be expanded at any time; S3. Using the Bayesian greedy learning algorithm, the quantified historical fault cause data sample set collected from each functional safety subsystem is trained to obtain the Bayesian fault diagnosis model: ； In the formula: A Directed Acyclic Graph (DAG) structure represents a Bayesian fault diagnosis model. A conditional probability table representing variables; The trained Bayesian fault diagnosis model is evaluated using the BIC scoring function, and the optimal Bayesian fault diagnosis model obtained from the training is saved. The BIC scoring function is shown below: ； ； In the formula: It is the maximum likelihood value of the model on the given data, that is, the likelihood function value after estimating the model parameters. It refers to the number of free parameters in the model, which is also the complexity of the model. This represents the dimension of the network structure of the Bayesian diagnostic model; The model with the smallest BIC value was selected as the final Bayesian fault diagnosis model for the functional safety subsystem.

3. The fault diagnosis decision system according to claim 2, characterized in that: Based on the set of fault probabilities corresponding to potential causes obtained from the fault diagnosis model, the method of arranging the probabilities of potential fault causes using the maximum likelihood estimation algorithm includes: When a subsystem triggers a safety function, the Bayesian fault diagnosis model of the current subsystem is switched on. Using the maximum likelihood estimation algorithm to investigate the causes Each set of fault causes is used to diagnose and infer the potential causes that lead to the safe state protection. S1. First, construct the likelihood function based on the independent observation data: ； In the formula: These are the parameters to be estimated. For observation samples, It is the likelihood function; S2. Taking the logarithm of the likelihood function yields the log-likelihood function: ； S3. Based on the objective requirement of maximizing the log-likelihood parameter value function, the following formula can be obtained: ； right about Take the derivative and set it to 0: ； S4. Parameters can be obtained by solving the equations. The estimated value This allows us to deduce the set of causes of failure through reverse reasoning. The corresponding probability value, using express: ； ； ； ； S5. After obtaining the probability of potential fault causes using the fault diagnosis model, the system enters the fault post-processing decision module, sorting the probability values ​​of each fault cause that triggers the safety protection state for the touch operation position monitoring from largest to smallest. Based on the probability values, the potential causes that trigger the safety protection are recommended to the authorized technicians one by one. The authorized technicians then handle and repair the faults based on the probability-ranked fault causes recommended by the diagnostic model.

Images