Self-adaptive tramcar operation optimization method and system suitable for tramcar

Abstract

The invention discloses a self-adaptive car operation optimization method and system suitable for a mine car. The method comprises the following steps: acquiring a multi-modal data set of the mine car; establishing a mine car kinematic model based on extended Kalman filtering in combination with abnormal fault-tolerant processing of the multi-modal data set; taking the multi-modal data set as the output of a mine car kinematics model, and outputting a real-time mine car motion state; according to the motion state of the mine car, motion environment perception and fuzzy logic reasoning decision based on dynamic score optimization are carried out, and a mine car control instruction is output; when a mine car is operated based on a mine car control instruction, a fuzzy logic reasoning decision dynamic feedback optimization mechanism based on error mode recognition is implemented and is applied to generation of a mine car control instruction in the next period. The stability, the reliability and the self-adaptability of mine car operation operation are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of mine control, and in particular to an adaptive operation optimization method and system applicable to a mine car. Background Art

[0002] In the mine hoisting system, the operation of the wellhead crane has a vital impact on the production safety and efficiency of the entire mine.

[0003] With the development of sensor technology and intelligent control technology, modern mine hoisting systems have begun to introduce automated and intelligent solutions. However, the following technical problems still exist:

[0004] Single-sensor systems lack detection accuracy: Existing mine hoisting systems typically rely on a single type of sensor (such as an infrared sensor) for position detection. This approach is susceptible to environmental interference in complex mine environments, resulting in insufficient detection accuracy and stability, which in turn affects system safety.

[0005] Lack of intelligent optimization capabilities: Traditional automated operating systems are often unable to intelligently adjust to real-time operating conditions. This results in rigid operating procedures that lack flexibility and adaptability, making them difficult to cope with complex and changing mine environments. Such systems lacking adaptive optimization capabilities cannot effectively improve mine production efficiency and safety.

[0006] Manual operation poses safety risks: Traditional mine operations rely on manual operation, which can easily lead to accidents due to operator errors or lack of experience. Furthermore, manual operation is susceptible to subjective judgment and limited vision, leading to frequent problems such as vehicle derailment, misplaced carts, misplaced carts, and overloaded carts.

[0007] In the prior art, the invention patent with the patent application publication number CN119659641A provides a vehicle state prediction method and related equipment. This method integrates vehicle sensor data from different sources and types, enabling the vehicle state prediction model to fully utilize multimodal fusion data for calculations and analysis, so as to adapt to different driving environments and working conditions for vehicle state prediction and improve the accuracy of vehicle state prediction. This patent applies multimodal data to improve the state prediction accuracy of road vehicles for use in automatic vehicle driving and intelligent driving assistance. However, during state prediction, there is no abnormal fault tolerance processing for multimodal data, and the continuity of multimodal data cannot be guaranteed, and a real-time control strategy cannot be given. Summary of the Invention

[0008] The technical problem to be solved by the present invention is to improve the reliability of mine car control in a complex mine environment.

[0009] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0010] An adaptive operation optimization method for a mine car, comprising:

[0011] Obtain a multimodal dataset of mine carts;

[0012] Based on the extended Kalman filter and combined with the abnormal fault tolerance processing of multimodal data sets, a mine car kinematic model is established; the multimodal data set is used as the output of the mine car kinematic model to output the real-time mine car motion status;

[0013] According to the movement state of the mine car, it performs motion environment perception and fuzzy logic reasoning decision-making based on dynamic scoring optimization, and outputs mine car control instructions;

[0014] When the mine car is operated based on the mine car control instructions, a fuzzy logic reasoning decision-making dynamic feedback optimization mechanism based on error pattern recognition is implemented and applied to the generation of the mine car control instructions for the next cycle.

[0015] Technical Effect: Dynamically adjusts the Kalman gain based on measurement residuals and switches to redundant data sources in the event of sensor failure, ensuring fusion continuity and high robustness. Based on the error-dominant identification results, the membership function or control rules of the fuzzy inference module are adaptively and dynamically adjusted. Control performance is continuously optimized through long-term experience accumulation, achieving adaptive evolution.

[0016] In this embodiment, obtaining a multimodal dataset of a mine car includes:

[0017] Multimodal sensors are set up on the mine car transportation path to collect the mine car arrival status signal, mine car absolute position information, obstacle relative distance information and track flatness change information. Based on the timestamp marking mechanism, the collected information is synchronized in time and spatially associated, and then standardized preprocessing is performed to obtain a multimodal data set.

[0018] Technical effect: In the mine vehicle operation system, sensors with three different detection principles, namely infrared photoelectric, laser ranging, and ultrasonic ranging, are combined to achieve temporal synchronization and spatial association of multimodal heterogeneous data through a unified timestamp marking mechanism.

[0019] In this embodiment, obtaining the real-time movement status of the mine car includes:

[0020] Based on the mine car's current position estimation and motion control input, a state prediction model is constructed to obtain the predicted state;

[0021] Construct observation equations for various sensors that collect multimodal data sets and obtain measurement data from each sensor;

[0022] Calculate the measurement residual based on the difference between the measurement data and the predicted state of each sensor;

[0023] Dynamically adjust the Kalman gain matrix based on measurement residuals and sensor type: When the measurement residuals continuously exceed the preset residual threshold range or the sensor signal quality index falls below the set confidence level, the weight of the data information collected by the sensor in the fusion process is automatically reduced, or the preset redundant sensor channel is switched;

[0024] Technical effect: Real-time monitoring of sensor measurement residuals, and dynamic adjustment of process noise covariance matrix and measurement noise covariance matrix based on the residuals. When the sensor is abnormal or its performance degrades, its data weight is automatically reduced or it is switched to redundant channel compensation data, ensuring high precision and high robustness of the fusion output, and significantly improving the reliability of state estimation of the mine vehicle operation system in complex environments.

[0025] The state prediction model is updated based on the dynamically adjusted Kalman gain matrix.

[0026] In this embodiment, the update equation of the state prediction model is:

[0027]

[0028] Where K k is the Kalman gain matrix, P k∣k-1 is the state prediction covariance matrix, H k is the observation matrix, R k is the measurement noise covariance matrix, r k is the measurement residual, They are the predicted states of the k-th and k-1-th mine car motion states respectively.

[0029] In this embodiment, outputting the mine car control instruction includes:

[0030] Collect the operating environment data of the mine car, including track flatness, obstacle information, and mine car running speed, as input for fuzzy logic reasoning decision-making;

[0031] The input quantity of fuzzy logic reasoning decision is divided into categories, and each category is logically distinguished based on the measured value relative to the set threshold range;

[0032] The input variables are fuzzy processed and triangular membership function with dynamically adjustable parameters is used for modeling. The triangular membership function parameters of each input variable are dynamically adjusted. The adjustable parameters of the triangular membership function are corrected online based on the statistical results of the real-time operation data of the mine car.

[0033] In the inference stage, according to the current input, the membership distribution results of the modified triangular membership function are comprehensively considered, and the confidence weighted centroid method is used for defuzzification to output continuous mine car control instructions.

[0034] In this embodiment, the adjustable parameters of the online correction triangular membership function include:

[0035] Based on the running status of the mine car collected during operation, the movement status evaluation indicators are set, including track running smoothness, obstacle avoidance success rate and speed control accuracy evaluation indicators;

[0036] According to the results of each driving operation, based on the motion state evaluation index, the experience score of each fuzzy inference rule is calculated;

[0037] The fuzzy inference rules are sorted according to their empirical scores within the statistical period. Inefficient fuzzy inference rules with scores lower than the set evaluation threshold are marked as candidates for elimination.

[0038] According to the elimination strategy, the inefficient fuzzy inference rules are deleted, and according to the real-time collected data characteristics and system optimization goals, new fuzzy inference rules are automatically introduced to correct the adjustable parameters of the triangular membership function online.

[0039] Technical Effect: By adjusting the center point and span parameters of the input variable membership function in real time, input fuzzy adaptation is achieved. At the same time, the fuzzy inference rule set is empirically scored and dynamically updated based on the actual vehicle operation results, inefficient rules are automatically eliminated and new control rules are introduced, continuously improving the accuracy and flexibility of decision-making in complex environments.

[0040] In this embodiment, the generation of the mine car control instruction for the next cycle includes:

[0041] Real-time acquisition of track surface changes, obstacle distance changes, and mine car speed changes; and calculation of the normalized deviation component of each change at each moment to construct a weighted comprehensive deviation;

[0042] According to the weighted comprehensive deviation and each individual deviation, the dominant deviation type is determined by using the deviation dominant judgment function;

[0043] Based on the identified dominant deviation type, the fuzzy logic reasoning decision is dynamically adjusted according to the following strategies:

[0044] If the dominant deviation type is track deviation, the center position and span parameters of the track flatness membership function are adjusted first. If the dominant deviation type is obstacle deviation, the obstacle-related fuzzy inference rules are reconstructed first. If the dominant deviation type is speed deviation, the triangular membership function of the speed input and the speed control output rules are adjusted first.

[0045] And when executing the adjustment of fuzzy logic reasoning decision, an adjustment increment is introduced; the updated fuzzy reasoning rule set and the adjustable parameters of the triangular membership function are applied in real time to the generation of the next cycle of mine car control instructions.

[0046] Technical effect: An error pattern recognition mechanism based on deviation vector calculation is constructed. After determining the dominant error type, the corresponding fuzzy membership function and inference rules are dynamically adjusted to form targeted optimized control instructions. Experience is accumulated through the feedback learning mechanism, and the adjustment strategy is gradually optimized to construct an intelligent closed-loop vehicle control system that integrates multimodal perception, state fusion, intelligent decision-making and adaptive feedback, which significantly improves the stability, reliability and adaptability of mine vehicle operation.

[0047] In this embodiment, the mine car control instructions for the next cycle are recorded, corresponding to the adjustment effect indicators of the mine car operation; based on the historical data of the adjustment effect indicators accumulated during long-term operation, the weights of each deviation in the weighted comprehensive deviation and the learning rate of the adjustment increment are dynamically optimized to achieve self-learning and adaptive evolution of the feedback mechanism.

[0048] In this embodiment, the dominant deviation type is obtained by the following formula:

[0049] M(t)=arg max{w s |Δs(t)|,w o |Δo(t)|,w v |Δv(t)|};

[0050] Where M(t) is the dominant deviation type, w s 、w o 、w v are the deviation weights corresponding to the track surface change, obstacle distance change, and minecart speed change, respectively. Δs(t), Δo(t), and Δv(t) are the track surface change, obstacle distance change, and minecart speed change, respectively. arg max is a function.

[0051] The present invention further provides an adaptive vehicle operation optimization system applicable to a mine car, which applies the above-mentioned adaptive vehicle operation optimization method applicable to a mine car, comprising:

[0052] Multimodal data module, used to obtain multimodal datasets of mine cars;

[0053] The vehicle operation model module is used to establish a mine car kinematic model based on the extended Kalman filter and combined with the abnormal fault tolerance processing of the multimodal data set. The multimodal data set is used as the output of the mine car kinematic model to output the real-time mine car motion status.

[0054] The control instruction module is used to perceive the motion environment and make fuzzy logic reasoning decisions based on dynamic scoring optimization according to the motion state of the mine car, and output the mine car control instructions;

[0055] The feedback module is used to implement a dynamic feedback optimization mechanism of fuzzy logic reasoning decision-making based on error pattern recognition when the mine car is operated based on the mine car control instructions, and is applied to the generation of the mine car control instructions for the next cycle.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] This solves the problems of incorrect operation and limited vision that may occur during manual operation, avoiding unsafe situations such as vehicle derailment, improperly pushed or released vehicles, and overloaded vehicles. Through precise control and real-time monitoring of the operation process, the safety of vertical shaft hoisting operations is effectively improved, and the risk of accidents caused by human error is reduced.

[0058] The automated vehicle operation system of this invention enables the entire vehicle operation process to be completed with a single click. The system operates continuously and smoothly, avoiding the pauses and interruptions that can occur during manual operation due to operational inconsistencies or operational errors. Through automated control, the vehicle operation process no longer relies on the operator's experience and skills, thus ensuring continuous and consistent operation.

[0059] Compared with traditional manual operation, this invention significantly shortens pithead operation time. The automated operation system can complete operations such as pushing, placing, and loading carts within the tank in the shortest possible time, reducing the time lost due to cumbersome manual procedures and slow response times. By optimizing the operation path and operating procedures, the system improves production efficiency and increases the overall operating speed of the mine hoisting system.

[0060] Reduced reliance on operators and lower human resource consumption. The automated operation system requires only a small number of operators for monitoring and exception handling, significantly reducing the number of personnel required at the wellhead. Furthermore, the system's intelligent control reduces operator training and skill requirements, further reducing human resource costs.

[0061] The use of multi-sensor fusion and fuzzy logic control enables the system to adaptively adjust operating parameters to adapt to different mine conditions, such as uneven tracks and complex ambient lighting. Compared with existing technologies, this invention is more adaptable and stable, ensuring safe and efficient operation under complex working conditions.

[0062] This invention has been successfully applied to the automatic vehicle operation system in a certain area of mines, showing significant economic and safety benefits. The specific application effects are as follows:

[0063] 1) Economic Benefits: Since the system was put into operation, the daily transport volume of mine cars has reached 200-300 cars, creating direct economic benefits of nearly 1 million yuan annually. The application of this system has significantly reduced the inefficiencies and operational errors caused by manual operation, significantly improving the overall operational efficiency and economic benefits of the mine.

[0064] 2) Safety Benefits: Compared to traditional manual operation, this invention achieves fully automated control of mine car operation through multi-sensor fusion and fuzzy logic adaptive optimization control. Through precise data monitoring and intelligent control, it effectively avoids potential safety hazards associated with manual operation, improves the safety factor of mine operations, and ensures the safety of operators.

[0065] 3) Social Benefits: The widespread application of this invention not only improves the operational efficiency of coal mining enterprises but also promotes the intelligent development of mining technology, reducing reliance on manual operations. By improving the safety and automation of mine operations, this invention contributes positively to the technological advancement and sustainable development of the coal mining industry.

[0066] 4) Promotion prospects: Currently, most vertical shaft auxiliary shafts in China are still controlled manually, which is cumbersome and has a low safety factor. As an efficient and safe mine automatic operation system, the present invention has broad promotion prospects. The promotion and application of this invention in vertical shaft auxiliary shafts nationwide can not only significantly improve mine production efficiency and safety levels, but also create huge economic and social benefits, and promote the modernization and transformation of the coal mining industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 The present invention is a flowchart of an adaptive vehicle operation optimization method applicable to a mine vehicle according to an embodiment of the present invention.

[0068] Figure 2 、 Figure 3 Schematic diagram showing multimodal data collected according to an embodiment of the present invention.

[0069] Figure 4 Schematic diagram of a vehicle operating system applying the method according to an embodiment of the present invention.

[0070] Figure 5 This is a block diagram of an adaptive vehicle operation optimization system applicable to a mine vehicle according to an embodiment of the present invention. DETAILED DESCRIPTION

[0071] To facilitate those skilled in the art to understand the technical solution of the present invention, the technical solution of the present invention is further described with reference to the accompanying drawings.

[0072] The terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, "plurality" means two or more, unless otherwise specifically defined.

[0073] See also Figure 1 As shown, the present invention provides an adaptive operation optimization method applicable to a mine car, comprising:

[0074] S10, obtaining a multimodal dataset of a mine car.

[0075] See also Figure 2 、 3 As shown, in one embodiment of the present invention, a multimodal sensor is set on the mine car transportation path to collect the mine car arrival status signal, the mine car absolute position information, the obstacle relative distance information and the track flatness change information. The collected information is synchronized in time and spatially associated based on the timestamp marking mechanism, and then standardized preprocessing is performed to obtain a multimodal data set.

[0076] In this embodiment, multiple types of sensors, including infrared photoelectric sensors, laser ranging sensors, and ultrasonic sensors, are deployed at key locations within the mine hoisting system. The infrared photoelectric sensors utilize a non-contact switch detection method to collect signals indicating the mine car's arrival status. The laser ranging sensors utilize the principle of high-frequency, narrow-pulse laser reflection ranging to collect information about the mine car's absolute position and track flatness changes. The ultrasonic sensors utilize the principle of ultrasonic time delay measurement to collect information about the relative distance to obstacles. A timestamp mechanism is used to synchronize and spatially correlate data from sensors in different modalities.

[0077] In this embodiment, the collected information is synchronized in time and spatially associated based on a timestamp marking mechanism, including:

[0078] S11, during the data collection process, assign a timestamp to each piece of data collected by the sensor, and synchronize and sort the modal data according to the timestamp. The synchronization conditions are:

[0079]

[0080] Where, d i ,d j are the data units of different sensors, t i ,t j is the corresponding timestamp, Δt sync is the synchronization tolerance threshold, Any number.

[0081] S12: spatially associate the synchronized modal data with each other according to the sensor spatial position calibration relationship to construct a multimodal data set.

[0082] D(t)={d IR (t),d Laser (t),d Ultra (t)};

[0083] Where D(t) is the multimodal dataset, d IR (t),d Laser (t),d Ultra (t) are the normalized data of infrared photoelectric, laser ranging and ultrasonic ranging sensors at the same time t.

[0084] S20, based on the extended Kalman filter and combined with the abnormal fault tolerance processing of the multimodal data set, a mine car kinematic model is established; the multimodal data set is used as the output of the mine car kinematic model to output the real-time mine car motion state.

[0085] In one embodiment of the present invention, a dynamic fault-tolerant acquisition strategy for interference from mine dust, humidity and strong vibration environments is acquired to achieve multi-modal composite real-time perception of the mine car position status, track environment status and obstacle distribution status.

[0086] In this embodiment, mine environmental parameters, including dust concentration Cd(t), air humidity H(t), and system vibration intensity V(t), are monitored in real time, and environmental status assessment is performed.

[0087] When any environmental parameter exceeds the set threshold, that is:

[0088] C d (t)>C d,thresh orH(t)>H thresh orV(t)>V thresh ;

[0089] Then dynamically adjust the sensor sampling strategy, increase the data sampling frequency of key sensors, or lengthen the stable sampling window length to ensure data quality. d,thresh 、H thresh 、V thresh The dust, humidity and vibration thresholds are set respectively.

[0090] When the confidence level of the main sensor signal Ri(t) is lower than the set reliability standard Rmin, that is:

[0091] Ri(t) <Rmin;

[0092] It automatically switches to the preset redundant sensor channel to maintain the continuity and accuracy of data collection. Among them, Ri(t) can be understood as the current sensor signal quality score, and Rmin is the set minimum acceptable confidence value.

[0093] In one embodiment of the present invention, obtaining the real-time movement status of the mine car includes:

[0094] S21, based on the current position estimation of the mine car and the motion control input, a state prediction model is constructed to obtain the predicted state.

[0095] In this embodiment, the state vector is predicted using the following state transition equation:

[0096]

[0097] Where, is the predicted state of the minecart motion, is the estimated state of the last minecart motion, u k-1 is the control input, including position, speed and direction angle, w k is the Gaussian process noise in the state transition, satisfying w k ~N(0,Q k ), that is, assuming that the mean is 0 and the process noise covariance is Q k Gaussian white noise, k is the number of times, N is the normal distribution, specifically represents w k The mean is 0 and the covariance matrix is Q k Normal distribution.

[0098] S22, construct observation equations for various sensors that collect multimodal data sets and obtain measurement data from each sensor.

[0099] In this embodiment, the observation equations for the infrared photoelectric sensor switch signal, the laser ranging sensor position information, and the ultrasonic sensor relative distance information are defined as follows:

[0100]

[0101] Where z k is the sensor observation data, h(·) is the nonlinear observation function, v k To measure noise.

[0102] S23, calculating the measurement residual according to the difference between the measurement data of each sensor and the predicted state.

[0103] In this embodiment, the measurement residual r k Obtained by the following formula:

[0104]

[0105] The measurement residual reflects the deviation between the actual observation value and the predicted value and is used for subsequent anomaly detection and gain adjustment.

[0106] S24, dynamically adjust the Kalman gain matrix based on the measurement residual and sensor type: when the measurement residual continuously exceeds the preset residual threshold range or the sensor signal quality index is lower than the set confidence standard, automatically reduce the weight of the data information collected by the sensor in the fusion process, or switch to the preset redundant sensor channel.

[0107] In this embodiment, the Kalman gain matrix K is adjusted based on the measurement residual size. k , perform the fusion update process, specifically, the update equation of the state prediction model is:

[0108]

[0109] Where K k is the Kalman gain matrix, P k∣k-1 is the state prediction covariance matrix, H k is the observation matrix, R k is the measurement noise covariance matrix, They are the predicted states of the k-th and k-1-th mine car motion states respectively.

[0110] In this embodiment, the process noise covariance matrix Q is dynamically adjusted according to the sensor type and the measurement residual distribution. k and the measurement noise covariance matrix R k , to enhance the fusion results The stability and robustness of the process noise covariance matrix Q k and the measurement noise covariance matrix R k They are used to describe the uncertainty of the system model, that is, the noise in the state prediction process, and the uncertainty of the sensor observation. If the process noise covariance matrix Q k The larger the value, the more trust the sensor observation; if the measurement noise covariance matrix R k The larger the value, the more trust the system model has. The two together determine the Kalman gain matrix K k , thus affecting the state update.

[0111] S25, updating the state prediction model based on the dynamically adjusted Kalman gain matrix.

[0112] In this embodiment, the fusion result is output As a high-precision state estimate of the mine car at the current moment, including position, speed and direction angle, it is used for subsequent control decisions.

[0113] S30, according to the movement state of the mine car, the movement environment perception and fuzzy logic reasoning decision based on dynamic score optimization are performed to output the mine car control instructions.

[0114] See also Figure 4 As shown, in one embodiment of the present invention, outputting a mine car control instruction includes:

[0115] S31, collects the operating environment data of the mine car, including track flatness, obstacle information, and mine car running speed, as input for fuzzy logic reasoning decision-making.

[0116] S32, the input quantity of the fuzzy logic reasoning decision is divided into categories, and each category is logically distinguished based on the measurement value relative to the set threshold range.

[0117] In this embodiment, the track flatness S is divided according to the height variation of the track surface, the obstacle information O is divided according to the detection distance of the obstacle in front of the mine car, and the mine car running speed V is divided according to the current mine car movement speed.

[0118] In this embodiment, track flatness S is categorized as flat, slightly undulating, and severely undulating. Obstacle information O is categorized as no obstacle, slight obstacle, and major obstacle. Car speed V is categorized as high, medium, and low. Each category is logically distinguished based on the measured value relative to a set threshold range, without requiring a fixed value.

[0119] S33, fuzzy processing is performed on the input quantity, and a triangular membership function with dynamically adjustable parameters is adopted for modeling. The triangular membership function parameters of each input variable are dynamically adjusted, and the adjustable parameters of the triangular membership function are corrected online based on the statistical results of the real-time operation data of the mine car.

[0120] In this embodiment, the triangular membership function μ(x) is:

[0121]

[0122] Where x is the input variable value, and a, b, and c are the left endpoint, vertex, and right endpoint position parameters of the triangular membership function, respectively. The triangular membership function parameters for each input variable are dynamically adjusted. Based on the real-time statistics of the mine car operation, the positions of parameters a, b, and c are corrected online to adapt to changes in the mine track status and environment.

[0123] In this embodiment, the fuzzy inference rule set is dynamically optimized based on the empirical scores of the actual operation effect of the vehicle. Specifically, the adjustable parameters of the online correction triangular membership function include:

[0124] S331, based on the running status of the mine car collected during the operation process, set the movement status evaluation indicators, including track running smoothness, obstacle avoidance success rate and speed control accuracy evaluation indicators.

[0125] S332, according to each driving result, based on the motion state evaluation index, calculate the experience score for each fuzzy inference rule.

[0126] In this embodiment, a single fuzzy inference rule R is defined. i The score is:

[0127]

[0128] Among them, n success (R i ) is the fuzzy inference rule R i The number of successful driving maneuvers under guidance, n total (R i ) is the fuzzy inference rule R i The total number of triggers.

[0129] S333, sorting the fuzzy inference rules according to their experience scores within the statistical period, and marking the inefficient fuzzy inference rules with scores lower than the set evaluation threshold as elimination candidates first.

[0130] S334, based on the elimination strategy, performs a deletion operation on the inefficient fuzzy inference rules, and automatically introduces new fuzzy inference rules based on the real-time collected data characteristics and system optimization goals, and corrects the adjustable parameters of the triangular membership function online.

[0131] In this embodiment, after reasoning based on the above-mentioned membership, the system evaluates the performance of each fuzzy reasoning rule in the actual driving process in real time and scores each fuzzy reasoning rule. For rules with scores below the threshold, they are marked, eliminated or updated, and new rules are automatically introduced based on the current operating data characteristics to maintain the efficiency of the rule base. When a, b, and c are adjusted, the membership calculation results will change, which will affect the triggering frequency and score distribution of the rules. When the rule base is optimized (new or eliminated rules), different rules cover different input intervals, which will change the need for re-tuning of a, b, and c, so as to more accurately capture the numerical range of the new rule logic. Take the track flatness S as an example to illustrate. Among them, the unit of track flatness S is: mm, and the value range is 0-10mm, as shown in Table 1.

[0132] Table 1 Membership degree of track flatness S.

[0133] fuzzy sets a b c Semantics flat 0 1 3 Almost no ups and downs Slight ups and downs 2 5 8 Small to moderate orbital altitude changes Severe ups and downs 7 9 10 Small to moderate orbital altitude changes

[0134] During operation, the system continuously fine-tunes a, b, and c based on the latest sensor data distribution to respond to orbital or environmental changes. Typically, the 5%, 50%, and 95% percentiles of the observed data are used as a, b, and c. Therefore, a, b, and c will adjust based on the observed values.

[0135] S34, in the inference stage, according to the current input, the membership distribution result of the triangular membership function output after comprehensive correction is used, and the confidence weighted centroid method is used for defuzzification to output continuous mine car control instructions.

[0136] In this embodiment, the mine car control instruction is obtained by the following formula:

[0137]

[0138] Where y is the final output of the mine car control instruction. i (x) is the membership degree of the i-th fuzzy inference rule under the current input x. i The output corresponding to the recommended value is applied in real time during the operation of the continuous mine car control instructions, achieving continuous dynamic adjustment of the mine car speed, push force, and path. G is the number of rules currently involved in the reasoning.

[0139] S40, when the mine car is operated based on the mine car control instruction, a fuzzy logic reasoning decision dynamic feedback optimization mechanism based on error pattern recognition is implemented and applied to the generation of the mine car control instruction for the next cycle.

[0140] In this embodiment, the generation of the mine car control instruction for the next cycle includes:

[0141] S41, collecting track surface variation, obstacle distance variation, and mine car speed variation in real time; and calculating the normalized deviation component of each variation at each moment to construct a weighted comprehensive deviation.

[0142] In this embodiment, based on the collected data, the normalized deviation components at each moment are calculated, namely the track surface change Δs(t), the obstacle distance change Δo(t), and the mine car speed change Δv(t). A weighted comprehensive deviation ΔE(t) is constructed and defined as:

[0143] ΔE(t)=w s |Δs(t)|+w o |Δo(t)|+w v |Δv(t)|

[0144] Among them, w s 、w o 、w vThese are the deviation weights corresponding to the track surface change, obstacle distance change, and minecart speed change, which are adaptively set according to the importance of the current environment.

[0145] S42, determining the dominant deviation type using a deviation dominant determination function based on the weighted comprehensive deviation amount and each individual deviation amount.

[0146] In this embodiment, the dominant deviation type is obtained by the following formula:

[0147] M(t)=arg max{w s |Δs(t)|,w o |Δo(t)|,w v |Δv(t)|};

[0148] Where M(t) is the dominant deviation type, which is used to guide subsequent optimization actions, and arg max is a function.

[0149] S43, based on the identified dominant deviation type, dynamically adjust the fuzzy logic reasoning decision according to the following strategy:

[0150] If the dominant deviation type is track deviation, priority is given to adjusting the center position and span parameters of the track flatness membership function; if the dominant deviation type is obstacle deviation, priority is given to reconstructing the obstacle-related fuzzy inference rules; if the dominant deviation type is speed deviation, priority is given to adjusting the triangular membership function of the speed input and the speed control output rules.

[0151] And when executing the adjustment of fuzzy logic reasoning decision, an adjustment increment is introduced; the updated fuzzy reasoning rule set and the adjustable parameters of the triangular membership function are applied in real time to the generation of the next cycle of mine car control instructions.

[0152] In this embodiment, when performing the adjustment, the adjustment increment δ is introduced. p , defined as:

[0153] δ p =k p Δp(t);

[0154] Among them, δ p It can be understood as the adjustment range of the corresponding membership function or fuzzy inference rule parameters, k p is the learning rate factor, and Δp(t) is the current deviation value, which can be understood as the current deviation value corresponding to the current deviation type determined after each deviation type is evaluated.

[0155] In this embodiment, the mine car control instructions for the next cycle are recorded, corresponding to the adjustment effect indicators of the mine car operation; based on the historical data of the adjustment effect indicators accumulated during long-term operation, the weights of each deviation in the weighted comprehensive deviation and the learning rate of the adjustment increment are dynamically optimized to achieve self-learning and adaptive evolution of the feedback mechanism.

[0156] See also Figure 5 As shown, the present invention further provides an adaptive vehicle operation optimization system applicable to a mine car, which applies the above-mentioned adaptive vehicle operation optimization method applicable to a mine car, including:

[0157] Multimodal data module, used to obtain multimodal datasets of mine carts.

[0158] The vehicle operation model module is used to establish a mine car kinematic model based on the extended Kalman filter and abnormal fault tolerance processing combined with multimodal data sets; the multimodal data set is used as the output of the mine car kinematic model to output the real-time mine car motion status.

[0159] The control instruction module is used to perceive the motion environment and make fuzzy logic reasoning decisions based on dynamic scoring optimization according to the motion state of the mine car, and output the mine car control instructions.

[0160] The feedback module is used to implement a dynamic feedback optimization mechanism of fuzzy logic reasoning decision-making based on error pattern recognition when the mine car is operated based on the mine car control instructions, and is applied to the generation of the mine car control instructions for the next cycle.

[0161] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims rather than the foregoing description. It is intended that all variations within the meaning and range of equivalents of the claims be embraced herein, and any reference signs in the claims should not be construed as limiting the claims to which they relate.

[0162] The above-mentioned embodiments merely represent the implementation methods of the invention. The protection scope of the present invention is not limited to the above-mentioned embodiments. For those skilled in the art, several variations and improvements can be made without departing from the concept of the present invention, which all fall within the protection scope of the present invention.

Claims

1. An adaptive operation optimization method for a mine car, characterized in that: include: Obtain a multimodal dataset of mine carts; A mine car kinematic model is established based on the extended Kalman filter and combined with the abnormal fault tolerance processing of multimodal data sets; The multimodal dataset is used as the output of the mine car kinematic model to output the real-time mine car motion status; According to the movement state of the mine car, it performs motion environment perception and fuzzy logic reasoning decision-making based on dynamic scoring optimization, and outputs mine car control instructions; When the mine car is operated based on the mine car control instructions, a fuzzy logic reasoning decision-making dynamic feedback optimization mechanism based on error pattern recognition is implemented and applied to the generation of the mine car control instructions for the next cycle.

2. The self-adaptive operation optimization method for a mine car according to claim 1, characterized in that: Get a multimodal dataset of minecarts, including: Multimodal sensors are set up on the mine car transportation path to collect the mine car arrival status signal, mine car absolute position information, obstacle relative distance information and track flatness change information. Based on the timestamp marking mechanism, the collected information is synchronized in time and spatially associated, and then standardized preprocessing is performed to obtain a multimodal data set.

3. The self-adaptive operation optimization method for a mine car according to claim 1, characterized in that: Get real-time minecart movement status, including: Based on the mine car's current position estimation and motion control input, a state prediction model is constructed to obtain the predicted state; Construct observation equations for various sensors that collect multimodal data sets and obtain measurement data from each sensor; Calculate the measurement residual based on the difference between the measurement data and the predicted state of each sensor; Dynamically adjust the Kalman gain matrix based on measurement residuals and sensor type: When the measurement residuals continuously exceed the preset residual threshold range or the sensor signal quality index falls below the set confidence level, the weight of the data information collected by the sensor in the fusion process is automatically reduced, or the preset redundant sensor channel is switched; The state prediction model is updated based on the dynamically adjusted Kalman gain matrix.

4. The self-adaptive operation optimization method for a mine car according to claim 3, characterized in that: The update equation of the state prediction model is: Where K k is the Kalman gain matrix, P k∣k-1 is the state prediction covariance matrix, H k is the observation matrix, R k is the measurement noise covariance matrix, r k is the measurement residual, They are the predicted states of the k-th and k-1-th mine car motion states respectively.

5. The self-adaptive operation optimization method for a mine car according to claim 1, characterized in that: Output minecart control instructions, including: Collect the operating environment data of the mine car, including track flatness, obstacle information, and mine car running speed, as input for fuzzy logic reasoning decision-making; The input quantity of fuzzy logic reasoning decision is divided into categories, and each category is logically distinguished based on the measured value relative to the set threshold range; The input variables are fuzzy processed and triangular membership function with dynamically adjustable parameters is used for modeling. The triangular membership function parameters of each input variable are dynamically adjusted. The adjustable parameters of the triangular membership function are corrected online based on the statistical results of the real-time operation data of the mine car. In the inference stage, according to the current input, the membership distribution results of the modified triangular membership function are comprehensively considered, and the confidence weighted centroid method is used for defuzzification to output continuous mine car control instructions.

6. The self-adaptive operation optimization method for a mine car according to claim 5, characterized in that: Adjustable parameters for online correction of triangular membership functions include: Based on the running status of the mine car collected during operation, the movement status evaluation indicators are set, including track running smoothness, obstacle avoidance success rate and speed control accuracy evaluation indicators; According to the results of each driving operation, based on the motion state evaluation index, the experience score of each fuzzy inference rule is calculated; The fuzzy inference rules are sorted according to their empirical scores within the statistical period. Inefficient fuzzy inference rules with scores lower than the set evaluation threshold are marked as candidates for elimination. According to the elimination strategy, the inefficient fuzzy inference rules are deleted, and according to the real-time collected data characteristics and system optimization goals, new fuzzy inference rules are automatically introduced to correct the adjustable parameters of the triangular membership function online.

7. The self-adaptive operation optimization method for a mine car according to claim 1, characterized in that: The generation of the minecart control instructions for the next cycle includes: Real-time acquisition of track surface changes, obstacle distance changes, and mine car speed changes; and calculation of the normalized deviation component of each change at each moment to construct a weighted comprehensive deviation; According to the weighted comprehensive deviation and each individual deviation, the dominant deviation type is determined by using the deviation dominant judgment function; Based on the identified dominant deviation type, the fuzzy logic reasoning decision is dynamically adjusted according to the following strategies: If the dominant deviation type is track deviation, the center position and span parameters of the track flatness membership function are adjusted first. If the dominant deviation type is obstacle deviation, the obstacle-related fuzzy inference rules are reconstructed first. If the dominant deviation type is speed deviation, the triangular membership function of the speed input and the speed control output rules are adjusted first. And when executing the adjustment of fuzzy logic reasoning decision, an adjustment increment is introduced; the updated fuzzy reasoning rule set and the adjustable parameters of the triangular membership function are applied in real time to the generation of the next cycle of mine car control instructions.

8. The self-adaptive operation optimization method for a mine car according to claim 7, characterized in that: Record the mine car control instructions for the next cycle, corresponding to the adjustment effect indicators of the mine car operation; based on the historical data of the adjustment effect indicators accumulated during long-term operation, dynamically optimize the weights of each deviation in the weighted comprehensive deviation and the learning rate of the adjustment increment to achieve self-learning and adaptive evolution of the feedback mechanism.

9. The self-adaptive operation optimization method for a mine car according to claim 7, characterized in that: The dominant deviation type is obtained by the following formula: M(t)=arg max{w s |Δs(t)|,w o |Δo(t)|,w v |Δv(t)|}; Where M(t) is the dominant deviation type, w s 、w o 、w v are the deviation weights corresponding to the track surface change, obstacle distance change, and minecart speed change, respectively. Δs(t), Δo(t), and Δv(t) are the track surface change, obstacle distance change, and minecart speed change, respectively. arg max is a function.

10. An adaptive vehicle operation optimization system suitable for a mine vehicle, characterized in that: The method for adaptive operation optimization of a mine vehicle according to any one of claims 1 to 9 comprises: Multimodal data module, used to obtain multimodal datasets of mine cars; The vehicle operation model module is used to establish a mine car kinematic model based on the extended Kalman filter and combined with the abnormal fault tolerance processing of the multimodal data set. The multimodal data set is used as the output of the mine car kinematic model to output the real-time mine car motion status. The control instruction module is used to perceive the motion environment and make fuzzy logic reasoning decisions based on dynamic scoring optimization according to the motion state of the mine car, and output the mine car control instructions; The feedback module is used to implement a dynamic feedback optimization mechanism of fuzzy logic reasoning decision-making based on error pattern recognition when the mine car is operated based on the mine car control instructions, and is applied to the generation of the mine car control instructions for the next cycle.

Images