A method for evaluating the hardware reliability of a fire reconnaissance robot system

Abstract

This invention discloses a method for evaluating the hardware reliability of a fire reconnaissance robot system. The steps include: collecting hardware performance data and key environmental stress data of the fire reconnaissance robot to construct basic evaluation data; extracting core evaluation factors that evolve with changes in fire conditions and determining reliability index data based on these core evaluation factors; determining subjective and objective weights based on real-time reliability index data under the current environment; dynamically adjusting the subjective and objective weights of the fire reconnaissance robot's indicators to obtain an adaptive comprehensive weight vector; and constructing a fuzzy relation matrix using the core evaluation factors and the comprehensive weight vector, and performing a synthesis operation to obtain a real-time updated system reliability level. This reliability evaluation method can provide real-time, dynamic, and accurate evaluation of the hardware reliability of fire reconnaissance robots, offering intelligent decision support for the robot's dynamic task decision-making and on-site maintenance.

Description

Technical Field

[0001] This invention relates to the field of robot reliability and safety evaluation technology, specifically to a method for evaluating the hardware reliability of a fire reconnaissance robot system. Background Technology

[0002] Firefighting reconnaissance robots, as an important branch of special rescue robots, are widely used in tasks such as fire fighting, hazardous environment detection, and personnel search and rescue. Due to the extreme characteristics of fire scenes, such as high temperatures, dense smoke, variable obstacles, and strong communication interference, robots often face risks of hardware performance degradation, system failure, and even damage during mission execution. Therefore, real-time, dynamic, and reliable hardware system status evaluation and risk warning for robots are of great significance for improving mission success rates and ensuring the safety of equipment and personnel.

[0003] Currently, there are still shortcomings in the reliability evaluation of such robots: (1) In terms of the construction of the evaluation system, the indicator dimensions of the existing methods are usually statically set, while the fire environment is highly time-varying and uncertain. Such indicator systems are difficult to dynamically match the reliability evolution process dominated by the environment, and the evaluation effectiveness is easily limited in complex fire environments; (2) In terms of weight allocation strategy, the existing methods rely more on expert prior knowledge. Its subjectivity may lead to the weight allocation being out of touch with the robot's real-time operating status and data characteristics. The weight adjustment mechanism based on real-time data distribution needs to be improved; (3) In terms of model real-time performance and adaptability, many reliability evaluation models are offline or batch processing modes. In the face of scenarios where fire information is updated rapidly and decision-making needs are urgent, the adaptive adjustment strategy of the existing methods still has room for improvement.

[0004] Therefore, there is an urgent need for a solution that can comprehensively address the above issues and provide a more accurate and reliable safety assessment of the hardware reliability of fire reconnaissance robots. Summary of the Invention

[0005] The purpose of this invention is to provide a method for evaluating the hardware reliability of a fire reconnaissance robot system, which can make a more accurate and reliable safety assessment of the hardware reliability of the fire reconnaissance robot.

[0006] Technical solution: The hardware reliability evaluation method for the fire reconnaissance robot system of the present invention includes the following steps:

[0007] Step 1: Collect hardware performance data of the fire reconnaissance robot in real time and key environmental stress data of the environment in which it is located, and use the hardware performance data and key environmental stress data to construct basic evaluation data.

[0008] Step 2: Extract the core evaluation factors that evolve with changes in fire conditions from the basic evaluation data, and determine the real-time reliability index data under the current environment based on the core evaluation factors.

[0009] Step 3: Determine the subjective and objective weights of the fire reconnaissance robot indicators based on real-time reliability index data under the current environment.

[0010] Step 4: Dynamically adjust the subjective and objective weights of the fire reconnaissance robot indicators based on real-time environmental data to obtain an adaptive comprehensive weight vector.

[0011] Step 5: Construct a fuzzy relation matrix using the core evaluation factors and the comprehensive weight vector, and perform a synthesis operation to obtain the real-time updated system reliability level.

[0012] Furthermore, in step 1, the hardware performance data of the body state includes environmental tolerance data, task execution capability data, and sensing and communication capability data; the key environmental stress data of the environment includes thermal environmental stress data, visibility and smoke stress data, chemical and gaseous environmental stress data, and structural and non-uniform field stress data.

[0013] Furthermore, in step 2, the specific steps for extracting the core evaluation factors that evolve with changes in fire conditions from the basic evaluation data are as follows:

[0014] Step 2.1: At the current evaluation time t, collect the basic evaluation data obtained in the most recent N sampling periods. The basic evaluation data obtained in each sampling period is a sample vector of m index values. Then, arrange the sample vectors of the N sampling periods in row-wise order according to the sampling time to form a real-time data matrix X. t for: In the formula, x ij This is the sample vector of the j-th index value in the i-th sampling period;

[0015] Step 2.2, for the real-time data matrix X t After standardizing each element, the standardized matrix Z is obtained. t Standardized matrix Z t Each element is: In the formula, For real-time data matrix X t The j-th column vector x j =[x 1j ,x 1j ,…,x Nj ] T mean S j For real-time data matrix X t The j-th column vector xj =[x 1j ,x 1j ,…,x Nj ] T standard deviation and S j The calculation formulas are as follows: , ;

[0016] Step 2.3, based on the normalized matrix Z t Calculate the covariance matrix R among the sample vectors of m index values. (t) for: In the formula, R (t) Z is an m×m matrix. t T For the normalized matrix Z t The transpose of the matrix, and then the covariance matrix R (t) Perform eigenvalue decomposition to obtain the covariance matrix R. (t) The decomposition formula for all eigenvalues ​​and their corresponding unit eigenvectors is: In the formula, λ q Let u represent the q-th non-zero eigenvalue. Sort all non-zero eigenvalues ​​in descending order of their numerical values. q The unit eigenvectors corresponding to the non-zero eigenvalues;

[0017] For the q-th non-zero eigenvalue λ q The corresponding principal component contribution rate w q (t) The cumulative contribution rate ρ of the first q principal components is defined as the proportion of the non-zero eigenvalue in the sum of all non-zero eigenvalues. q (t) The ratio of the cumulative sum of the first q non-zero eigenvalues ​​after permutation to the sum of all non-zero eigenvalues ​​is expressed by the following formulas: , In the formula, Let j be the j-th non-zero eigenvalue;

[0018] Step 2.4: Set the cumulative contribution rate threshold to 85%, and find the value that makes ρ... q (t) The smallest integer k that is true with ≥85% probability t The integer is then used to determine the number of core evaluation factors to be retained at the current evaluation time t. The non-zero eigenvalues ​​are then sorted in descending order from largest to smallest, and the top k are selected. t The unit eigenvector corresponding to the largest eigenvalue is used as the core evaluation factor, and a set of eigenvectors is constructed based on this.

[0019] Furthermore, step 2, extracting the core evaluation factors that evolve with changes in fire conditions from the basic evaluation data, also includes the following steps:

[0020] Step 2.6, calculate the mean cosine similarity change Δcos in the principal component directions as follows: In the formula, V i (t) V is the principal component direction matrix at the current evaluation time t. t (k) The i-th principal component direction vector, V i (t-1) V is the principal component direction matrix at the previous evaluation time t-1. t-1 (k) The direction vector of the i-th principal component;

[0021] Step 2.7: Set the update threshold γ. If Δcos > γ or the rate of change of any core evaluation factor exceeds the corresponding safety threshold, return to step 2.1.

[0022] Furthermore, in step 3, the specific steps for determining the subjective weights of the fire reconnaissance robot indicators based on real-time reliability index data under the current environment are as follows:

[0023] Step 3.1, obtain k for the current evaluation time t. t The core evaluation factors are scaled using a 1-9 method to form a judgment matrix for the current evaluation time t. for: In the formula, b ij (t) This represents the importance scale of the i-th core evaluation factor relative to the j-th core evaluation factor at the current evaluation time t.

[0024] Step 3.2, for the judgment matrix Each row is normalized to obtain the normalized value for each row: In the formula, The importance scale b in the row ij (t) The normalized value is then summed across all rows after normalization to obtain the sum eigenvector of each row: In the formula, The sum of the eigenvectors for each row;

[0025] Step 3.3: Normalize the obtained rows and eigenvectors to obtain the judgment matrix. The feature vectors are: In the formula, W s (t) To determine the matrix The feature vectors are the subjective weights.

[0026] Furthermore, step 3, determining the subjective weights of the fire reconnaissance robot indicators, also includes the following steps:

[0027] Step 3.4, calculate the consistency index. The specific calculation formula is as follows: , In the formula, To determine the matrix Consistency indicators To determine the matrix The consistency ratio, λ max (t) To determine the matrix The largest eigenvalue, RI(k) t ) is the average random consistency index. If CR (t) If <0.10, then accept the judgment matrix. Consistency, if CR (t) If ≥0.10, then for the judgment matrix Adjustments were made to make CR (t) <0.10.

[0028] Furthermore, in step 3, the specific steps for determining the objective weights of the fire reconnaissance robot's indicators based on real-time reliability index data under the current environment are as follows:

[0029] Step 3.5, calculate the principal component score matrix F t The specific calculation formula is as follows: The principal component score matrix F t As a data source for objective weighting, the number of core evaluation factors k t The number of principal components selected at the current evaluation time t;

[0030] Step 3.6, analyze the principal component score matrix F. t To perform nonnegation processing on the negative values ​​present in the data, let: In the formula, f ij (t) Principal Component Score Matrix F t elements, Let be the maximum value of all elements in the i-th row. Let f' be the minimum value of all elements in the i-th row, and let α be a positive number used to ensure that all f'' are true. ij (t) If the value is greater than 0, the processed matrix is ​​denoted as matrix F. t ';

[0031] Step 3.7: Calculate the index contribution p of the i-th sample value under the j-th principal component. ij (t) for: In the formula, Let matrix F t The element in the i-th row and j-th column;

[0032] Step 3.8, calculate the information entropy of the j-th principal component. for: The above formula shows that the smaller the entropy value, the greater the difference in the observed values ​​of each sample under this factor, and the more information it contains.

[0033] Step 3.9: Calculate the information utility value, i.e., the coefficient of difference. for: ;

[0034] Step 3.10: Normalize the difference coefficients to obtain the objective weight vector: , In the formula, Let the objective weight of the j-th principal component be . Let be the objective weight at time t. These are the weight values ​​corresponding to each principal component.

[0035] Furthermore, in step 4, the specific steps for dynamically adjusting the subjective and objective weights of the fire reconnaissance robot's indicators based on real-time environmental data to obtain an adaptive comprehensive weight vector are as follows:

[0036] Step 4.1, construct the linear combination W' of subjective and objective weights as follows: In the formula, α s and α o Let W be the coefficients of the linear combination to be optimized. s (t) ) T Let W be the transpose of the subjective weights. o (t) ) T This is the transpose of the objective weights;

[0037] Step 4.2, for the linear combination coefficients α s and α o Optimization is performed by minimizing the deviation in the objective function to find the optimal linearity of the weights. In the formula, Let L be the L2 norm of the vector, i.e., the Euclidean distance. This indicates minimizing the sum of the distances;

[0038] Step 4.3, based on the properties of matrix differentiation, convert the linear combination coefficients α s and α o The optimization problem is transformed into a system of linear differential equations with first-order derivative conditions for optimization: In the formula, w s and w oSubjective weights W s (t) and objective weight W o (t) Weighting adjustment factor;

[0039] Step 4.4: Solve the system of linear differential equations to obtain the optimal linear combination coefficients α. s 'and α o ';

[0040] Step 4.5, for the linear combination coefficients α s 'and α o Normalization is performed to obtain the comprehensive weight vector W based on game theory weighting. c for: In the formula, These are the normalized subjective weight coefficients. These are the normalized objective weighting coefficients.

[0041] Furthermore, step 4, when dynamically adjusting to obtain the adaptive comprehensive weight vector, also includes the following steps:

[0042] Step 4.6, for each core evaluation factor d (d=1,2,…,k) t Based on the sensitivity of the basic assessment data associated with the core assessment factors to environmental risks, a risk impact factor r is defined for each factor. d (t) An environmental risk vector is constructed using the risk impact factors of all core evaluation factors. for: In the formula, These are the risk impact factor values ​​of each core evaluation factor at time t, with an initial value of 0;

[0043] Step 4.7, using environmental risk vectors For the comprehensive weight vector W c Dynamic correction is performed, and the correction formula is as follows: In the formula, To determine the corrected weight value of the j-th factor, w cj It is W c The j-th component, β is the global risk sensitivity coefficient, and 0≤β≤1, is used to control the strength of risk correction;

[0044] Step 4.8, adjust the corrected comprehensive weight vector W. c Normalization is performed to generate the final comprehensive weight vector W for the fuzzy comprehensive evaluation at the current evaluation time t. T for: , In the formula, This represents the final weight of the j-th factor after normalization.

[0045] Furthermore, in step 5, the specific steps for constructing a fuzzy relation matrix using the core evaluation factors and the comprehensive weight vector, and performing a synthesis operation to obtain the real-time updated system reliability level are as follows:

[0046] Step 5.1, using k t The core evaluation factors constitute the dynamic factor set U at the current evaluation time t. t for: In the formula, Each of these represents a dynamic factor, and each represents a dynamically refined dimension of reliability impact.

[0047] Step 5.2: Define a comment set V containing 5 evaluation levels, and assign a standard score to each evaluation level to form a score vector V. s for: , In the formula, Comments for each rating level. These correspond to the five rating levels mentioned above;

[0048] Step 5.3, for dynamic factor u i (t) and rating level v s Let the corresponding trapezoidal membership function parameter range be [a j ,b j ,c j ], of which it belongs entirely to the evaluation level v j The interval is [b j ,c j ], between rating level v j-1 With v j The transition interval between them is [a j ,b j ], between rating level v j With v j+1 The transition interval between them is [c j ,d j For dynamic factors u i (t) and rating level v s The trapezoidal membership function is used to calculate the membership degree r of the factor to that level. ij (t) The calculation formula is: ;

[0049] Step 5.4, based on k t The membership degrees of each dynamic factor to the five evaluation levels are used to construct the fuzzy relation matrix R at the current evaluation time t. t for: ;

[0050] Step 5.5, using the final comprehensive weight vector W at the current evaluation time t. T With fuzzy relation matrix R t Fuzzy synthesis operations are performed, using a weighted average synthesis operator, to obtain the comprehensive membership vector B for each evaluation level. t for: In the formula, " " indicates fuzzy composition operation;

[0051] Step 5.6: Assign scores to the comment set V, and calculate the final score S as follows: ;

[0052] Step 5.7: Obtain the comprehensive reliability evaluation score based on the final score S, and then obtain the corresponding reliability level through the preset level standard.

[0053] Compared with the prior art, the beneficial effects of this invention are:

[0054] (1) The evaluation mode has been changed from "static" to "dynamic", which improves the timeliness and environmental adaptability of the evaluation. This invention introduces a sliding time window mechanism, real-time environmental data coupling and update trigger conditions, so that the entire evaluation system can run online and be updated in real time. The core evaluation factors, index weights and final results all evolve dynamically with the current environment and robot state, ensuring that the safety assessment conclusions are always synchronized with the real-time and most important threats faced by the robot. (2) An evaluation system with deep coupling of "environment-state" is constructed, which makes the reliability assessment more scientific and targeted. Key environmental stress indicators (such as temperature and smoke concentration) are integrated in the indicator system construction stage, and a direct data correlation between the environment and hardware reliability is established. This design enables the subsequent dynamic principal component analysis to automatically identify the hardware performance variation pattern driven by specific environmental stress (such as high temperature), thereby extracting the core evaluation factors that are truly representative of the scenario. (3) A dynamic fusion and adaptive adjustment mechanism of subjective and objective weights is proposed to make the weight allocation more reasonable. This invention introduces a real-time environmental risk perception module, which dynamically corrects the final combined weights based on whether the environmental monitoring data exceeds the threshold. This means that when a certain environmental risk (such as extreme high temperature) becomes prominent, the weight of the related hardware reliability indicators will automatically increase, thereby enabling the evaluation focus to quickly focus on the most pressing risk point, realizing intelligent and adaptive weight allocation. (4) It provides a comprehensive evaluation output with diagnostic and early warning capabilities. This invention not only outputs real-time reliability levels and scores through dynamic fuzzy comprehensive evaluation, but also identifies key vulnerable links with "high weight and low membership" and rapid performance degradation points with "sudden drop in membership". The output results are upgraded from a single "state description" to an integrated decision support information package that includes "state assessment, root cause location, and trend prediction". Attached Figure Description

[0055] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0056] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the embodiments described.

[0057] like Figure 1 As shown, the hardware reliability evaluation method for the fire reconnaissance robot system disclosed in this invention includes the following steps:

[0058] Step 1: Collect hardware performance data of the fire reconnaissance robot in real time and key environmental stress data of the environment in which it is located, and use the hardware performance data and key environmental stress data to construct basic evaluation data.

[0059] Step 2: Extract the core evaluation factors that evolve with changes in fire conditions from the basic evaluation data, and determine the real-time reliability index data under the current environment based on the core evaluation factors.

[0060] Step 3: Determine the subjective and objective weights of the fire reconnaissance robot indicators based on real-time reliability index data under the current environment.

[0061] Step 4: Dynamically adjust the subjective and objective weights of the fire reconnaissance robot indicators based on real-time environmental data to obtain an adaptive comprehensive weight vector.

[0062] Step 5: Construct a fuzzy relation matrix using the core evaluation factors and the comprehensive weight vector, and perform a synthesis operation to obtain the real-time updated system reliability level.

[0063] Furthermore, in step 1, the hardware performance data of the body state includes environmental tolerance data, task execution capability data, and sensing and communication capability data; the key environmental stress data of the environment includes thermal environmental stress data, visibility and smoke stress data, chemical and gaseous environmental stress data, and structural and non-uniform field stress data.

[0064] For ground robots, environmental tolerance data includes operating time, explosion-proof performance, dust and water resistance, impact resistance, and overall fire resistance and heat insulation; task execution capability data includes walking speed, obstacle crossing height, drag force, load capacity, and climbing gradient; and perception and communication capability data includes temperature measurement range, detection distance, path planning and autonomous obstacle avoidance capabilities, communication establishment time, communication distance, and communication link latency.

[0065] For unmanned aerial vehicles (flying robots), environmental tolerance data includes endurance, wind and disturbance resistance, and structural integrity; mission execution capability data includes maximum level flight speed, hovering accuracy, minimum obstacle avoidance distance, and payload operation capability; and perception and communication capability data includes temperature measurement range, ranging range, autonomous positioning accuracy, path planning and autonomous obstacle avoidance capability, communication establishment time, communication distance, and communication link latency.

[0066] Thermal environment stress data includes instantaneous temperature field, temperature gradient, thermal radiation flux, and thermal contact risk; visibility and smoke stress data includes smoke concentration (optical density), visibility level, and smoke deposition rate; chemical and gaseous environment stress data includes toxic and harmful gas concentration, oxygen concentration, and air humidity (high temperature and high humidity); structural and non-uniform field stress data includes ground (space) passability, airflow turbulence intensity, and noise level.

[0067] By establishing a "performance-environmental stress" mapping, the key environmental stress data of the environment is used as modulation variables to dynamically correct the theoretical value of the hardware performance data of the robot's state, thereby obtaining its actual usable value in the current harsh environment (such as "actual effective detection distance" and "dynamic remaining endurance"). This forms an evaluation data basis that can truly reflect "what the robot can do in a specific fire scene". The following table is an example.

[0068] Table 1 shows the integration relationship between some "performance-environmental stress" parameters.

[0069] Status indicators Associated environmental stress Relevant evaluation logic Overall fire resistance and heat insulation Instantaneous temperature field; thermal radiation flux <![CDATA[When the ambient temperature > T1 (a certain upper temperature limit), active cooling is triggered and the battery life is reduced; when the temperature > T2 (the material critical point), the durability accelerates to deteriorate.]]> Dustproof and waterproof performance Smoke and dust settling rate; air humidity High dust settling rates may clog ventilation holes and joints; high temperature and humidity environments test sealing performance. The actual effectiveness of the protection level decreases. Walking speed; obstacle crossing height Ground passability index The expected "travel speed" and obstacle-crossing success rate are dynamically adjusted based on obstacle density and height. Hovering accuracy (drones) airflow turbulence intensity Strong turbulence causes positioning drift. Actual hovering accuracy = nominal accuracy + variance of error caused by turbulence. Communication distance Smoke Concentration and Visibility Level Smoke and walls have an attenuating effect on wireless signals.

[0070] Furthermore, in step 2, the specific steps for extracting the core evaluation factors that evolve with changes in fire conditions from the basic evaluation data are as follows:

[0071] Step 2.1: At the current evaluation time t, collect the basic evaluation data obtained in the most recent N sampling periods. The basic evaluation data obtained in each sampling period is a sample vector of m index values. Then, arrange the sample vectors of the N sampling periods in row-wise order according to the sampling time to form a real-time data matrix X. t for: In the formula, x ij This is the sample vector of the j-th index value in the i-th sampling period;

[0072] Step 2.2, for the real-time data matrix X t After standardizing each element, the standardized matrix Z is obtained. t Standardized matrix Z t Each element is: In the formula, For real-time data matrix X t The j-th column vector x j =[x 1j ,x 1j ,…,x Nj ] T mean S j For real-time data matrix X t The j-th column vector x j =[x 1j ,x 1j ,…,x Nj ] T standard deviation and S j The calculation formulas are as follows: , ;

[0073] Step 2.3, based on the normalized matrix Z tCalculate the covariance matrix R among the sample vectors of m index values. (t) for: In the formula, R (t) Z is an m×m matrix. t T For the normalized matrix Z t The transpose of the matrix, and then the covariance matrix R (t) Perform eigenvalue decomposition to obtain the covariance matrix R. (t) The decomposition formula for all eigenvalues ​​and their corresponding unit eigenvectors is: In the formula, λ q Let u represent the q-th non-zero eigenvalue. Sort all non-zero eigenvalues ​​in descending order of their numerical values. q Let λ be the unit eigenvector corresponding to the non-zero eigenvalues, with dimension m×1; for the q-th non-zero eigenvalue λ q The corresponding principal component contribution rate w q (t) The cumulative contribution rate ρ of the first q principal components is defined as the proportion of the non-zero eigenvalue in the sum of all non-zero eigenvalues. q (t) The ratio of the cumulative sum of the first q non-zero eigenvalues ​​after permutation to the sum of all non-zero eigenvalues ​​is expressed by the following formulas: , In the formula, Let j be the j-th non-zero eigenvalue;

[0074] Step 2.4: Set the cumulative contribution rate threshold to 85%, and find the value that makes ρ... q (t) The smallest integer k that is true with ≥85% probability t This integer is then used to determine the number of core evaluation factors to be retained at the current evaluation time t. Finally, the top k non-zero eigenvalues, arranged in descending order, are... t The unit eigenvectors corresponding to the non-zero eigenvalues ​​are used as core evaluation factors. The eigenvector set is constructed using each core evaluation factor, and each column in the set is the eigenvector corresponding to a core evaluation factor.

[0075] Furthermore, when determining the real-time reliability index data under the current environment based on the core evaluation factors, the top k... t The eigenvectors corresponding to the core evaluation factors constitute the principal component direction matrix V. t (k) And the principal component direction matrix V t (k) As the core evaluation factor loading matrix at the current evaluation time t, the standardized matrix Z t Projected onto the principal component direction matrix V t (k)Then, by analyzing the principal component direction matrix V... t (k) The size of the elements in each column of the feature vector is used to identify the basic evaluation data that contributes the most to each core evaluation factor, which serves as the reliability index data affecting the system reliability under the current environment.

[0076] Furthermore, step 2, extracting the core evaluation factors that evolve with changes in fire conditions from the basic evaluation data, also includes the following steps:

[0077] Step 2.6, calculate the mean cosine similarity change Δcos in the principal component directions as follows: In the formula, V i (t) V is the principal component direction matrix at the current evaluation time t. t (k) The i-th principal component direction vector, V i (t-1) V is the principal component direction matrix at the previous evaluation time t-1. t-1 (k) The direction vector of the i-th principal component;

[0078] Step 2.7: Set the update threshold γ. If Δcos > γ or the rate of change of any core evaluation factor exceeds the corresponding safety threshold, it is determined that the system state or environment has changed significantly. Immediately return to step 2.1 to trigger a new dynamic PCA calculation and update the core evaluation factor set.

[0079] Furthermore, in step 3, the specific steps for determining the subjective weights of the fire reconnaissance robot indicators based on real-time reliability index data under the current environment are as follows:

[0080] Step 3.1, obtain k for the current evaluation time t. t The core evaluation factors are scaled using a 1-9 method to form a judgment matrix for the current evaluation time t. for: In the formula, b ij (t) To determine each element in the matrix, we define the importance scale of the i-th core evaluation factor relative to the j-th core evaluation factor at the current evaluation time t.

[0081] Step 3.2, for the judgment matrix Each row is normalized to obtain the normalized value for each row: In the formula, The importance scale b in the row ij (t) The normalized value is then summed across all rows after normalization to obtain the sum eigenvector of each row: In the formula, The sum of the eigenvectors for each row;

[0082] Step 3.3: Normalize the obtained rows and eigenvectors to obtain the judgment matrix. The feature vectors are: In the formula, W s (t) To determine the matrix The feature vectors are the subjective weights.

[0083] Furthermore, step 3, determining the subjective weights of the fire reconnaissance robot indicators, also includes the following steps:

[0084] Step 3.4, calculate the consistency index. The specific calculation formula is as follows: , In the formula, To determine the matrix Consistency indicators To determine the matrix The consistency ratio, λ max (t) To determine the matrix The largest eigenvalue, RI(k) t ) is the average random consistency index. If CR (t) If the value is less than 0.10, then the judgment matrix is ​​accepted. Consistency; if CR (t) If ≥0.10, then for the judgment matrix Adjustments were made to make CR (t) <0.10.

[0085] During adjustment, the judgment matrix is ​​solved first. The normalized eigenvector corresponding to the largest eigenvalue : This normalized eigenvector is the initial weight vector for each reliability index to be evaluated, obtained through calculation. Assuming the initial weight for the i-th indicator, the theoretical weight ratio is then calculated. Then calculate the judgment matrix. Each element b in ij (t) The absolute deviation from the corresponding theoretical weight ratio Based on this, select the element b with the largest absolute deviation. ij (t) Perform targeted fine-tuning to bring its scale value closer to the theoretical ratio. Approaching, then using the adjusted new judgment matrix. Recalculate CR (t) Repeat the above process until CR (t)<0.10.

[0086] Furthermore, in step 3, the specific steps for determining the objective weights of the fire reconnaissance robot's indicators based on real-time reliability index data under the current environment are as follows:

[0087] Step 3.5, calculate the principal component score matrix F t The specific calculation formula is as follows: The principal component score matrix F t As a data source for objective weighting, the number of core evaluation factors k t The number of principal components selected at the current evaluation time t;

[0088] Step 3.6, analyze the principal component score matrix F. t To perform nonnegation processing on the negative values ​​present in the data, let: In the formula, f ij (t) Principal Component Score Matrix F t elements, Let be the maximum value of all elements in the i-th row. Let f' be the minimum value of all elements in the i-th row, and let α be a positive number used to ensure that all f'' are true. ij (t) If the result is greater than 0, the processed matrix is ​​denoted as matrix Ft'.

[0089] Step 3.7: Calculate the index contribution p of the i-th sample value under the j-th principal component. ij (t) for: In the formula, Let be the element in the i-th row and j-th column of matrix Ft';

[0090] Step 3.8, calculate the information entropy of the j-th principal component. for: This indicates that the smaller the entropy value, the greater the difference in the observed values ​​of each sample under this factor, and the more information it contains;

[0091] Step 3.9: Calculate the information utility value, i.e., the coefficient of difference. for: ;

[0092] Step 3.10: Normalize the difference coefficients to obtain the objective weight vector: , In the formula, Let the objective weight of the j-th principal component be . Let be the objective weight at time t. These are the weight values ​​corresponding to each principal component.

[0093] Furthermore, in step 4, the specific steps for dynamically adjusting the subjective and objective weights of the fire reconnaissance robot's indicators based on real-time environmental data to obtain an adaptive comprehensive weight vector are as follows:

[0094] Step 4.1, construct the linear combination W' of subjective and objective weights as follows: In the formula, α s and α o Let W be the coefficients of the linear combination to be optimized. s (t) ) T Let W be the transpose of the subjective weights. o (t) ) T This is the transpose of the objective weights;

[0095] Step 4.2: The optimal strategy should minimize the overall difference between the combined weights and the basic weights. For the linear combination coefficient α... s and α o Optimization is performed by minimizing the deviation in the objective function to find the optimal linearity of the weights. In the formula, Let L be the L2 norm of the vector, i.e., the Euclidean distance. This indicates minimizing the sum of the distances;

[0096] Step 4.3, based on the properties of matrix differentiation, convert the linear combination coefficients α s and α o The optimization problem is transformed into a system of linear differential equations with first-order derivative conditions for optimization: In the formula, w s and w o Subjective weights W s (t) and objective weight W o (t) Weighting adjustment factor;

[0097] Step 4.4: Solve the system of linear differential equations to obtain the optimal linear combination coefficients α. s 'and α o ';

[0098] Step 4.5: To meet the normalization requirement of the weight vector, adjust the linear combination coefficients α... s 'and α o Normalization is performed to obtain the comprehensive weight vector W based on game theory weighting. c for: In the formula, These are the normalized subjective weight coefficients. These are the normalized objective weighting coefficients.

[0099] Furthermore, step 4, when dynamically adjusting to obtain the adaptive comprehensive weight vector, also includes the following steps:

[0100] Step 4.6, for each core evaluation factor d (d=1,2,…,k) t Based on the sensitivity of the basic assessment data associated with the core assessment factors to environmental risks, a risk impact factor r is defined for each factor. d (t) For example, real-time reading of environmental monitoring data (such as temperature Tt, smoke concentration St, toxic gas concentration Gt, etc.) at the current evaluation time t; if the current Tt exceeds the threshold T... th Then, the r of all indicators directly related to heat resistance (such as motor temperature, circuit board temperature) j (t) The value will increase according to preset rules, and an environmental risk vector will be constructed using the risk impact factors of all core evaluation factors. for: In the formula, These are the risk impact factor values ​​of each core evaluation factor at evaluation time t, with an initial value of 0;

[0101] Step 4.7, using environmental risk vectors For the comprehensive weight vector W c Dynamic correction is performed, and the correction formula is as follows: In the formula, To determine the corrected weight value of the j-th factor, w cj It is W c The j-th component, β, is the global risk sensitivity coefficient, where 0 ≤ β ≤ 1, used to control the intensity of risk correction. It can be adjusted according to the overall risk level of the task. After correcting all indicators, the unnormalized correction weight vector W is obtained. t ';

[0102] Step 4.8, adjust the corrected comprehensive weight vector W. c Normalization is performed to generate the final comprehensive weight vector W for the fuzzy comprehensive evaluation at the current evaluation time t. T for: , In the formula, This represents the final weight of the j-th factor after normalization.

[0103] Furthermore, in step 5, the specific steps for constructing a fuzzy relation matrix using the core evaluation factors and the comprehensive weight vector, and performing a synthesis operation to obtain the real-time updated system reliability level are as follows:

[0104] Step 5.1, using k t The core evaluation factors constitute the dynamic factor set U at the current evaluation time t. tfor: In the formula, Each of these represents a dynamic factor, and each represents a dynamically refined dimension of reliability impact.

[0105] Step 5.2: Define a comment set V containing 5 evaluation levels, and assign a standard score to each evaluation level to form a score vector V. s for: , In the formula, Comments for each rating level. Each factor corresponds to one of five rating levels. Based on the physical meaning of the factor (e.g., "thermal stress factor," "sensory performance factor"), it is associated with a set of specific, measurable underlying hardware or environmental indicators. The f value is determined by querying a pre-defined "indicator value-reliability level" mapping knowledge base. i (t) The corresponding reliability level range. This knowledge base is built based on national standards, industry specifications, historical data, and expert experience, and is configured before the task according to the specific robot model and task type;

[0106] Step 5.3, for dynamic factor u i (t) and rating level v s Let the corresponding trapezoidal membership function parameter range be [a j ,b j ,c j ], of which it belongs entirely to the evaluation level v j The interval is [b j ,c j ], between rating level v j-1 With v j The transition interval between them is [a j ,b j ], between rating level v j With v j+1 The transition interval between them is [c j ,d j ];

[0107] For dynamic factors u i (t) and rating level v s The trapezoidal membership function is used to calculate the membership degree r of the factor to that level. ij (t) The calculation formula is: ;

[0108] Step 5.4, based on k t The membership degrees of each dynamic factor to the five evaluation levels are used to construct the fuzzy relation matrix R at the current evaluation time t. tfor: ;

[0109] Step 5.5, using the final comprehensive weight vector W at the current evaluation time t. T With fuzzy relation matrix R t Fuzzy synthesis operations are performed, using a weighted average synthesis operator, to obtain the comprehensive membership vector B for each evaluation level. t for: In the formula, " " indicates fuzzy composition operation;

[0110] Step 5.6: Assign scores to the comment set V, and calculate the final score S as follows: ;

[0111] Step 5.7: Obtain the comprehensive reliability evaluation score based on the final score S, and then obtain the corresponding reliability level through the preset grading standards. The correspondence between the preset scoring range and the reliability level is as follows:

[0112] Rating range grade [85,100) excellent [70,85) good [55,70) medium [40,55) Pass [0,40) Poor

[0113] Furthermore, after obtaining the real-time updated system reliability level, it also provides the location of critical vulnerable links and short-term risk warnings. The specific steps are as follows: First, identify those that meet the w i (t) >θ w And max j (r ij (t) )<θ r Factor u i (t) , where θ w θ r For a preset threshold, such factors represent key vulnerabilities that are considered important (high weight) but are currently in poor condition (low membership). Then, the difference between the membership degree of each factor to the "poor" level at the current moment and its short-term historical mean is calculated. If it exceeds the threshold, the subsystem corresponding to that factor is marked as having rapidly deteriorating performance.

[0114] As described above, although the invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the invention itself. Various changes in form and detail may be made without departing from the spirit and scope of the invention as defined in the appended claims.

Claims

1. A method for evaluating the hardware reliability of a fire reconnaissance robot system, characterized in that, Includes the following steps: Step 1: Collect hardware performance data of the fire reconnaissance robot in real time and key environmental stress data of the environment in which it is located, and use the hardware performance data and key environmental stress data to construct basic evaluation data. Step 2: Extract the core evaluation factors that evolve with changes in fire conditions from the basic evaluation data, and determine the real-time reliability index data under the current environment based on the core evaluation factors. Step 3: Determine the subjective and objective weights of the fire reconnaissance robot indicators based on real-time reliability index data under the current environment. Step 4: Dynamically adjust the subjective and objective weights of the fire reconnaissance robot indicators based on real-time environmental data to obtain an adaptive comprehensive weight vector. Step 5: Construct a fuzzy relation matrix using the core evaluation factors and the comprehensive weight vector, and perform a synthesis operation to obtain the real-time updated system reliability level; In step 3, the specific steps for determining the subjective weights of the fire reconnaissance robot's indicators based on real-time reliability index data under the current environment are as follows: Step 3.1, obtain k for the current evaluation time t. t The core evaluation factors are scaled using a 1-9 method to form a judgment matrix for the current evaluation time t. for: In the formula, b ij (t) This represents the importance scale of the i-th core evaluation factor relative to the j-th core evaluation factor at the current evaluation time t. Step 3.2, for the judgment matrix Each row is normalized to obtain the normalized value for each row: In the formula, The importance scale b in the row ij (t) The normalized value is then summed across all rows after normalization to obtain the sum eigenvector of each row: In the formula, The sum of the eigenvectors for each row; Step 3.3: Normalize the obtained rows and eigenvectors to obtain the judgment matrix. The feature vectors are: In the formula, W s (t) To determine the matrix The feature vectors are the subjective weights; In step 3, the specific steps for determining the objective weights of the fire reconnaissance robot's indicators based on real-time reliability index data under the current environment are as follows: Step 3.5, calculate the principal component score matrix F t The specific calculation formula is as follows: The principal component score matrix F t As a data source for objective weighting, the number of core evaluation factors k t The number of principal components selected at the current evaluation time t; Step 3.6, analyze the principal component score matrix F. t To perform nonnegation processing on the negative values ​​present in the data, let: In the formula, f ij (t) The elements of the principal component score matrix Ft, Let be the maximum value of all elements in the i-th row. Let f' be the minimum value of all elements in the i-th row, and let α be a positive number used to ensure that all f'' are true. ij (t) If the result is greater than 0, the processed matrix is ​​denoted as matrix Ft'. Step 3.7: Calculate the index contribution p of the i-th sample value under the j-th principal component. ij (t) for: In the formula, Let be the element in the i-th row and j-th column of matrix Ft'; Step 3.8, calculate the information entropy of the j-th principal component. for: The above formula shows that the smaller the entropy value, the greater the difference in the observed values ​​of each sample under this factor, and the more information it contains. Step 3.9: Calculate the information utility value, i.e., the coefficient of difference. for: Step 3.10: Normalize the difference coefficients to obtain the objective weight vector: , In the formula, Let the objective weight of the j-th principal component be . Let be the objective weight at time t. These are the weight values ​​corresponding to each principal component.

2. The method for evaluating the hardware reliability of a fire reconnaissance robot system according to claim 1, characterized in that, In step 1, the hardware performance data of the host body includes environmental tolerance data, task execution capability data, and sensing and communication capability data; the key environmental stress data of the environment includes thermal environmental stress data, visibility and smoke stress data, chemical and gaseous environmental stress data, and structural and non-uniform field stress data.

3. The method for evaluating the hardware reliability of a fire reconnaissance robot system according to claim 1, characterized in that, Step 2 involves extracting the core evaluation factors that evolve with changes in fire conditions from the basic evaluation data. The specific steps are as follows: Step 2.1: At the current evaluation time t, collect the basic evaluation data obtained in the most recent N sampling periods. The basic evaluation data obtained in each sampling period is a sample vector of m index values. Then, arrange the sample vectors of the N sampling periods in row-wise order according to the sampling time to form a real-time data matrix X. t for: In the formula, x ij This is the sample vector of the j-th index value in the i-th sampling period; Step 2.2, for the real-time data matrix X t After standardizing each element, the standardized matrix Z is obtained. t Standardized matrix Z t Each element is: In the formula, For real-time data matrix X t The j-th column vector x j =[x 1j ,x 1j ,…,x Nj ] T mean S j For real-time data matrix X t The j-th column vector x j =[x 1j ,x 1j ,…,x Nj ] T standard deviation and S j The calculation formulas are as follows: , Step 2.3, based on the normalized matrix Z t Calculate the covariance matrix R between the sample vectors of m index values. (t) for: In the formula, R (t) Z is an m×m matrix. t T For the normalized matrix Z t The transpose of the matrix, and then the covariance matrix R (t) Perform eigenvalue decomposition to obtain the covariance matrix R. (t) The decomposition formula for all eigenvalues ​​and their corresponding unit eigenvectors is: In the formula, λ q Let u represent the q-th non-zero eigenvalue. Sort all non-zero eigenvalues ​​in descending order of their numerical values. q The unit eigenvectors corresponding to the non-zero eigenvalues; For the q-th non-zero eigenvalue λ q The corresponding principal component contribution rate w q (t) The cumulative contribution rate ρ of the first q principal components is defined as the proportion of the non-zero eigenvalue in the sum of all non-zero eigenvalues. q (t) The ratio of the cumulative sum of the first q non-zero eigenvalues ​​after permutation to the sum of all non-zero eigenvalues ​​is expressed by the following formulas: , In the formula, Let j be the j-th non-zero eigenvalue; Step 2.4: Set the cumulative contribution rate threshold to 85%, and find the value that makes ρ... q (t) The smallest integer k that is true with ≥85% probability t The integer is then used to determine the number of core evaluation factors to be retained at the current evaluation time t. The non-zero eigenvalues ​​are then sorted in descending order from largest to smallest, and the top k are selected. t The unit eigenvector corresponding to the largest eigenvalue is used as the core evaluation factor, and a set of eigenvectors is constructed based on this.

4. The method for evaluating the hardware reliability of a fire reconnaissance robot system according to claim 3, characterized in that, Step 2, extracting the core evaluation factors that evolve with changes in fire conditions from the basic evaluation data, also includes the following steps: Step 2.5, calculate the mean cosine similarity change Δcos in the principal component directions as follows: In the formula, V i (t) V is the principal component direction matrix at the current evaluation time t. t (k) The i-th principal component direction vector, V i (t-1) V is the principal component direction matrix at the previous evaluation time t-1. t-1 (k) The direction vector of the i-th principal component; Step 2.6: Set the update threshold γ. If Δcos > γ or the rate of change of any core evaluation factor exceeds the corresponding safety threshold, return to step 2.

1.

5. The method for evaluating the hardware reliability of a fire reconnaissance robot system according to claim 1, characterized in that, Step 3, determining the subjective weights of the fire reconnaissance robot indicators, also includes the following steps: Step 3.4, calculate the consistency index. The specific calculation formula is as follows: , In the formula, To determine the matrix Consistency indicators To determine the matrix The consistency ratio, λ max (t) To determine the matrix The largest eigenvalue, RI(k) t ) is the average random consistency index. If CR (t) If <0.10, then accept the judgment matrix. Consistency, if CR (t) If the value is ≥0.10, then the judgment matrix... Adjustments were made to make CR (t) <0.

10.

6. The method for evaluating the hardware reliability of a fire reconnaissance robot system according to claim 1, characterized in that, In step 4, the specific steps for dynamically adjusting the subjective and objective weights of the fire reconnaissance robot's indicators based on real-time environmental data to obtain an adaptive comprehensive weight vector are as follows: Step 4.1, construct the linear combination W' of subjective and objective weights as follows: In the formula, α s and α o Let W be the coefficients of the linear combination to be optimized. s (t) ) T Let W be the transpose of the subjective weights. o (t) ) T This is the transpose of the objective weights; Step 4.2, for the linear combination coefficients α s and α o Optimization is performed by minimizing the deviation in the objective function to find the optimal linearity of the weights. In the formula, Let L be the L2 norm of the vector, i.e., the Euclidean distance. This means minimizing the sum of those distances; Step 4.3, based on the properties of matrix differentiation, the linear combination coefficients α s and α o The optimization problem is transformed into a system of linear differential equations with first-order derivative conditions for optimization: In the formula, w s and w o Subjective weights W s (t) and objective weight W o (t) Weighting adjustment factor; Step 4.4: Solve the system of linear differential equations to obtain the optimal linear combination coefficients α. s 'and α o '; Step 4.5, for the linear combination coefficients α s 'and α o Normalization is performed to obtain the comprehensive weight vector W based on game theory weighting. c for: In the formula, These are the normalized subjective weight coefficients. These are the normalized objective weighting coefficients.

7. The method for evaluating the hardware reliability of a fire reconnaissance robot system according to claim 6, characterized in that, Step 4, when dynamically adjusting to obtain the adaptive comprehensive weight vector, also includes the following steps: Step 4.6, for each core evaluation factor d, d=1,2,…,k t Based on the sensitivity of the basic assessment data associated with the core assessment factors to environmental risks, a risk impact factor r is defined for each factor. d (t) An environmental risk vector is constructed using the risk impact factors of all core evaluation factors. for: In the formula, These are the risk impact factor values ​​of each core evaluation factor at evaluation time t, with an initial value of 0; Step 4.7, using environmental risk vectors For the comprehensive weight vector W c Dynamic correction is performed, and the correction formula is as follows: In the formula, To determine the corrected weight value of the j-th factor, w cj It is W c The j-th component, β is the global risk sensitivity coefficient, and 0≤β≤1, is used to control the strength of risk correction; Step 4.8, adjust the corrected comprehensive weight vector W. c Normalization is performed to generate the final comprehensive weight vector W for the fuzzy comprehensive evaluation at the current evaluation time t. T for: , In the formula, This represents the final weight of the j-th factor after normalization.

8. The method for evaluating the hardware reliability of a fire reconnaissance robot system according to claim 1, characterized in that, In step 5, the specific steps for constructing a fuzzy relation matrix using the core evaluation factors and the comprehensive weight vector, and then performing a synthesis operation to obtain the real-time updated system reliability level are as follows: Step 5.1, using k t The core evaluation factors constitute the dynamic factor set U at the current evaluation time t. t for: In the formula, Each of these represents a dynamic factor, and each represents a dynamically refined dimension of reliability impact. Step 5.2: Define a comment set V containing 5 evaluation levels, and assign a standard score to each evaluation level to form a score vector V. s for: , In the formula, Comments for each rating level. These correspond to five different levels of feedback; Step 5.3, for dynamic factor u i (t) and rating level v s Let the corresponding trapezoidal membership function parameter range be [a j ,b j ,c j ], of which it belongs entirely to the evaluation level v s The interval is [b j ,c j ], between rating level v j-1 With v j The transition interval between them is [a j ,b j ], between rating level v j With v j+1 The transition interval between them is [c j ,d j For dynamic factors u i (t) and rating level v s The trapezoidal membership function is used to calculate the membership degree r of the factor to that level. ij (t) The calculation formula is: Step 5.4, based on k t The membership degrees of each dynamic factor to the five evaluation levels are used to construct the fuzzy relation matrix R at the current evaluation time t. t for: Step 5.5, using the final comprehensive weight vector W at the current evaluation time t. T With fuzzy relation matrix R t Fuzzy synthesis operations are performed, using a weighted average synthesis operator, to obtain the comprehensive membership vector B for each evaluation level. t for: In the formula, " " indicates fuzzy composition operation; Step 5.6: Assign scores to the comment set V, and calculate the final score S as follows: Step 5.7: Obtain the comprehensive reliability evaluation score based on the final score S, and then obtain the corresponding reliability level through the preset level standard.

Images