FMECA method for precision reliability of safe and credible numerical control machine tool

Abstract

The invention discloses an FMECA method for the precision reliability of a safe and credible numerical control machine tool, and the method comprises the steps: giving out the definition of the precision reliability of the safe and credible numerical control machine tool, classifying faults affecting the precision reliability of the safe and credible numerical control machine tool, constructing an FMECA knowledge multi-ontology frame for the precision reliability of the safe and credible numerical control machine tool, and carrying out the fault diagnosis of the precision reliability of the safe and credible numerical control machine tool. Forming a fault mode library influencing the precision reliability of the safe and credible numerical control machine tool; the method comprises the steps of analyzing fault modes influencing reliability, analyzing fault modes, reasons and influences by using a full fault mode method based on a meta-action theory aiming at mechanical faults, sorting the fault modes by using a method combining an entropy weight method and a fuzzy TOPSIS method aiming at harmfulness analysis of the mechanical faults, and putting forward improvement measures according to a harmfulness analysis result; and a deep neural network technology is used to identify newly collected fault modes, and continuous updating of the fault mode library is realized. According to the method, the FMECA analysis of the precision reliability of the safe and credible numerical control machine tool can be realized by improving the traditional FMECA method.

Description

Technical Field

[0001] The invention belongs to the research field of FMECA of safe and reliable CNC machine tools, and in particular relates to an FMECA method for the precision reliability of safe and reliable CNC machine tools. Background Art

[0002] As the working machine of the manufacturing industry, CNC machine tools are the material basis of the modern manufacturing system. With the advancement of technology, the personalized needs of users have made the structure of CNC machine tools more complex and the working conditions more variable. This multi-load and multi-task working mode has also led to the characteristics of multiple fault modes of CNC machine tools, which has brought many difficulties to the use and maintenance of machine tools.

[0003] Failure Modes, Effects, and Criticality Analysis (FMECA) is a method for discovering potential failures. Its main purpose is to discover potential failure modes that may have adverse effects on system functions or performance. Through FMECA, the possible failure modes of each unit and their impact on the system can be determined, so that targeted improvement measures can be taken to reduce reliability problems caused by unclear understanding of unit failure modes.

[0004] The traditional FMECA method is a bottom-up analysis method, which is usually carried out in the requirements analysis and outline design phase of software development. Its analysis objects are mainly components, functional modules or subsystems at the architecture layer. Therefore, the real purpose of the FMECA method is to predict and quantify various possible failures in the early stage of software development, and if necessary, take corresponding measures to prevent and avoid them. Therefore, FMECA can also be considered as a quality management tool, which is essentially a multi-objective decision-making (MODM) problem, which requires a reasonable evaluation method and structure.

[0005] The shortcomings of the traditional FMECA method include: 1) the knowledge in related fields is scattered and unsystematic, which makes the analysis results highly dependent on the experience of the analyst and his familiarity with the field to which the analysis system belongs; 2) it is usually based on text descriptions and has no tool support; 3) the scoring strategy uses exact values ​​to represent fuzzy semantics, and the ranking of fault mode priorities ignores the weight differences of various indicators, making it difficult to obtain more accurate analysis results; 4) the identification and judgment of new fault modes depends on manual work, which is labor-intensive and time-consuming, and requires personnel to have rich engineering experience and high technical level, increasing the risk of errors in the fault mode classification and identification process. Summary of the invention

[0006] The purpose of the present invention is to provide a FMECA method for the precision reliability of a safe and reliable CNC machine tool, which can realize the FMECA analysis of the precision reliability of a safe and reliable CNC machine tool by improving the traditional FMECA method.

[0007] In order to achieve the above object, one aspect of the present invention provides a safe and reliable FMECA method for the precision reliability of CNC machine tools, comprising:

[0008] Step S1: Give a definition of the precision reliability of safe and reliable CNC machine tools, classify the faults that affect the precision reliability of safe and reliable CNC machine tools, build an FMECA knowledge multi-ontology framework of the precision reliability of safe and reliable CNC machine tools, and form a fault mode library that affects the precision reliability of safe and reliable CNC machine tools;

[0009] Step S2: Perform FMECA analysis on the failure modes that affect the precision reliability of safe and reliable CNC machine tools, wherein the failure modes include mechanical failures. For mechanical failures, the full failure mode method based on meta-action theory is used to analyze the failure modes, causes and effects. For criticality analysis of mechanical failures, the entropy weight method and the fuzzy TOPSIS method are combined to rank the failure modes, and improvement measures are proposed according to the criticality analysis results.

[0010] Step S3: Use deep neural network technology to identify newly collected fault modes to achieve continuous updating of the fault mode library.

[0011] Preferably, in step S1, a definition of the precision reliability of a safe and reliable CNC machine tool is given as: the probability that the machining precision of a CNC machine tool remains within a specified range under specified machining conditions and within a specified machining time, and faults that affect the precision reliability of a safe and reliable CNC machine tool are classified into machine tool body faults and electrical system and CNC system faults, wherein machine tool body faults are all mechanical faults, and electrical system and CNC system faults include mechanical faults, motor and electronic component faults, and software faults.

[0012] Preferably, mechanical failure modes include motion trajectory failure modes, motion speed failure modes, and position positioning failure modes; software failure modes include processing program failure modes related to processing program compilation and processing, and safety protection failure modes caused by the lack of network security protection measures; safety protection failure modes include intentional failure modes and negligent failure modes.

[0013] Preferably, in step S1, constructing a FMECA knowledge multi-ontology framework for the precision reliability of safe and reliable CNC machine tools includes:

[0014] Construct the generalized ontology of the precision reliability FMECA knowledge ontology of safe and reliable CNC machine tools;

[0015] Based on the collected failure mode data, by enriching the core concepts and associations of the domain, a domain ontology concept space with the FMECA causal chain of the precision reliability of safe and reliable CNC machine tools as the core is formed, thereby constructing the domain ontology of the precision reliability FMECA knowledge ontology of safe and reliable CNC machine tools;

[0016] Based on the collected failure mode data, the failure mode data and improvement measures of the application domain are refined. By enriching the core concepts and associations of the application domain, an application ontology concept space with the FMECA causal chain of the precision reliability of safe and reliable CNC machine tools as the core is formed, thereby constructing the application ontology of the precision reliability FMECA knowledge ontology of safe and reliable CNC machine tools.

[0017] Preferably, in step S2, the full failure mode method based on the meta-action theory is used to analyze the failure mode, cause and impact of the mechanical failure, including: defining the failure mode of the mechanical transmission system on the motion form of the meta-action, defining the failure cause on the mechanical structure of the meta-action unit, and defining the failure impact as the impact of each failure mode of the meta-action on its own agreed level and product function.

[0018] Preferably, in step S2, the method of combining the entropy weight method and the fuzzy TOPSIS method to sort the fault modes includes:

[0019] Assume m fault modes FMi, i = 1, 2, ..., m and n evaluation indicators, the jth indicator value C of each fault mode j , j = 1, 2, ..., n, obtained by the evaluation of multiple experts. Suppose there are k experts to conduct the evaluation. The evaluation results of k groups of experts are converted into a fuzzy number set, and the evaluation fuzzy result of the jth indicator of the i-th fault mode of the k-th group is obtained as follows:

[0020]

[0021] According to the fuzzy decision matrix based on the fuzzy evaluation results, the entropy weight method is used to obtain the weights of the fuzzy risk factors of the failure modes, and then the fuzzy TOPSIS method is used to rank all the failure modes.

[0022] Preferably, the weight of the fuzzy risk factor of the failure mode is obtained using the entropy weight method as follows:

[0023] make represents entropy, where pi is the probability of a random variable calculated from the probability density function P, using Convert fuzzy numbers into exact values ​​and use i=1,2,...,m,j=1,2,...,n are normalized, where x ijrepresents the exact value of the i-th fault mode for the j-th indicator;

[0024] set up Where k>0,e j ≥0, ln is the natural logarithm, and g is defined j =1-e j is a diversification indicator, when g j When the value is larger, the weight of index j is larger. When the diversified indexes of all indexes are obtained, the weight of the fuzzy risk factor of the failure mode is j=1,2,…,n.

[0025] Preferably, all failure modes are ranked using the fuzzy TOPSIS method as follows:

[0026] Normalize the fuzzy decision matrix;

[0027] Calculate the weighted fuzzy decision matrix Among them, w j is the weight of the fuzzy risk factor of the failure mode, r ij is the element in the normalized fuzzy decision matrix;

[0028] Determine the positive ideal solution and negative ideal solution

[0029] Calculate the ideal solution distance and get i=1,2,…,m, i=1,2,...,m,where Represents the distance between two fuzzy numbers;

[0030] Compute closeness to rank all failure modes: i=1,2,…,m.

[0031] Preferably, in step S3, using deep neural network technology to identify newly collected fault modes includes: data set construction, fault mode neural network construction, model training and model identification, wherein the fault mode neural network construction uses a DS-cascaded regional convolutional neural network to perform machine tool fault mode identification.

[0032] Preferably, ResNet101 is used as the feature extraction network, and the convolution layer of the last part of the residual network 101 is replaced by a deformable convolution, and the feature map output by the deformable convolution is input into the spatial attention module. After the candidate region is selected by the region proposal network, the features of the candidate target region and the feature map output by the spatial attention module are fused after unifying the size through the pooling layer, and then input into the fully connected layer for classification; in the regression branch part, a convolution layer is used to retain more spatial information, and the cascade detector connects three detectors in series, resamples the border regression domain output by the previous layer, and sends it to the next layer of detector as a new input, so as to gradually improve the sample quality and network training effect, and obtain a series of machine tool fault mode classification and regional prediction frames.

[0033] According to the FMECA method for the precision reliability of safe and reliable CNC machine tools of the above aspects of the present invention, FMECA analysis of the precision reliability of safe and reliable CNC machine tools is achieved through three aspects: FMECA knowledge ontology construction, criticality analysis of mechanical failures, and continuous updating of the failure mode library using deep neural network technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings used in the description of the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work:

[0035] Figure 1 It is a flow chart of a FMECA method for the precision reliability of a safe and reliable CNC machine tool according to an embodiment of the present invention;

[0036] Figure 2 It is a principle diagram of the FMECA method for the accuracy reliability of a safe and reliable CNC machine tool according to an embodiment of the present invention;

[0037] Figure 3 It is a schematic diagram of a FMECA knowledge multi-ontology framework for the precision reliability of a safe and reliable CNC machine tool according to an embodiment of the present invention;

[0038] Figure 4 is a schematic diagram of an FMECA cause-and-effect relationship chain according to an embodiment of the present invention;

[0039] Figure 5 is a flow chart of constructing a fault mode recognition model according to an embodiment of the present invention;

[0040] Figure 6 This is a DS-Cascade R-CNN model diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] An embodiment of the present invention provides a safe and reliable FMECA method for the precision reliability of CNC machine tools, such as Figure 1 As shown, the FMECA method for the precision reliability of safe and reliable CNC machine tools in the embodiment of the present invention includes steps S1 to S3. Figure 2 The schematic diagram of the present invention is used to describe in detail each step of the method of the embodiment of the present invention.

[0043] In step S1, the definition of the precision reliability of safe and reliable CNC machine tools is given, the faults that affect the precision reliability of safe and reliable CNC machine tools are classified, the FMECA knowledge multi-ontology framework of the precision reliability of safe and reliable CNC machine tools is constructed, and a fault mode library that affects the precision reliability of safe and reliable CNC machine tools is formed.

[0044] In this step, the definition of the precision reliability of safe and reliable CNC machine tools is first given: the precision reliability of safe and reliable CNC machine tools refers to the probability that the machining accuracy of the CNC machine tool remains within the specified range under the specified machining conditions and within the specified machining time [0, t]. If a non-negative random variable X is used to describe the precision life of the CNC machine tool, the corresponding distribution function (fault distribution function) of X is:

[0045] F(t)=P(X≤t)(1)

[0046] Where F(0) = 0,

[0047] F(t) is the probability that the CNC machine tool will fail before time t.

[0048] Then the probability that the CNC machine tool does not fail before time t is:

[0049]

[0050] R(t) in formula (2) is called the accuracy reliability or reliability function of the CNC machine tool. R(t)≥0, R(0)=1, and

[0051] From this definition, it can be seen that when the machining accuracy of the CNC machine tool is not within the specified range, that is, it exceeds the threshold range, the CNC machine tool is considered to be faulty.

[0052] Then, the faults that affect the precision reliability of safe and reliable CNC machine tools are classified. The reliability of the product is divided into two categories: functional reliability and performance reliability. The corresponding product faults can also be divided into functional faults and performance faults. For systems such as CNC machine tools, the most common problem during use is not the functional faults that make the machine tool unable to continue to be used. The biggest problem is often caused by unreliable performance, such as poor accuracy and short accuracy life. The precision reliability of safe and reliable CNC machine tools reflects the accuracy stability and accuracy life of CNC machine tools, and the faults that affect the precision reliability belong to the category of performance faults. The above faults are divided into the following two categories according to the location of occurrence: machine tool body faults and electrical system and CNC system faults. Among them, machine tool body faults are all mechanical faults, and electrical system and CNC system faults include mechanical faults, motor and electronic component faults, software faults, etc.

[0053] CNC machine tools are electromechanical composite equipment with complex structures and diverse functions. At the same time, the cognitive uncertainty caused by the superposition of a large number of uncertain factors also affects the distinction of fault states, the identification of fault modes, and the analysis of fault mechanisms and fault effects. Therefore, it is necessary to adopt a scientific and reasonable method to represent fault-related knowledge, which can not only accurately and clearly express the meaning of factors such as fault characterization and fault effects, but also avoid ambiguity, incompleteness and inconsistency in the description process to the greatest extent. The embodiment of the present invention constructs the FMECA knowledge ontology by using the ontology method based on the Knowledge Aided Design System (KADS). In the knowledge engineering community, ontology is considered to be a clear and standardized description of concepts. By adopting the ontology method, FMECA knowledge is represented as ontology concepts and associations, so it is clear (unambiguous), complete and consistent, and is conducive to knowledge sharing and reuse.

[0054] Constructing the FMECA knowledge multi-ontology framework for the precision reliability of safe and reliable CNC machine tools Figure 3 As shown in the figure, the generalized ontology describes generalized concepts that are independent of specific problems or fields and is located at the top level of the ontology structure. The domain ontology is a description of the general domain and a specialization of the generalized ontology. The application ontology depends on a specific domain and is a specialization of the domain ontology. It can be seen that there are classes and inheritance relationships between ontologies. The domain ontology can be reused in the same domain, converted into a domain model through domain analysis, and obtained as an application model through application analysis. In the process of building a multi-ontology framework, domain experts, customers, users, and system developers can all participate in this process. Therefore, this framework integrates multiple perspectives of stakeholders.

[0055] First, the generalized ontology of the precision reliability FMECA knowledge ontology of safe and reliable CNC machine tools is constructed. The generalized ontology (GO) reflects the common ontology elements of this type of system, which can be realized by constructing its concepts (classes), concept class hierarchies, associations, attributes and attribute characteristics.

[0056] Secondly, the domain ontology (DO) of the precision reliability FMECA knowledge ontology of safe and reliable CNC machine tools is constructed. The domain ontology is a special ontology used to describe the knowledge in a specified domain. It gives a description of the domain concepts and their relationships, domain activities, and the characteristics and laws of the domain. These concepts and associations are obtained by instantiating the corresponding contents in the generalized ontology with various documents, domain knowledge and industry standards as the source. The domain ontology concept space should contain domain concepts, attributes and associations, and provide activities and main theories occurring in the domain. The core of the domain ontology of the precision reliability FMECA knowledge ontology of safe and reliable CNC machine tools is essentially to describe the failure modes, causes, effects, degrees of harm and corresponding improvement measures that affect the precision reliability of safe and reliable CNC machine tools, so it can be characterized from both static and dynamic aspects.

[0057] The embodiment of the present invention builds the FMEA knowledge ontology of CNC machine tool precision maintenance reliability based on GJBZ1391-2006 "Guide to Failure Mode, Effect and Criticality Analysis" and "IEEESTD 1044-1993", and its core concepts include: failure mode, failure cause, failure impact, failure risk number (SRPN), occurrence frequency (SOPR), severity (SESR) and detection index (SDDR), etc. This is also a characterization of the field from a static perspective.

[0058] In addition, this field can also be characterized from a dynamic perspective. From an ontological point of view, the causes of failures and their impacts can be modeled as causal relationships. Therefore, the core of the ontology of this field is a chain structure composed of a series of such causal relationships and the domain concepts and associations surrounding this structure. The formation of this chain structure is caused by the transitivity of fault impacts. Fault impacts can be divided into local impacts, higher-level impacts, and final impacts. For a given level, its fault impact is the fault manifestation of the immediately upper level, that is, the fault mode. This transitivity between different levels also enables the formation of the causal chain of FMECA, such as Figure 4 As shown. Figure 4In the causal chain, the change of state (characterized by constraints) leads to the occurrence of events, and the event in turn acts on the state, causing the state to change further, and finally forming the structure of the fault causal chain. In engineering practice, the change of state is characterized by the fault cause (mechanism) or fault mode, and the occurrence of events can be characterized by the fault impact, including local impact, higher-level impact and final impact. From the bottom state 1, through constraint 1, event 1 is caused, and through transmission to the middle state i, through constraint i, event i is caused, and finally through transmission to the top state n, through constraint n, event n is caused.

[0059] Based on the collected failure mode data, by enriching the core concepts and associations in the field, a DO concept space with the FMECA causal chain of safe and reliable CNC machine tool precision reliability as the core is formed.

[0060] Finally, the application ontology (AO) of the precision reliability FMECA knowledge ontology of safe and reliable CNC machine tools is constructed. Since the FMECA activities in the embodiment of the present invention are based on ontology, it is necessary to take the FMECA goal as a guide, within the defined system boundary, based on the generalized ontology, and according to the mapping relationship of the concept class, collect domain-related concept terms and associations, and classify them. Further, based on the generalized ontology and the domain ontology, according to the mapping relationship of the concept class, with various documents, domain knowledge, application domain knowledge, industry standards, fault data and existing systems as the source, collect application domain-related concept terms and associations, and classify them. Based on the collected failure mode data, the failure mode data and improvement measures of the application domain are refined. Through the above-mentioned enrichment of the core concepts and associations of the application domain, an AO concept space with the FMECA causal chain of the precision reliability of safe and reliable CNC machine tools as the core is formed.

[0061] Protégé is an extensible, cross-platform and open source development environment for generating and editing ontologies. It can define classes and class hierarchies, attribute relationships and attribute-value constraints, as well as relationships between classes and attributes. It can obtain instances of classes defined in the ontology through instance tags. Currently, Protégé has been widely used and promoted, so the embodiment of the present invention uses Protégé to edit the ontology. After the ontology is built, it is also necessary to use the improved FOCA method to evaluate the ontology quality from three aspects: ontology type verification, problem verification and quality verification.

[0062] By constructing the FMECA knowledge ontology of the precision reliability of safe and reliable CNC machine tools, the reuse of FMECA knowledge becomes possible and helps to improve the quality and efficiency of FMECA. According to the FMECA knowledge multi-ontology framework of the precision reliability of safe and reliable CNC machine tools constructed above, a fault mode library that affects the precision reliability of safe and reliable CNC machine tools is formed. The fault mode library contains the fault mode data that affects the precision reliability of safe and reliable CNC machine tools under the definition of precision reliability, including mechanical fault mode and other fault modes. Mechanical fault mode includes motion trajectory fault mode, motion speed fault mode, and position positioning fault mode; other fault modes include software fault mode and motor and electronic component fault mode. Software fault mode is further divided into safety protection fault mode and processing program fault mode. Safety protection fault mode is further divided into intentional fault mode and negligent fault mode. Through the obtained fault information, including the analysis of fault classification, fault mode and fault cause within the scope of the research on faults affecting the precision reliability of safe and reliable CNC machine tools, the knowledge base is enriched.

[0063] In step S2, an FMECA analysis is performed on the failure modes that affect the precision reliability of the safe and reliable CNC machine tool. The failure modes include mechanical failures. The full failure mode method based on the meta-action theory is used to analyze the failure modes, causes and effects of mechanical failures. For the criticality analysis of mechanical failures, a method based on entropy weight and fuzzy TOPSIS is used to rank the failure modes, and improvement measures are proposed according to the criticality analysis results.

[0064] For FMECA of mechanical failure, the full failure mode method based on meta-action theory is used to analyze the failure mode, cause and impact of mechanical failure. According to meta-action theory, meta-action is the most basic motion unit of mechanical transmission system. As long as the meta-action unit fails, the product is in a faulty state. Therefore, the failure mode of mechanical transmission system can be defined on the motion form of meta-action, the cause of failure can be defined on the mechanical structure of meta-action unit, and the failure impact refers to the impact of each failure mode of meta-action on its own agreed level (itself and its next meta-action) and product function. The motion form of a meta-action part (including two types of moving meta-action parts and rotating meta-action parts) is characterized by three most basic kinematic parameters: "motion trajectory", "motion speed" and "position positioning". As long as any of these three motion parameters varies, the motion of the meta-action will deviate from the predetermined requirements, and then a motion failure occurs. The full failure mode analysis method can be used to complete the analysis of the mechanical failure mode and cause of safe and reliable CNC machine tools, and covers most types of mechanical parts.

[0065] In order to avoid the shortcomings of the traditional risk priority number (RPN) method for criticality analysis of mechanical failures, a method based on entropy weight and fuzzy TOPSIS (Technique for Order Preference by Similarity to an Ideal Solution) is used to rank the failure modes. The results of the failure mode ranking will serve as the basis for expert evaluation of the application effect of subsequent FMECA activities and the basis for determining the key subsystems of the machine tool. This method assumes that there are m failure modes FM i (i=1,2,...,m) and n evaluation indicators. The jth indicator value C of each fault mode j (j=1, 2, ..., n) is obtained by the evaluation of multiple experts. Assuming that there are k experts making decisions, the semantic evaluation results of k groups of experts can be converted into a set of fuzzy numbers. The calculation results of each indicator for each fault mode are obtained:

[0066]

[0067] represents the evaluation fuzzy result of the jth index of the ith fault mode of the kth group. The ranking problem of fault modes can be regarded as a fuzzy multi-objective decision-making problem.

[0068] The first step is to calculate the entropy weight. The specific process is as follows:

[0069] make represents entropy, where p i (i=1, 2, ..., n) is the probability of a random variable calculated from the probability density function P. Use Convert fuzzy numbers into exact values. And use i=1,2,...,m,j=1,2,...,n for normalization. ij represents the exact value of the i-th fault mode for the j-th indicator. This step converts the different units or scales of different indicators into the same measurement unit. Where k>0,e j ≥0, 1n is the natural logarithm. If for index j, all x ij are all equal, that is Then e j Get the maximum value: Here, let k = 1 / lnm, then 0 ≤ e j ≤1. For a given index j, when x ij The smaller the change, the greater the entropy value e j The larger the value, the greater the ij When equal, e j =max ej =1. When x ij The greater the change, the greater the entropy value e j The smaller it is, the greater the weight of indicator j should be. So define g j =1-e j is a diversification indicator, when g j The larger is the value, the greater the weight of the corresponding indicator should be. After obtaining the diversified indicators of all indicators, the weight of the fuzzy risk factor of the failure mode of the CNC machine tool is j = 1, 2, ..., n. That is, the diversification index is calculated by the entropy value, and then the weight of the risk factor is calculated by the diversification index.

[0070] The second step is to perform fuzzy TOPSIS. The specific process is as follows:

[0071] TOPSIS is a solution to multi-objective decision-making problems that allows for trade-offs between various indicators. It is more reasonable than the traditional FMECA RPN (Risk Priority Number). The fuzzy TOPSIS method based on entropy weights first normalizes the fuzzy decision matrix. Then, the weighted fuzzy decision matrix is ​​calculated. Among them, w j is the weight of the fuzzy risk factor of the failure mode of the CNC machine tool, r ij is the element in the normalized fuzzy decision matrix, which is obtained by normalization based on the previous formula (3). Then determine the positive ideal solution and negative ideal solution After that, the ideal solution distance is calculated and we get i=1,2,…,m, i=1,2,...,m,where represents the distance between two fuzzy numbers, p and q are two triangular fuzzy numbers. The embodiment of the present invention uses the calculation formula of Euclidean distance: in, and are two triangular fuzzy numbers. After obtaining the ideal solution distance, the closeness is calculated to rank all failure modes: i = 1, 2, ..., m. The value of the progress CCi is calculated. The larger the value, the higher the ranking, which also means that the importance (criticality) of the corresponding CNC machine tool failure mode is higher.

[0072] FMECA for other faults other than mechanical faults is similar to the traditional method, but the specific content varies according to different research objects. Figure 1As shown in the figure, other faults besides mechanical faults include motor and electronic component faults and software faults. The FMECA of motor and electronic component faults will not be described here, and the focus will be on the FMECA of software faults. This type of fault is divided into two categories: faults related to the compilation and processing of machining programs (machining technology) and faults caused by the lack of network security protection measures. With the development of technology and the improvement of software engineering level, the occurrence of the former type of faults is showing a downward trend; while the occurrence of the latter type of faults is showing an upward trend. The latter type of faults can be divided into two categories according to the intention of the operator: "intentional" and "negligent". "Intentional" faults refer to the occurrence of CNC system faults that affect the accuracy and reliability level of CNC machine tools by the operator's artificial network attack for the purpose of attack. "Negligence" refers to the occurrence of CNC system faults that affect the accuracy and reliability level of CNC machine tools, which are not necessarily for the purpose of attack. Generally, the operator's misoperation causes the occurrence of CNC system faults that affect the accuracy and reliability level of CNC machine tools. In engineering practice, it is found that most of the cases of "intentional" faults are caused by the lack of safety protection of the CNC system. In addition to the simplest case of a single CNC machine tool, in more complex scenarios such as CNC production lines and CNC workshops, due to the communication between various equipment such as CNC systems, robots, and transmission equipment from different manufacturers and supporting different communication protocols, and most communication protocols use plain text transmission, it provides a suitable soil for the occurrence of "intentional" faults. As for the case of "negligent" faults, most of them are due to the failure to clearly set various operating roles (operators, maintenance engineers, process engineers, workshop directors, machine tool factory debuggers, machine tool factory administrators, CNC system debuggers, CNC system administrators, safety administrators, auditors, etc.) and corresponding authority levels. Since different roles have different business scopes and operating permissions, unclear role positioning will lead to personnel without operating permissions operating the CNC system (such as parameter setting), which will affect the processing accuracy of the CNC machine tools and cause damage to the CNC machine tools.

[0073] The fundamental solution to the failures caused by the lack of network security protection means is to use security protection means on the basis of the existing CNC machine tools (systems). Especially in the industrial Internet environment, CNC machine tools (systems) face greater network security threats. It is necessary to add a secure and trusted module on the basis of the existing CNC system architecture. Based on the existing computer active immune mechanism, use trusted measurement, trusted verification, trust chain transmission and other methods to ensure the security and trustworthiness of the CNC system. Study the identity authentication mechanism and access control strategy for the operation authority of the CNC system; at the same time, provide security protection for local and remote key business data in the CNC system to ensure data integrity and confidentiality (such as ciphertext transmission and security analysis of process files, processing codes, etc.).

[0074] In step S3, deep neural network technology is used to identify newly collected fault modes to achieve continuous updating of the fault mode library.

[0075] Since the work of fault collection is ongoing, the increase of faults makes the fault mode library in a dynamic update state. There are various types of fault modes for safe and reliable CNC machine tools, and each type has several subcategories. Manual identification alone is labor-intensive and time-consuming, and requires personnel with rich engineering experience and high technical level, which increases the risk of errors in the process of fault mode classification and identification. Therefore, how to identify new fault modes is a difficult point. The embodiment of the present invention uses deep neural network technology to identify new fault modes more accurately and intelligently to achieve Figure 1 The function of failure mode library update in .

[0076] like Figure 5 As shown, building a fault mode recognition model includes: data set construction, fault mode neural network construction, model training and model recognition, among which the fault mode neural network construction is the focus.

[0077] (1) Establishing data sets: CNC machine tool fault data sets are mainly divided into two categories: one is a text-based fault mode data set; the other is a picture-based fault mode data set. After collecting these data sets, since artificially collected data sets inevitably have some problems and flaws, these data sets need to be preprocessed by classification, cleaning, and enhancement to ensure the accuracy of the data sets.

[0078] Since the machine tool fault dataset contains text datasets and image datasets, general artificial intelligence models cannot input both image and text datasets for training at the same time, so the image dataset needs to be converted into a vector dataset, that is, a text dataset. The first few layers of the neural network of the artificial intelligence model convert image pixels into vectors, so the image dataset is directly input into the neural network, while the text dataset is input into the neural network for model training after the first few layers.

[0079] (2) Construction of Fault Mode Neural Network

[0080] Machine tool fault pattern recognition is a target detection and classification problem. Based on the research and comparison of various mainstream artificial intelligence detection and classification models, the embodiment of the present invention combines the inherent characteristics of machine tool fault pattern recognition and uses a machine tool fault pattern recognition method based on DS-cascade region-convolutional neural networks (DS-cascadeR-CNN). Figure 6As shown. Based on the original cascade R-CNN structure, deformable convolution is added, and the spatial attention mechanism is adopted to enhance the perception area of ​​the feature map, propose decoupled classification and regression branches, and finally optimize the model to reduce redundant parameters. DS-cascade R-CNN uses ResNet101 as the feature extraction network, and replaces the 3×3 convolution of the last part conv5 of the residual network 101 (residual network 101, ResNet101) with deformable convolution. In order to learn the offset on relatively stable features, the deformable convolution is set as the convolution layer behind ResNet101. DS-cascade R-CNN inputs the feature map output by deformable convolution (deformableconvnets, DCN) into the spatial attention module to learn the spatial dependency of features in spatial positions. In addition, an embodiment of the present invention also proposes decoupled classification and regression branches. After the candidate regions are selected by the region proposal network (RPN), the features of the candidate target regions and the feature maps output by the spatial attention module are fused after being resized through the pooling layer, and then input into the fully connected layer (FC) for classification. In the regression branch, the convolution layer is used to retain more spatial information. The cascade detector connects three detectors in series, resamples the bounding box regression domain output by the previous layer, and fine-tunes it as a new input to the next layer of detectors, gradually improving the sample quality and network training effect, and obtaining a series of machine tool fault mode classification and regional prediction frames. Under the cascade structure, through step-by-step regression, the positioning of the candidate frame will tend to be accurate, and the threshold between it and the true reference frame will also increase accordingly.

[0081] (3) Model training: First, based on existing experience and data from most engineering experiments, the initial parameters and hyperparameters of the model are set to train the model. Then, based on the size of the loss function of the model training, the hyperparameters and parameters are gradually tuned. Finally, after multiple rounds of iterations, the loss function of the model reaches the minimum or smaller value, and the optimal model parameters are obtained.

[0082] (4) Model fault identification: After the model is trained, it is tested and verified on the test set and validation set to verify the recognition and detection capabilities of the constructed machine tool fault detection model. If the machine tool fault category can be correctly identified and the detection accuracy and precision are high, the trained model can be used for deployment and identification; if the fault category can be partially identified correctly and partially identified incorrectly, and the recognition and detection accuracy and precision are low, the cause of the problem is analyzed and the model is retrained.

[0083] In summary, the method of the embodiment of the present invention proposes the concept of precision reliability of safe and reliable CNC machine tools based on the confidence reliability theory, studies the FMECA method of faults that affect this reliability, and improves the traditional FMECA method from three aspects: FMECA knowledge ontology construction, criticality analysis of mechanical faults, and continuous updating of the fault mode library using deep neural network technology.

[0084] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A safe and reliable FMECA method for the precision reliability of CNC machine tools, characterized in that: include: Step S1: Give a definition of the precision reliability of safe and reliable CNC machine tools, classify the faults that affect the precision reliability of safe and reliable CNC machine tools, build an FMECA knowledge multi-ontology framework of the precision reliability of safe and reliable CNC machine tools, and form a fault mode library that affects the precision reliability of safe and reliable CNC machine tools; Step S2: Perform FMECA analysis on the failure modes that affect the precision reliability of safe and reliable CNC machine tools, wherein the failure modes include mechanical failures. For mechanical failures, the full failure mode method based on meta-action theory is used to analyze the failure modes, causes and effects. For criticality analysis of mechanical failures, the entropy weight method and the fuzzy TOPSIS method are combined to rank the failure modes, and improvement measures are proposed according to the criticality analysis results. Step S3: Use deep neural network technology to identify newly collected fault modes to achieve continuous updating of the fault mode library.

2. The method according to claim 1, characterized in that In step S1, the definition of the precision reliability of safe and reliable CNC machine tools is given as: the probability that the machining accuracy of the CNC machine tool remains within the specified range under the specified machining conditions and within the specified machining time. The faults that affect the precision reliability of safe and reliable CNC machine tools are classified into machine tool body faults and electrical system and CNC system faults. Among them, machine tool body faults are all mechanical faults, and electrical system and CNC system faults include mechanical faults, motor and electronic component faults, and software faults.

3. The method according to claim 2, characterized in that Mechanical failure modes include motion trajectory failure modes, motion speed failure modes, and position positioning failure modes. Software failure modes include processing program failure modes related to processing program compilation and processing, as well as safety protection failure modes caused by the lack of network security protection measures. Safety protection failure modes include intentional failure modes and negligent failure modes.

4. The method according to any one of claims 1 to 3, characterized in that In step S1, the FMECA knowledge multi-ontology framework for the accuracy reliability of safe and reliable CNC machine tools is constructed, including: Construct the generalized ontology of the precision reliability FMECA knowledge ontology of safe and reliable CNC machine tools; Based on the collected failure mode data, by enriching the core concepts and associations of the domain, a domain ontology concept space with the FMECA causal chain of the precision reliability of safe and reliable CNC machine tools as the core is formed, thereby constructing the domain ontology of the precision reliability FMECA knowledge ontology of safe and reliable CNC machine tools; Based on the collected failure mode data, the failure mode data and improvement measures of the application domain are refined. By enriching the core concepts and associations of the application domain, an application ontology concept space with the FMECA causal chain of the precision reliability of safe and reliable CNC machine tools as the core is formed, thereby constructing the application ontology of the precision reliability FMECA knowledge ontology of safe and reliable CNC machine tools.

5. The method according to any one of claims 1 to 4, characterized in that In step S2, the full failure mode method based on the meta-action theory is used to analyze the failure mode, cause and impact of the mechanical failure, including: defining the failure mode of the mechanical transmission system on the motion form of the meta-action, defining the failure cause on the mechanical structure of the meta-action unit, and defining the failure impact as the impact of each failure mode of the meta-action on its own agreed level and product function.

6. The method according to any one of claims 1 to 5, characterized in that In step S2, the method of combining the entropy weight method and the fuzzy TOPSIS method to sort the fault modes includes: Assume m failure modes FM i , i = 1, 2, ..., m and n evaluation indicators, the jth indicator value C of each fault mode j , .j=1, 2, ..., n, obtained by the evaluation of multiple experts. Suppose there are k experts to conduct the evaluation. The evaluation results of k groups of experts are converted into fuzzy number sets, and the evaluation fuzzy result of the jth indicator of the i-th fault mode of the k-th group is obtained as follows: According to the fuzzy decision matrix based on the fuzzy evaluation results, the entropy weight method is used to obtain the weights of the fuzzy risk factors of the failure modes, and then the fuzzy TOPSIS method is used to rank all the failure modes.

7. The method according to claim 6, characterized in that The entropy weight method is used as follows to obtain the weight of the fuzzy risk factor of the failure mode: make represents entropy, where pi is the probability of a random variable calculated from the probability density function P, using Convert fuzzy numbers into exact values ​​and use Normalization is performed, where xij represents the exact value of the jth indicator for the i-th fault mode; set up Where k>0,e j ≥0, ln is the natural logarithm, and g is defined j =1-e j is a diversification indicator, when g j When the value is larger, the weight of index j is larger. When the diversified indexes of all indexes are obtained, the weight of the fuzzy risk factor of the failure mode is 8. The method according to claim 7, characterized in that All failure modes are ranked using the fuzzy TOPSIS method as follows: Normalize the fuzzy decision matrix; Calculate the weighted fuzzy decision matrix Among them, w j is the weight of the fuzzy risk factor of the failure mode, r ij is the element in the normalized fuzzy decision matrix; Determine the positive ideal solution and negative ideal solution Calculate the ideal solution distance and get in, Represents the distance between two fuzzy numbers; Compute closeness to rank all failure modes:

9. The method according to any one of claims 1 to 8, characterized in that In step S3, the newly collected fault modes are identified using deep neural network technology, including: data set construction, fault mode neural network construction, model training and model identification, wherein the fault mode neural network construction uses a DS-cascaded regional convolutional neural network to perform machine tool fault mode identification.

10. The method according to claim 9, characterized in that ResNet101 is used as the feature extraction network, and the convolution layer of the last part of the residual network 101 is replaced by deformable convolution. The feature map output by the deformable convolution is input into the spatial attention module. After the candidate region is selected by the region proposal network, the features of the candidate target region and the feature map output by the spatial attention module are fused after unifying the size through the pooling layer and then input into the fully connected layer for classification. In the regression branch part, the convolution layer is used to retain more spatial information. The cascade detector connects three detectors in series, resamples the bounding box regression domain output by the previous layer, and sends it to the next layer of detector as a new input, gradually improving the sample quality and network training effect, and obtaining a series of machine tool fault mode classification and regional prediction boxes.

Images