An industrial alarm analysis method and system

Abstract

The application discloses an industrial alarm analysis method and system, which comprises the following steps: deploying a master node and a collection node; generating alarm data; operating the alarm data by the master node and the collection node; generating alarm sequences; splicing alarm variables and alarm types of each alarm in Ra1 to form a new sequence in the original order and placing the new sequence into res1; placing alarm types of each alarm in Rb1 into res2 in the original order, recording the corresponding first and last alarms of the new sequence, updating the value of start1 to start1+l1, and updating the value of start2 to start2+l2; calculating alarm sequence support; and calculating alarm sequence similarity. The application can generate alarm sequences in a variable manner on the basis of guaranteeing real alarm time sequences, and then analyzes industrial alarms, gives analysis results, and assists operators in decision-making. The application solves the problem that real alarm time sequences are difficult to record by setting the master node and the collection node and synchronizing the time of the master node and the collection node. The application reduces the amount of alarm data by issuing rules to the collection node by the master node. According to the characteristics of industrial control alarms, the application defines an alarm sequence generation method, and can effectively solve the dynamic change problem of alarms.

Description

Technical Field

[0001] This invention relates to an alarm method and system, and more particularly to an industrial alarm analysis method and system. Background Technology

[0002] In industrial production, abnormal equipment operation, production process problems, or other abnormal situations trigger automatic alarm signals to prompt staff to take timely measures to prevent accidents or reduce losses. Industrial alarms are typically implemented using various sensors and monitoring devices through digital technology. These sensors and devices monitor parameters such as temperature, pressure, liquid level, and flow rate, converting them into electrical signals that are sent to the control system. When an abnormality is detected, an alarm signal is triggered. Industrial alarms are an important component of the industrial production safety assurance system, improving the safety and reliability of the production process, preventing and reducing accidents, and protecting the safety of personnel and equipment. Ideally, industrial alarm data is small; however, in real-world environments, alarm signals may be triggered frequently, generating too many industrial alarms, making it difficult for operators to make decisions, and even threatening production safety. Current technologies struggle to guarantee the true timing of alarm occurrences, and are unable to generate reasonable alarm sequences for analysis, handle large amounts of alarm data, and are easily affected by the alarm order during sequence analysis. Summary of the Invention

[0003] Purpose of the invention: The purpose of this invention is to provide an industrial alarm analysis method and system that, while ensuring the true timing of alarms, generates alarm sequences in a variable manner, and then analyzes industrial alarms, providing analysis results to assist operators in decision-making.

[0004] Technical solution: This invention includes: deploying master nodes and acquisition nodes; generating alarm data; the master node and acquisition nodes operating on the alarm data; generating alarm sequences: concatenating the alarm variables and alarm types of each alarm in Ra1, forming a new sequence in the original order and placing it in res1; forming a new sequence of the alarm types of each alarm in Rb1 in the original order and placing it in res2, while recording which alarm corresponds to the first and last of the new sequence, updating the value of start1 to start1+l1, and updating the value of start2 to start2+l2; calculating the support of alarm sequences; calculating the similarity of alarm sequences.

[0005] The alarm sequence is generated as follows:

[0006] Initialize two windows W1 and W2, with window widths of s1 and s2 respectively, maximum window widths of ms1 and ms2 respectively, and each movement distance of the windows of l1 and l2 (l1≤ms1, l2≤ms2), and the starting positions of the two windows satisfy start1=start2=a1;

[0007] Record the alarms within windows W1 and W2 respectively, and the results are Ra1 and Rb1; let the size of Ra1 be t1 and the size of Rb1 be t2, and the calculation method of the alarm window metric value is as follows:

[0008]

[0009]

[0010] In equations (1) and (2), totali represents the number of alarms whose alarm type is i within the window, and T1 and T2 represent the total number of alarm types in windows W1 and W2 respectively; update the value of start1 to start1+l1, and update the value of start2 to start2+l2;

[0011] Move windows W1 and W2 to the positions of start1 and start2 respectively, record the alarms located in the two windows respectively, calculate new values ny1 and ny2 according to equations (1) and (2) at the same time, and update the settings of windows W1 and W2.

[0012] The settings of the windows W1 and W2 are updated in the following manner:

[0013] If ny1>y1, update each movement distance of window W1 to l1=l1-1, where the minimum value of l1 is 1, and update the width of window W1 to s1=s1+2; if ny2>y2, update each movement distance of window W2 to l2=l2-1, where the minimum value of l2 is 1;

[0014] If ny1=y1, the settings of window W1 remain unchanged; if ny2=y2, the settings of window W2 also remain unchanged;

[0015] If ny1<y1, update each movement distance of window W1 to l1=min(l1+2, ms1), and update the width of window W1 to s1=s1-2; if ny2<y2, update each movement distance of window W2 to l2=min(l2+2, ms2);

[0016] Repeat the above steps until windows W1 and W2 respectively contain the last alarm in A, so as to obtain the final alarm sequence sets res1 and res2.

[0017] The calculation of alarm sequence support is as follows: For alarm sequences B and C in res1, if all elements appearing in B also appear in C, and the order of appearance is consistent, then sequence C supports sequence B. Let num1 be the number of sequences in the sequence set res1 that support B, and num2 be the total number of alarms in the sequence set. Then the support of sequence B is...

[0018]

[0019] In calculating the support of alarm sequences, a threshold min_sup is set, and alarm patterns are generated in the following manner:

[0020] 1) Calculate the support of each alarm in res1, remove alarms with support less than min_sup, and obtain a pattern set of length 1;

[0021] 2) Concatenate the alarm patterns of length 1 obtained in the previous step to obtain an alarm pattern set of length 2. Calculate the support of each alarm pattern and remove alarm patterns with support less than min_sup.

[0022] 3) For alarm patterns with a length greater than or equal to 2, generate longer alarm patterns by combining patterns with a common prefix, calculate their support, and remove alarm patterns with a support less than min_sup.

[0023] 4) Repeat step 3) until no new alarm pattern sequence can be generated.

[0024] The calculation of alarm sequence similarity specifically involves: alarm sequence set res2 = {R1, R2, R3, ..., R...} m} satisfies R i = {ri1, ri2, ..., rin} (i∈m), set a level for each rin, and calculate the similarity sim between alarm sequences in res2; the specific calculation method is as follows: the initial value of sim is 0, traverse the two alarm sequences, if the alarms at corresponding positions of the two alarm sequences are the same, then increment sim by 1, otherwise sim remains unchanged.

[0025] The deployment of the master node and the data acquisition node specifically involves the master node periodically detecting the online status of the data acquisition node while maintaining time synchronization between the data acquisition node and the master node.

[0026] The generation of alarm data specifically involves the master node formulating alarm rules based on the area where the data acquisition node is located, issuing different alarm rules for different data acquisition nodes, and the data acquisition node generating alarms according to the alarm rules.

[0027] The master node and the acquisition node operate on the alarm data. Specifically, the acquisition node compresses and deduplicates the alarm data according to the alarm rules and then uploads it to the master node at regular intervals. The master node arranges the alarm data of each acquisition node in the order of the time when each alarm was generated.

[0028] An industrial alarm analysis system, comprising:

[0029] Master node and data acquisition node deployment module: The master node periodically detects the online status of the data acquisition nodes, while maintaining time synchronization between the data acquisition nodes and the master node;

[0030] Alarm data generation module: The master node formulates alarm rules based on the area where the acquisition node is located, and issues different alarm rules for different acquisition nodes. The acquisition node generates alarms according to the alarm rules.

[0031] The main node and the acquisition node operate on alarm data: The acquisition node compresses and deduplicates the alarm data according to the alarm rules and uploads it to the main node at regular intervals. The main node arranges the alarm data of each acquisition node in the order of the time when each alarm was generated.

[0032] Alarm sequence generation module: Concatenate the alarm variables and alarm types of each alarm in Ra1, and form a new sequence in the original order and put it into res1; Form a new sequence of the alarm types of each alarm in Rb1 in the original order and put it into res2, and record which alarm corresponds to the first and last of the new sequence; Update the value of start1 to start1+l1, and update the value of start2 to start2+l2.

[0033] Alarm sequence support calculation module: For alarm sequences B and C in res1, if all elements appearing in B also appear in C, and the order of appearance is consistent, then sequence C supports sequence B. Let num1 be the number of sequences in the sequence set res1 that support B, and num2 be the total number of alarms in the sequence set. Then the support of sequence B is:

[0034]

[0035] Alarm sequence similarity calculation module: Alarm sequence set res2 = {R1, R2, R3, ..., R...} m} satisfies R i = {ri1, ri2, ..., rin} (i∈m), assign a level to each rin, and calculate the similarity sim between alarm sequences in res2.

[0036] Beneficial effects: This invention can generate alarm sequences in a variable manner while ensuring the true timing of alarms, thereby analyzing industrial alarms, providing analysis results, and assisting operators in decision-making; by setting a master node and acquisition nodes and synchronizing their time, it solves the problem of the difficulty in recording the true timing of alarm generation; by sending rules from the master node to the acquisition nodes, it reduces the amount of alarm data; and by defining an alarm sequence generation method specifically for the characteristics of industrial control alarms, it can effectively handle the dynamic changes of alarms. Attached Figure Description

[0037] Figure 1 This is a deployment architecture diagram of the present invention;

[0038] Figure 2 This is a flowchart of the present invention. Detailed Implementation

[0039] The invention will now be further described with reference to the accompanying drawings.

[0040] like Figure 1 and Figure 2 As shown, the industrial alarm analysis method of the present invention includes the following steps:

[0041] Step 1: Deploy the master node and acquisition nodes. The acquisition nodes are deployed in the industrial field to communicate with PLCs and other equipment and collect process data. The master node is deployed in the monitoring layer. The master node periodically detects the online status of the acquisition nodes and keeps the acquisition nodes and master node synchronized in time. The master node ensures the time synchronization of each acquisition node and detects whether the acquisition nodes are online normally, so as to ensure the accurate timing of alarms.

[0042] Step 2: Generate alarm data: The master node formulates alarm rules based on the area where the data acquisition node is located, and issues different alarm rules to different data acquisition nodes. The data acquisition node generates alarms according to the alarm rules, including attributes such as alarm variables, alarm type, and alarm time.

[0043] Step 3: Operations of alarm data by the master node and acquisition nodes: The acquisition nodes compress and deduplicate the alarm data according to the alarm rules and then upload it to the master node at regular intervals. The master node arranges the alarm data of each acquisition node in the order of the time when each alarm was generated.

[0044] Step 4: Let A = {a1, a2, a3, ..., a...} m} represents the alarm data generated by the system, where each data entry a m Each alarm includes attributes such as alarm variable val, alarm type type, and alarm time time, and generates an alarm sequence according to the following method:

[0045] S41. Initialize two windows W1 and W2, with window widths s1 and s2 respectively, maximum window widths ms1 and ms2 respectively, and each movement distance of the windows being l1 and l2 (l1≤ms1, l2≤ms2), and the starting positions of the two windows satisfy start1=start2=a1.

[0046] S42. Record the alarms in windows W1 and W2 respectively, with the results being Ra1 and Rb1. Let the size of Ra1 be t1 and the size of Rb1 be t2, and the calculation method of the alarm window metric value is as follows:

[0047]

[0048]

[0049] In Equation (1) and Equation (2), totali represents the number of alarms with the alarm type i in the window, and T1 and T2 represent the total number of alarm types in windows W1 and W2 respectively. Update the value of start1 to start1+l1, and update the value of start2 to start2+l2. Splice the alarm variable and alarm type of each alarm in Ra1, and form a new sequence in the original order to place into res1. Form a new sequence with the alarm type of each alarm in Rb1 in the original order to place into res2, and record which alarm the start and end of the new sequence correspond to at the same time.

[0050] S43. Move windows W1 and W2 to the positions of start1 and start2 respectively, record the alarms located in the two windows respectively, calculate new values ny1 and ny2 according to Equation (1) and Equation (2), and update the settings of windows W1 and W2 in the following manner:

[0051] 1) Adjust the alarm window when the alarm window metric value increases: if ny1>y1, update each movement distance of window W1 to l1=l1-1, where the minimum value of l1 is 1, and update the width of window W1 to s1=s1+2; if ny2>y2, update each movement distance of window W2 to l2=l2-1, where the minimum value of l2 is 1;

[0052] 2) Adjust the alarm window when the alarm window metric value remains unchanged: if ny1=y1, keep the settings of window W1 unchanged. If ny2=y2, also keep the settings of window W2 unchanged;

[0053] 3) Adjust the alarm window when the alarm window metric value decreases: if ny1<y1, update each movement distance of window W1 to l1=min(l1+2, ms1), and update the width of window W1 to s1=s1-2; if ny2<y2, update each movement distance of window W2 to l2=min(l2+2, ms2);

[0054] 4) Repeat step 3) until windows W1 and W2 each contain the last alarm from A, thus obtaining the final alarm sequence sets res1 and res2.

[0055] Step 5: Calculate the support of alarm sequences: For alarm sequences B and C in res1, if all elements appearing in B also appear in C, and the order of appearance is consistent, then sequence C supports sequence B. Let num1 be the number of sequences in the sequence set res1 that support B, and num2 be the total number of alarms in the sequence set. Then the support of sequence B is... Set the threshold min_sup and generate alarm modes as follows:

[0056] 1) Calculate the support of each alarm in res1, remove alarms with support less than min_sup, and obtain a pattern set of length 1;

[0057] 2) Concatenate the alarm patterns of length 1 obtained in the previous step to obtain an alarm pattern set of length 2. Calculate the support of each alarm pattern and remove alarm patterns with support less than min_sup.

[0058] 3) For alarm patterns with a length greater than or equal to 2, generate longer alarm patterns by combining patterns with a common prefix, and calculate their support. Then, remove alarm patterns with a support less than min_sup.

[0059] 4) Repeat step 3) until no new alarm pattern sequence can be generated.

[0060] Step Six: Calculate the similarity of alarm sequences: Alarm sequence set res2 = {R1, R2, R3, ..., R...} m} satisfies R i = {ri1, ri2, ..., rin} (i∈m), and set a level for each rin. The similarity sim between alarm sequences in res2 is calculated as follows: the initial value of sim is 0. Iterate through the two alarm sequences. If the alarms at corresponding positions of the two alarm sequences are the same, then increment sim by 1. Otherwise, sim remains unchanged.

[0061] Obtain the similarity matrix, cluster res2 using the similarity matrix, and record the cluster centers of each class. When a new alarm sequence is generated, calculate the distance between the new alarm sequence and each cluster center, select the class with the closest distance as the class of the new alarm sequence, and update the cluster centers of that class.

[0062] The industrial alarm analysis system of the present invention includes:

[0063] Master node and data acquisition node deployment module: The master node periodically detects the online status of the data acquisition nodes and keeps the time of the data acquisition nodes and the master node synchronized. The master node ensures the time synchronization of each data acquisition node and detects whether the data acquisition nodes are online normally, so as to ensure the accurate timing of alarms.

[0064] Alarm data generation module: The master node formulates alarm rules based on the area where the acquisition node is located, and issues different alarm rules for different acquisition nodes. The acquisition node generates alarms according to the alarm rules, including attributes such as alarm variables, alarm type, and alarm time.

[0065] The master node and the acquisition node operate on alarm data: The acquisition node compresses and deduplicates the alarm data according to the alarm rules and then uploads it to the master node at regular intervals. The master node arranges the alarm data of each acquisition node in the order of the time when each alarm was generated.

[0066] Alarm sequence generation module: Concatenates the alarm variables and alarm types of each alarm in Ra1, forming a new sequence in the original order and placing it in res1. Forms a new sequence of alarm types for each alarm in Rb1 in the original order and places it in res2, while recording the alarm number corresponding to the beginning and end of the new sequence. Updates the value of start1 to start1+l1, and updates the value of start2 to start2+l2.

[0067] Alarm sequence support calculation module: For alarm sequences B and C in res1, if all elements appearing in B also appear in C, and the order of appearance is consistent, then sequence C supports sequence B. Let num1 be the number of sequences in the sequence set res1 that support B, and num2 be the total number of alarms in the sequence set. Then the support of sequence B is:

[0068]

[0069] Alarm sequence similarity calculation module: Alarm sequence set res2 = {R1, R2, R3, ..., R...} m} satisfies R i = {ri1, ri2, ..., rin} (i∈m), and set a level for each rin. The similarity sim between alarm sequences in res2 is calculated as follows: the initial value of sim is 0. Iterate through the two alarm sequences. If the alarms at corresponding positions of the two alarm sequences are the same, then increment sim by 1. Otherwise, sim remains unchanged.

Claims

1. An industrial alarm analysis method, characterized in that, comprising: deploying a master node and an acquisition node; generating alarm data; operating, by the master node and the acquisition node, on the alarm data; generating an alarm sequence: splicing the alarm variable and alarm type of each alarm in Ra1, forming a new sequence in the original order and placing the new sequence into res1; forming a new sequence with the alarm types of each alarm in Rb1 in the original order and placing the new sequence into res2, meanwhile recording the serial numbers of the alarms corresponding to the start and end of the new sequence, updating the value of start1 to start1+l1, and updating the value of start2 to start2+l2; the alarm sequence is generated in the following manner: initializing two windows W1 and W2, with window widths being s1 and s2 respectively, maximum window widths being ms1 and ms2 respectively, and each window movement distance being l1 and l2, where l1≤ms1, l2≤ms2, and the starting positions of the two windows satisfy start1=start2=a1; recording the alarms in the windows W1 and W2 respectively, with the results being Ra1 and Rb1; letting the size of Ra1 be t1 and the size of Rb1 be t2, the calculation method of the alarm window metric value is as follows: y1 （1） y2 （2） in formula (1) and formula (2), totali represents the number of alarms whose alarm type type is i in the window, T1 and T2 represent the total number of alarm types in the windows W1 and W2 respectively; updating the value of start1 to start1+l1, and updating the value of start2 to start2+l2; moving the windows W1 and W2 to the positions of start1 and start2 respectively, recording the alarms located in the two windows respectively, calculating new values ny1 and ny2 according to formula (1) and formula (2), and updating the settings of the windows W1 and W2; the settings of the windows W1 and W2 are updated in the following manner: if ny1>y1, updating the each movement distance of the window W1 to l1=l1-1, where the minimum value of l1 is 1, and updating the width of the window W1 to s1=s1+2; if ny2>y2, updating the each movement distance of the window W2 to l2=l2-1, where the minimum value of l2 is 1; if ny1=y1, keeping the settings of the window W1 unchanged; if ny2=y2, keeping the settings of the window W2 unchanged as well; if ny1<y1, updating the each movement distance of the window W1 to l1=min(l1+2, ms1), and updating the width of the window W1 to s1=s1-2; if ny2<y2, updating the each movement distance of the window W2 to l2=min(l2+2, ms2); repeating the above steps until the windows W1 and W2 each contain the last alarm in A, to obtain the final alarm sequence sets res1 and res2; calculating the support degree of the alarm sequence; calculating the similarity of the alarm sequence.

2. The industrial alarm analysis method according to claim 1, characterized in that, The calculation of alarm sequence support is as follows: For alarm sequences B and C in res1, if all elements appearing in B also appear in C, and the order of appearance is consistent, then sequence C supports sequence B. Let num1 be the number of sequences in the sequence set res1 that support B, and num2 be the total number of alarms in the sequence set. Then the support of sequence B is sup = 100%.

3. The industrial alarm analysis method according to claim 2, characterized in that, In the step of calculating the support degree of the alarm sequence, a threshold min_sup is set, and an alarm pattern is generated in the following manner: 1) calculating the support degree of each alarm in res1, removing the alarms whose support degree is less than min_sup, to obtain a pattern set with a length of 1; 2) Concatenate the alarm patterns of length 1 obtained in the previous step to obtain an alarm pattern set of length 2. Calculate the support of each alarm pattern and remove alarm patterns with support less than min_sup. 3) For alarm patterns with a length greater than or equal to 2, generate longer alarm patterns by combining patterns with a common prefix, calculate their support, and remove alarm patterns with a support less than min_sup. 4) Repeat step 3) until no new alarm pattern sequence can be generated.

4. The industrial alarm analysis method according to claim 2, characterized in that, The calculation of alarm sequence similarity specifically involves: alarm sequence set res2 = {R1, R2, R3, ..., R...} m } Satisfies R i ={ri1, ri2, ..., rin}, i∈[1,m], and R is a denoted R i Each element in the algorithm is assigned a level, and the similarity sim between alarm sequences in res2 is calculated. The specific calculation method is as follows: the initial value of sim is 0. The two alarm sequences are traversed. If the alarms at corresponding positions of the two alarm sequences are the same, sim is incremented by 1. Otherwise, sim remains unchanged.

5. The industrial alarm analysis method according to claim 1, characterized in that, The deployment of the master node and the data acquisition node specifically involves the master node periodically detecting the online status of the data acquisition node while maintaining time synchronization between the data acquisition node and the master node.

6. The industrial alarm analysis method according to claim 5, characterized in that, The generation of alarm data specifically involves the master node formulating alarm rules based on the area where the data acquisition node is located, issuing different alarm rules for different data acquisition nodes, and the data acquisition node generating alarms according to the alarm rules.

7. The industrial alarm analysis method according to claim 6, characterized in that, The master node and the acquisition node operate on the alarm data. Specifically, the acquisition node compresses and deduplicates the alarm data according to the alarm rules and then uploads it to the master node at regular intervals. The master node arranges the alarm data of each acquisition node in the order of the time when each alarm was generated.

8. An industrial alarm analysis system, applicable to the industrial alarm analysis method according to any one of claims 1 to 7, characterized in that, include: Master node and data acquisition node deployment module: The master node periodically detects the online status of the data acquisition nodes, while maintaining time synchronization between the data acquisition nodes and the master node; Alarm data generation module: The master node formulates alarm rules based on the area where the acquisition node is located, and issues different alarm rules for different acquisition nodes. The acquisition node generates alarms according to the alarm rules. The main node and the acquisition node operate on alarm data: The acquisition node compresses and deduplicates the alarm data according to the alarm rules and then uploads it to the main node at regular intervals. The main node arranges the alarm data of each acquisition node in the order of the time when each alarm was generated. Alarm sequence generation module: Concatenates the alarm variables and alarm types of each alarm in Ra1, and puts them into res1 in the original order to form a new sequence; puts the alarm types of each alarm in Rb1 into res2 in the original order, and records which alarm corresponds to the first and last of the new sequence; updates the value of start1 to start1+l1, and updates the value of start2 to start2+l2. Alarm sequence support calculation module: For alarm sequences B and C in res1, if all elements appearing in B also appear in C, and the order of appearance is consistent, then sequence C supports sequence B. Let num1 be the number of sequences in the sequence set res1 that support B, and num2 be the total number of alarms in the sequence set. Then the support of sequence B is sup = 100%; Alarm sequence similarity calculation module: Alarm sequence set res2={R1, R2, R3, ..., R m } Satisfies R i ={ri1, ri2, ..., rin}, i∈[1,m], and R is a denoted R i Each element in the algorithm is assigned a level, and the similarity sim between alarm sequences in res2 is calculated.

Images