Leak test method and apparatus

[0002] The present invention relates to a leak test method and apparatus which are used to check various containers or vessels for leaks.

[0003] In the manufacture of products or parts required to be free of leaks, it is general practice in the prior art to inspect them in succession on production lines and compare the inspection data with preset reference values to determine if they are leak-free or not. With the conventional method, products under test for leaks (hereinafter referred to as works), such as vessels or containers, are tested successively on production lines by introducing thereinto compressed gas and making a check to see if the compressed gas leaks out thereof or if the amount of leakage is smaller than a reference value. A leak testing apparatus that has been used in the past is a differential pressure type leak tester that detects leaks of the compressed gas from the work based on variations in the pressure difference between a master tank (hereinafter referred to as a master) and the work.

[0004] Referring now to FIGS. 1 and 2, the differential pressure type leak tester will be described in brief. FIG. 1 is a piping diagram of a prior art example and FIG. 2 a graph showing how the abovementioned pressure difference changes with the lapse of time.

[0005] In FIG. 1, a pressurized gas source 10 is piped via a pressure regulation valve 12 and a three-way electromagnetic valve 14 to electromagnetic valves 16W and 16M leading to a work 22W and a master 22M, respectively. Connected between the work 22W and the master 22M is a differential pressure sensor 18. The three-way electromagnetic valve 14 normally vents the conduits on both its sodes to the atmospheric pressure, but in response to a drive voltage, conducts both the conduits to each other.

[0006] The actual leak test starts with opening the valves 12, 14, 16W, and 16M to introduce pressurized gas into the work 22W and the master 22M from the pressurized gas source 10, followed by regulating the regulation valve 12 to set a pressure gauge 13 at a desired testing pressure. Next, the electromagnetic valves 16W and 16M are closed and after a certain elapsed time the pressure difference between the work 22W and the master 22M is measured by the differential pressure sensor 18. If the work 22W leaks, the pressure on the work side becomes gradually lower than the pressure on the master side. The detected pressure difference is compared with a threshold value L_thin a decision part 19. When the detected pressure difference is smaller than the threshold value L_th, it is decided that the work 22W is leak-free or that the leak is negligibly small, and when the pressure difference is larger than the threshold value L_th, the work 22W is decided to be leaky. In this case, if the work 22W does not leak, the pressure difference ought to be zero. In practice, however, the pressure difference frequently develops even if the work 22W does not leak. Such a situation might be the case wherein when the temperature of the work 22W heated on the production line is still higher than the temperature of the master 22M (room temperature, for instance), the measurement of the pressure difference is started and thereafter the pressure in the work 22W drops as its temperature is gradually reduced toward that of the master 22M as a result of thermal radiation. Even if no leaks are detected, the pressure difference usually varies due to temporal changes in the temperature difference between the work 22W and the master 22M. A description will be given below, with reference to FIG. 2, of how the pressure difference varies during measurements.

[0007] In FIG. 2, the curve 2A indicates the pressure difference when the work 22W does not leak and the curve 2B the pressure difference when the work 22W leaks. As shown, the pressure difference develops at and after time t_swhen the electromagnetic valves 16W and 16M depicted in FIG. 1 are closed, and thereafter the pressure difference varies unstably until time t_a. This is primarily due to a shock resulting from the closing of the electromagnetic valves 16W and 16M. Then, the pressure difference undergoes substantially linear variation during the time interval from t_ato t_b. The reason for this is, for example, that the temperature of the pressurized gas introduced into the work 22W as mentioned above is gradually lowered. And the pressure difference varies in a smooth curve from time t_bto t_c, because the cooling rate of the pressurized gas reduces as its temperature approaches room temperature.

[0008] After time t_c, the pressure difference does not vary when the work is free of leaks, but in the case of a leaky work, the pressure difference further undergoes linear variation. This period will hereinafter be referred to as a “stabilization period.” Since the gas temperature in either of the work and the master is considered to be equal to room temperature during this period, pressure difference variation per unit time is in proportion to the amount of leakage (cm³/sec). The amount of leakage that is intended to be detected is appreciably small, and it can be regarded as substantially constant from time t_sto the stabilization period after time t_c. Through utilization of this phenomenon it is possible to decide that the work is leak-free or leaky, depending upon whether the detected pressure difference variation per unit time after time t_cis close to zero or not.

[0009] However, the conventional method is time-consuming since the measurement cannot be started until after time t_c. A differential pressure type leak tester that has been proposed as a solution to this problem is disclosed in Japanese Patent Application Laid-Open Gazette No. 4-506262. With this leak tester, the pressure difference variation per unit time is premeasured using a leak-free, non-defective work in the time interval t_ato t_bdepicted in FIG. 2 during which the pressure difference varies linearly after time t_awhen it ceases from sharp variations. The abovementioned period from time t_ato t_bwill hereinafter be referred to as a “measurement period.”

[0010] In the actual leak test, pressure difference variation per unit time, Δp/Δt, is detected in the measuring period t_ato t_b, and is compared with the pressure difference variation premeasured using the leak-free, non-defective work. It is possible to decide that the product under test is leak-free or leaky, depending upon whether the pressure difference values compared are nearly equal or not. This enables the measurement to start prior to time t_cand hence permits reduction of the measurement time.

[0011] This prior art method is effective in reducing the measurement time but requires the preparation of a non-defective work. To solve this problem there has been proposed such a method as described below.

[0012] To begin with, two pressure differences p₁and p₂are measured, using a sample work, at a predetermined time interval Δt (two seconds, for instance) in the measurement period from time t_ato t_bshown in FIG. 3, and pressure difference variation per unit time, δp₁=(p₂−p₁)/Δt=δp₁/δt, is calculated from the two pressure differences p₁and p₂. This is followed by calculating pressure difference variation per unit time, δp₂, from two pressure differences p₃and p₄similarly measured at the predetermined time interval Δt in the stabilization period after time t_c. The pressure difference variations per unit time, δp₁and δp₂, are calculated as follows:

[0013] δp₁=(p₂−p₁)/Δt=Δp₁/Δt

[0014] δp₂=(p₄−p₃)/Δt=Δp₂/Δt

[0015] The pressure difference variation δp₂in the stable period can be regarded as a variation attributable to leaks of the work. The pressure difference variation δp₁in the measurement period can be regarded as the sum of the abovementioned variation δp₂and the amount of variation which occurs even in the absence of leaks (which variation will hereinafter referred to as the “amount of drift” or simply as “drift”). Thus, the drift p_dcan be obtained by subtracting the pressure difference variation δp₂from δp₁. In this way, the drift, p_d=δp₁−δp₂, is precalculated.

[0016] In the actual leak test, the pressure difference variation per unit time is calculated first on the work under test during the time interval between t₁to t₂, and the abovementioned drift is subtracted from the pressure difference variation to obtain a correction result. Finally, the presence or absence of leaks is decided, depending on whether the correction result falls within a predetermined range of values.

[0017] This method permits correction using one sample work (without distinction of leakage), and hence does not involve the preparation of leak-free, non-defective work. Further, the pressure difference variation needs only to be measured on the sample work until the stabilization period, and the measurements can be made on other works at earlier timing to determine if they are leak-free or leaky.

[0018] With the prior art, however, it is necessary to measure the pressure differences p₁and p₂in the measurement period and the pressure differences p₃and p₄in the stabilization period by a single differential pressure sensor as depicted in FIG. 3. In general, there is a limit to the number of digits of each value that a single differential pressure sensor can measure with accuracy. The minimum resolution of a differential pressure sensor capable of measuring large pressure differences is more coarse than the minimum resolution of a differential pressure sensor for measuring small pressure differences. For this reason, a small pressure difference (for example, 0.1 mmHg) cannot be detected with high accuracy by a differential pressure sensor for large pressure differences (for example, 100 mmHg). Hence, the prior art encounters difficulty in calculating drift with high accuracy.

[0019] On the other hand, the proposal made in Japanese Patent Application Laid-Open Gazette No. 4-506262 mentioned above involves the preparation of a leak-free, non-defective work.

[0020] It is therefore an object of the present invention to provide a leak test method and apparatus with which it is possible to make an accurate determination whether or not pressurized gas introduced in a work leaks out therefrom, without involving the preparation of a leak-free, non-defective sample work and the measurement of a large pressure difference.

[0021] According to a first aspect of the present invention, there are provided a leak test method and apparatus in which: pressurized gas is introduced into a master and a work; the pressure difference therebetween is measured by a differential pressure sensor; pressure difference variation per unit time is calculated from the measured pressure difference variation; the amount of drift is subtracted from the calculated pressure difference variation to obtain the amount of leakage; and based on the leakage, it is determined whether or not the pressurized gas leaks out from the work. In this instance, first and second pressure difference variations per unit time are each measured immediately after a pressure equilibrium is established between the master and the work.

[0022] With the method and apparatus according to the first aspect of the invention, the absolute value of the pressure difference to be detected for the detection of the first and the second pressure difference variation can be held small by the pressure equilibration between the master and the work during the time interval from when the first pressure difference variation is measured in the measurement period and the second pressure difference variation is measured in the stabilization period. This allows the use of a fine-resolution, sensitive differential pressure sensor, providing increased accuracy in making drift calculation. Besides, the drift calculation can be made using a given work instead of using a leak-free, non-defective work.

[0023] According to a second aspect of the present invention, the drift obtained by a drift calculation process in the above is stored in a drift table. In a leak test mode, the data stored in the drift table is used to calculate a moving average of drift, then a third pressure difference variation is measured on the work under test, then the moving average drift is subtracted from the third pressure difference variation to obtain leakage, and when the leakage is equal to or smaller than a predetermined threshold value, the third pressure difference variation is stored as drift in the drift table.

[0024] With the above scheme, an initial drift is stored first in the drift table and then the third pressure difference variation is stored therein only when the work under test is decided to be free of leaks. The leakage is calculated from the third pressure difference variation and the moving average calculated base on the data stored in the drift table. Since the third pressure difference variation is susceptible to the influence of a temperature change or the like, the moving average varies with a temperature change or the like, accordingly. It is possible, therefore, to accurately determine if the work leaks out therefrom the pressurized gas independently of temperature changes.

[0025] According to a third aspect of the present invention, the drift is given, by a linear approximate expression p_d=K(T_A−T_B)+C, as a linear function of the difference between the temperature T_Aof the work and the temperature T_Bof a jig. The constants C and K are obtained in advance on various types of works and stored in a memory. In the leak test mode, the constants C and K corresponding to the type of work under test are read out of the memory and are used to calculate drift from the detected work and jig temperatures T_Aand T_Bby the linear approximate expression.

[0026] With this scheme, the works of any types with the constants C and K already measured thereon can be tested for leaks at any temperatures even after they are heat-processed on production lines.

[0027] FIG. 1 is a block diagram illustrating the basic configuration of a conventional differential pressure type leak test apparatus;

[0028] FIG. 2 is a graph showing changes in the pressure difference with time in the cases of leak-free and leaky works in FIG. 1;

[0029] FIG. 3 is a graph for explaining the principle of leakage measurement in FIG. 1;

[0030] FIG. 4 is a block diagram illustrating the leak test apparatus according to the present invention;

[0031] FIG. 5 is a block diagram depicting a control 100 in FIG. 4 formed by a computer;

[0032] FIG. 6 is a flowchart showing a leak test procedure in a first embodiment of the leak test method according to the present invention;

[0033] FIG. 7 is a flowchart showing a pass/fail decision process in the flowchart of FIG. 6;

[0034] FIG. 8 is a flowchart showing a drift calculation process in the flowchart of FIG. 6;

[0035] FIG. 9A is a graph showing changes in the pressure difference in the drift calculation procedure based on the principle of the present invention;

[0036] FIG. 9B is a graph showing changes in the pressure difference in a conventional drift calculation procedure;

[0037] FIG. 10A is a graph showing changes in the pressure difference, for explaining a modified form of the drift calculation procedure in the first embodiment;

[0038] FIG. 10B is a graph showing changes in the pressure difference in a drift calculation procedure which is actually used;

[0039] FIG. 10C is a graph showing changes in the pressure difference variation in the first embodiment in terms of an exponential function;

[0040] FIG. 11 is a graph showing changes in the pressure difference in a modified form of the drift calculation procedure in the first embodiment;

[0041] FIG. 12 is a flowchart showing a modified form of the drift calculation procedure in the first embodiment;

[0042] FIG. 13 is a flowchart showing a leak test procedure in a second embodiment of the present invention;

[0043] FIG. 14 is a flowchart showing an initial drift calculation procedure in the flowchart of FIG. 13;

[0044] FIG. 15A is a diagram depicting an example of a drift table;

[0045] FIG. 15B is a diagram depicting updating of the drift table;

[0046] FIG. 16 is a flowchart showing pass/fail decision process in the flowchart of FIG. 13;

[0047] FIG. 17 is a diagram showing the range of correction results which are stored in the drift table;

[0048] FIG. 18 is a diagram depicting the relationship between temperature and drift measured for each type of work;

[0049] FIG. 19 is a block diagram illustrating a leak test apparatus according to a third embodiment of the present invention;

[0050] FIG. 20 is a flowchart depicting a leak test procedure in the third embodiment;

[0051] FIG. 21 is a flowchart depicting a drift linear approximation procedure in the flowchart of FIG. 20; and

[0052] FIG. 22 is a flowchart depicting a pass/fail decision process in the flowchart of FIG. 20.

[0053] With reference to the accompanying drawings, embodiments of the present invention will hereinafter be described.

[0054] Referring first to FIGS. 4 to 9, a first embodiment of the present invention will be described. FIG. 4 depicts the pipe arrangement of the leak test apparatus of this embodiment and a functional block configuration of a control 100 and FIG. 5 a functional block configuration of the control 100 formed by a microcomputer. FIGS. 6, 7 and 8 are flowcharts showing a leak test procedure, a decision procedure, and a drift calculation procedure, respectively. FIGS. 9A and 9B are graphs showing changes in pressure differences that are measured in the procedures depicted in FIGS. 6 to 8.

[0055] In FIG. 4 the pressurized gas source 10 is connected via the pressure regulating valve 12 and the three-way electromagnetic valve 14 to the electromagnetic valve 16W and the electromagnetic valve 16M by branch lines 15W and 15M, respectively. The electromagnetic valves 16W and 16M are connected by conduits 17W and 17M to the work 22W and the 22M, respectively. Connected between the work 22W and the master 22M is the differential pressure sensor 18. A thermometer 24 measures the temperature of the work 22W and room temperature. The differential pressure sensor 18 is connected between the conduits 17W and 17M to detect the pressure difference therebetween. According to the present invention, an electromagnetic valve 20 is connected between the conduits 17W and 17M in parallel relation to the differential pressure sensor 18.

[0056] The control 100 has, as its functional blocks, a valve control part 26, a drift calculating part 31, a leakage calculating part 32, a pass/fail decision part 33, and a sequence control part 34. The valve control part 26 drives the electromagnetic valves 14, 16W, 16M, and 20 to open and close them under the control of the sequence control part 34. The drift calculating part 31 calculates an initial drift amount on the work 22W on the basis of pressure differences detected by the differential pressure sensor 18 several times in drift calculating mode under the control of the sequence control part 34. The leakage calculating part 32 calculates the amount of leakage by correcting the detected initial pressure difference variation with the drift amount in a leak test mode under the control of the sequence control part 34. The pass/fail decision part 33 compares the absolute value of the calculated drift with a reference value, and decides that the product under test is leak-free or leaky, depending upon whether the absolute value is smaller or larger than the reference value. Under the control of a leak test program the sequence control part 34 decides whether to execute the drift calculating mode or the leak test mode according to the detected temperature the work under test and its type and, at the timing sequence corresponding to the mode decided to execute, applies an electromagnetic valve open/close control signal to the valve control part 26 or instructs the drift calculating part 31 and the leakage calculating part 32 to take in the detected pressure difference and calculate the amounts of drift and leakage, respectively.

[0057] The functional configurations of the valve control part 26, the drift calculating part 31, the leakage calculating part 32, the pass/fail decision part 33, and the sequence control part 34 of the control 100 depicted in FIG. 4 are computerized as shown in FIG. 5. Accordingly, the control 100 has a CPU 110, a ROM 102, a RAM 112, an input processing circuit 104, an output processing circuit 106, a display control circuit 114, a display unit 116, and a timer 118 which are components of ordinary computers. Incidentally, other function control parts for the mounting and exchange of the work 22W and the like are not shown for the sake of brevity.

[0058] The CPU 110 controls the electromagnetic valves 14, 16W, 16M, and 20 under the control of a leak test program stored in the ROM 102. The leak test program contains various programs for leak test, pass/fail decision, drift calculating, and other procedures in the flowcharts described later on. In the RAM 112, there are stored various kinds of data, such as the temperature of the work 22W, room temperature, the pressure difference, the pressure difference variation, the amount of drift, correction results, a drift table, a moving average, and a threshold value. The following embodiments will be described in connection with the case of storing various kinds of data in the RAM 112.

[0059] The input processing circuit is 104 A/D converts the detected signals from the differential pressure sensor 18 and the thermometer 24, and provides them to the CPU 110 or stores them in the RAM 112. The input processing circuit 104, which A/D converts the pressure difference signal from the differential pressure sensor 18, and the CPU 110, which uses the pressure difference data from the input processing circuit to calculate drift under the control of a program stored in the ROM 102, constitute the drift calculating part 31 in FIG. 4. The output processing circuit 106 outputs drive voltages for opening or closing the electromagnetic valves 14, 16W, 16M, and 20 in response to an open/close control signal sent from the CPU 110 via a bus 108 at predetermined timing. That is, the CPU 110, which outputs the valve control signal under the control of the program stored in the ROM 102, and the output processing circuit 106, which responds to the control signal to output the valve drive voltages, constitute the valve control part 26. The display control circuit 114 displays test conclusions on the display unit 116 based on display data sent from the CPU 110 or the like via the bus 108.

[0060] The final decision on whether the work 22W leaks or not is made in the control 100 described above. These processes are implemented by executing the procedures depicted in FIGS. 6 to 8. A detailed description will be given of each procedure.

[0061] In the leak test procedure depicted in FIG. 6, the following steps are executed:

[0062] Step S1: The procedure begins with initialization. This initialization step is, for example, to clear various kinds of data stored in the RAM 112 and set room temperature as a reference temperature.

[0063] Step S2: Next, the work 22W to be tested for leaks is mounted in position. Then, the electromagnetic valves 14, 16W, and 16M are opened to introduce pressurized gas into the work 22W and the master 22M, and the electromagnetic valves 16W, and 16M are closed.

[0064] Step S3: The type of the work 22W is detected, and the current room temperature is measured using the thermometer 24. In the case where the work 22W is of the same type as the previous one tested for leaks and a room temperature change falls within a predetermined range relative to the reference temperature (YES), the electromagnetic valve 20 is closed in the aforementioned measurement period and then the procedure goes to step S4; if NOT, then the procedure goes to step S8.

[0065] Step S4: The procedure waits for a predetermined time interval between the closure of the electromagnetic valve 20 and the measurement period.

[0066] Step S5: The pressure difference variation per unit time, δp, is calculated in the measurement period. More specifically, the pressure difference variation δp is calculated as described below. That is, the first pressure difference p₁between the work 22W and the master 22M is measured in the measurement period, and after the predetermined time Δt has elapsed, the second pressure difference p₂is measured. These two measurements used are to derive the respective pressure differences from the values of signals that are both provided from the differential pressure sensor 18. Based on the first-measured pressure difference p₁, the second pressure difference p₂measured after the predetermined time interval, and the predetermined time Δt, the pressure difference variation per unit time, δp, is computed by the following equation.

[0067] δp=(p₂−p₁)/Δt  (1)

[0068] Step S6: The pressure difference variation δp thus obtained is corrected using the amount of drift, and based on the result of correction, a pass/fail decision process is executed to make a determination whether or not the pressurized gas leaks out of the work 22W. This decision process will be described later on with reference to FIG. 7.

[0069] Step S7: The electromagnetic valve 14 is vented and the electromagnetic valves 16W, and 16M are opened to restore the atmospheric pressure in the work 22W and the master 22M. A check is made to see if there is another work to be tested, and if so, then the procedure returns to step S2, followed by repeating steps S2 to S6 described above. In the absence of such another work, the test is finished.

[0070] Step S8: When the type of work currently under test differs from the type of work tested, and/or if the measured room temperature falls far outside the predetermined range relative to the reference temperature, the drift p_dis calculated by the drift calculation process described later on with reference to FIG. 8 and is used to update the previously calculated drift, and the settings of the type of work and the reference temperature are updated, after which the procedure returns to step S3.

[0071] Turning now to FIG. 7, the pass/fail decision process of step S6 will be described below.

[0072] S61: To determine if the pressurized gas leaks out of the work 22W, a drift component is removed from the pressure difference variation δp measured in step S5 to obtain a corrected pressure difference variation p_L. In this case, it is assumed that the amount of drift was measured in step S8.

[0073] Step S62: Next, a check is made to see if the corrected pressure difference variation p_Lis within the range of a threshold value. The corrected pressure difference variation p_Lis calculated by the following equation based on the pressure difference variation per unit time, δp, and the drift p_dcalculated by the drift calculating process described below.

[0074] p_L=δp−p_d  (2)

[0075] The corrected pressure difference variation p_Lcorresponds to the amount of leakage, and hence it will hereinafter be referred to as leakage.

[0076] Step S63: If it is found in step S62 that the absolute value of the leakage p_Lis within the range of the threshold value L_th(YES), it is decided that the work under test is leak-free.

[0077] Step S64: If the absolute value of the leakage p_Lis outside the range of the threshold value L_th(NO), the work is decided as leaky.

[0078] The drift calculation process of step S8 in FIG. 6 is a process that the drift calculating part 31 in FIG. 4 executes. In the measurement period starting at the timing t_a, this process is executed using the procedure depicted in FIG. 8. In FIG. 9A, there are shown variations that the output from the differential pressure sensor 18 undergoes as the drift calculating process is executed. In this drift calculating process, the pressure difference variation δp=(p₂−p₁)/Δt is measured N times. The entire length t₀to t_2Nof the measurement sequence for the drift calculation is predetermined to make the first (n=1) measurement in the measurement period t_ato t_band the last (n=N, where N is an integer equal to or greater than 2) measurement in the stable period (after t_c). In the following description, the first (n=1) pressure difference variation will be referred to as the initial pressure difference variation and identified by δp_sand the last (n=N) pressure difference variation as the final pressure difference variation and identified by δp_e.

[0079] Step S81: An n-th pressure difference variation measurement begins with the application of a signal from the CPU 110 via the output processing circuit 106 to the electromagnetic valve 20 at time t_2(n−1)to open it to equilibrate the pressures in the work 22W and the master 22M, immediately followed by closing the electromagnetic valve 20. Then the pressure difference value (i.e. a voltage value) p_2n−1, which is output from the differential pressure sensor 18 at time t_2n−1, is scored in the RAM 112.

[0080] Step S82: Next, the procedure waits for the time interval Δt between time t_2n−1and the measurement time t_2n.

[0081] Step S83: At time t_2n, the pressure difference value p_2ndetected by the differential pressure sensor 18 is read into the input processing circuit 104. At this time, the n-th pressure difference variation is calculated as given below.

[0082] δp_n=(p_2n−p_2n−1)/Δt  (3)

[0083] where p_2n−1and p_2nare pressure differences measured n-th time at measurement times t_2n−1and t_2nset Δt apart.

[0084] In practice, the detected pressure value p_2n−1at time t_2n−1in step S81 need not be stored in the memory 112 and the detected pressure difference p_2n−1at time t_2n−1needs only to be reset to zero by digital processing. The subsequent detected pressure differences are obtained as variations relative to the true pressure difference at the reset time t_2n−1. Accordingly, the detected pressure difference at time t_2nin step S8 is the pressure difference variation (p_2n−p_2n−1) itself in Eq. (3).

[0085] Step S84: A check is made to see if the current measurement is the first (n=1) or final (n=N) measurement.

[0086] Step S85: Only when the measurement is the first or final one, the pressure difference variation is stored as δp_sor δp_ein the RAM 112.

[0087] Step S86: The above-described processes of steps S81 to S85 are repeated from the first (n=1) to final (n=N) measurements to determine whether or not the last (n=N) pressure difference variation has been measured. In this way, the initial pressure difference variation per unit time, δp_s=p₁, measured first and the pressure difference variation per unit time, δp_e=p_N, measured last are stored. Step S87: Where the current measurement timing is the last one t_2N, the drift p_dcontained in the initial pressure difference variation δp_sis computed. The drift p_dis obtained as the difference between the initial pressure difference variation δp_s=δp₁and the last pressure difference variation δp_e=p_Nboth stored in the RAM 112, and is given by the following equation.

[0088] p_d=δp_s−δp_e  (4)

[0089] The drift calculating process ends with the calculation of the initial drift p_d, and the procedure returns to step S3 of the pass/fail decision process depicted in FIG. 6, starting the process on the work 22W. The processing from steps S3 to S7 will hereinafter be called a leak test mode and the processing of step S8 a drift calculation mode.

[0090] In the prior art, the pressure difference between the work 22W and the master 22M increases with the lapse of time as depicted in FIG. 9B. In contrast thereto, according to the above-described first embodiment, the pressures in the work 22W and the master 22M are equilibrated by opening the electromagnetic valve 20 at the measurement of each pressure difference (time t₂, t₄, t₆, t₈, . . . ) in step S81 in FIG. 8. As a result, the pressure difference between the work 22W and the master 22M repeats increasing from zero as shown in FIG. 9A. That is, after executing the process of step S83 for obtaining the n-th pressure difference variation δp_n, the procedure returns via steps G84, G85 and S86 to step S81, wherein the electromagnetic valve 20 is opened for pressure equilibration to start measuring the nth-plus-one pressure difference variation δp_n+1and is closed thereafter. On this account, the absolute value of the pressure difference which is provided to the differential pressure sensor 18 does not become so large; therefore, a differential pressure sensor of high sensitivity can be employed. Since the pressure difference variations δp_sand δp_ecan be measured with higher accuracy, the initial drift p_dis also accurate. Further, the calculation of the drift p_ddoes not require the use of a leak-free, non-defective work and can be made using a given work. Incidentally, since the amounts calculated by Eqs. (1) to (4) are variations per unit time Δt, the value Δt can be freely chosen as long as it held constant. In the actual measurement, the value δp or δp_nmay be defined as given below, setting Δt=1 in Eq. (1) or (3) regardless of the constant value Δt used.

[0091] δp=p₂−p₁  (1′)

[0092] δp_n=p_2n−p_2n−1  (3′)

[0093] The same holds true with regard to all the embodiments described later on.

[0094] Modified Form 1: In the initial drift calculation procedure of FIG. 8, the pressure difference between the work 22W and the master 22M which is measured by the differential pressure sensor 18 slightly fluctuates due to shock by the opening of the electromagnetic valve 20 immediately thereafter. The influence of such fluctuations in the pressure difference can be avoided by a method shown in FIG. 10A, wherein the pressures are equilibrated by opening the electromagnetic valve 20 and kept in equilibrium by a fixed time interval bt during which to allow the pressure difference fluctuations to be damped sufficiently, and the electromagnetic valve 20 is closed, followed by measuring the pressure difference. This method prevents the influence of the pressure difference fluctuations by the opening of the electromagnetic valve 20 and hence provides increased accuracy in the detection of the pressure difference (and consequently the pressure difference variation per unit time and the amount of drift).

[0095] Modified Form 2: The measurement sequence in the drift calculating procedure depicted in FIG. 9A is its simplified version intended to facilitate a better understanding of the principle of the present invention. In the actual measurement sequence, measurement is repeated at time intervals T longer than the predetermined period Δt as depicted in FIG. 10B. At the start point of each measurement period T, the electromagnetic valve 20 is closed, and after elapsed time t_d, the reading of the first pressure difference value p_2n−1output from the differential pressure sensor is reset to zero at time t_2n−1. The second pressure difference p_2nis detected at time t_2n, the predetermined time Δt after time t_2n−1. After the detection of the pressure difference p_2nat time t_2n, the electromagnetic valve 20 is opened, and after the elapse of time Δt the electromagnetic valve 20 is closed as is the case with the modification described above in respect of FIG. 10A, and the same measurement is repeated in the next measurement cycle T. Since the reading of the pressure difference p_2n−1detected by the differential pressure sensor at time t_2n−1is reset to zero, the reading of the differential pressure sensor output at the next measurement time t_2nis also offset by the same value p_2n−1. Hence, the value of the true pressure difference variation δp_n=p_2n−p_2n−1is equal to the reading of the offset value p_2n. In the other embodiments described below, too, the measurement in every measurement cycle T in the initial drift measurement sequence is carried out at the same time as described above with respect to FIG. 10B.

[0096] Modified Form 3: The pressure difference variation at fixed time intervals Δt in the drift calculation procedure depicted in FIG. 8 exponentially decreases as indicated by the curve f in FIG. 10C, and converges to a virtually fixed value. The reason for this is that the temperature of the work connected, in heated state, to the conduit of the leak test apparatus gradually lowers due to thermal radiation. The value of convergence corresponds to the amount of leakage, and the exponential decrease in the pressure difference variation means an exponential reduction in the amount of drift attributable to an exponential decrease in the temperature difference between the work and the master—this means the pressure difference variation converges to zero with a decrease in the temperature difference. Now, approximate the curve f, for example, by the following exponential function and set an equation for computing the pressure difference p.

[0097] p=a×exp (−bt)+c  (5)

[0098] where a, b, and c are coefficients and t time. In the measurement period from ta to tb in FIG. 2, steps S81 to S86 in FIG. 8 are executed a plurality of times (concretely, three to five times) to measure the pressure difference between the work 22W and the master 22M. By applying the pressure difference p thus obtained to Eq. (5), the coefficients a, b, and c are determined. The coefficient c corresponds to a pressure difference variation by leaks of the work. Accordingly, even if works under test are those of the same type and whose initial temperatures are within a prescribed range with respect to a reference value, the constant c related to the amount of leakage differs from work to work. By applying desired points in time t_e−Δt and t_ein the stabilization period (after time tc in FIG. 2) to the above equation by which the coefficients a, b, and c are determined, it is possible to estimate and calculate pressure differences p_e−1and p_eat he time t_e−1=t_e−Δt and t_ein the stabilization period.

[0099] With this method, since the pressure differences in the stabilization period are obtained by several pressure difference measurements in the measurement period, the pressure difference variation per unit time in the stabilization period can be calculated. Thus, the time for calculating the amount of drift can be reduced significantly. In this case, too, the pressure equilibrium is established between the work 22W and the master 22M just prior to each measurement of the pressure difference in the measurement period; hence, the drift can be calculated with higher accuracy.

[0100] Modified Form 4: In the initial drift calculation procedure depicted in FIG. 8, a step of establishing the pressure equilibrium between the work 22W and the master 22M, closing the valve 20 and then measuring the pressure difference variation δp_nby the differential pressure sensor 18 is executed N times. By this, the pressure difference measured by the differential pressure sensor 18 can be obtained with high accuracy. However, the calculation of the initial drift p_d=δp_s−δp_e=p₁−δ_Nrequires only the first (n=1) and last (n=N) pressure difference variations δp₁=δp_sand δp_N=δp_eand does not require the values of the pressure difference variations at points in time where n is 2 to N−1. Accordingly, by additionally performing only once the step of establishing the pressure equilibrium between the work 22W and the master 22M at point in time t_e−Δt, that is, Δt before the last measurement time t_eas depicted in FIG. 11, the pressure difference can be detected more accurately than in the past. In FIG. 12, there is shown a flowchart for the drift calculation procedure in this case.

[0101] Step S81: In the measurement period, the electromagnetic valve 20 is opened first to establish the pressure equilibrium between the work 22W and the master 22M, then the valve 20 is closed, and the reading of the detected pressure difference value of the differential pressure sensor 18 is reset to zero.

[0102] Step S82: The procedure waits for the time interval Δt until the measurement time t_s.

[0103] Step S83: When the measurement time t_sis reached, the difference pressure p_sis measured, and a first differential pressure difference is calculated as the initial pressure difference variation δp_s. The method for obtaining this pressure difference variation is the same as in Step S5 in FIG. 6.

[0104] Step S84: Thereafter, the procedure waits until the stabilization period.

[0105] Step S85: In the stabilization period, the electromagnetic valve 20 is opened again to establish an equilibrium of pressures as in step S81, then the valve 20 is closed, and the reading of the detected pressure difference value of the differential pressure sensor 18 is reset to zero at time t_e−Δt.

[0106] Step S86: Following this, the procedure waits by the time interval Δt until the measurement time t_e.

[0107] Step S87: Δt the measurement time t_ethe pressure difference p_eis measured and a second pressure difference variation is calculated as the last pressure difference variation δp_e.

[0108] Step S88: The amount of drift is calculated based on the two pressure difference variations δp_sand δp_eobtained as mentioned above. The method for the drift calculation is the same as in step S88 in FIG.8.

[0109] With the drift calculating procedure shown in FIG. 12, the pressure difference detected by the differential pressure sensor 18 varies in such a curve as depicted in FIG. 11. That is, immediately before time t_s−Δt the valve 20 opened until then is closed and the pressure difference starts to increase, and at time t_s−Δt, the reading of the output value of the differential pressure sensor 18 is reset to zero, and at time t_sthe process of step S83 in FIG. 12 is executed to calculate the first pressure difference variation δp_s. Next, immediately before time t_e−Δt in the stabilization period after time t_c, the valve 20 is opened again to provide the pressure equilibrium and then the valve 20 is closed, and at time t_e−Δt, the reading of the output value of the differential pressure sensor 18 is reset again to zero, and at time t_sthe process of step S87 is executed to obtain the second pressure difference variation Δp_e. This procedure also provides an accurate value of drift because of the pressure equilibration by the opening of the valve 20 just prior to the calculation of the second pressure difference variation. Accordingly, the drift can be obtained with higher accuracy than in the conventional method described previously with respect to FIG. 3.

[0110] In the first embodiment, the drift p_dis calculated each time a measurement environmental condition (room temperature) changes by a fixed quantity (steps S3 and S8 in FIG. 6). In practice, air temperature changes continuously, but the drift p_dwhich is measured in the above embodiment follows the temperature change or the like on a stepwise basis. If the reference range of temperature variations is narrowed with a view to reducing the step size of the stepwise drift variation, the calculation of the drift p_dneeds to be done frequently and hence time-consuming, resulting in a decrease in the number of works that can be tested for leaks per unit time. From this point of view, a second embodiment performs a drift calculation more accurately responsive to temperature changes without decreasing (or for increasing) the number of works to be tested for leaks. Turning now to FIGS. 13 to 17, the second embodiment will be described below.

[0111] In the first embodiment, too, if the work under test is leak-free, the pressure difference variation measured within the measurement period t_ato t_bin step S6 in FIG. 6 represents the initial drift itself. In the second embodiment, even when the temperature environment change does not fall outside the predetermined range (that is, even when step S4 in FIG. 6 is not executed), if the current work of the same type as the immediately preceding tested work is decided to be free of leaks, the initial pressure difference variation δp_smeasured on the previous work is used as an estimated initial drift. By this, it is possible to determine the initial drift finely in response to the environmental temperature change.

[0112] FIG. 13 is a flowchart showing a leak test procedure of the second embodiment based on the idea referred to above, FIG. 14 a flowchart showing an initial drift calculating process of step S8′ in FIG. 13, and FIG. 16 a flowchart showing a pass/fail decision process of step S6′ in FIG. 13. In these flowcharts, the same processes as those in FIGS. 6 to 8 are identified by the same reference numerals and no description will be given of them. FIG. 15 depicts an example of a drift table and FIG. 17 the range of corrected amounts of drift which are stored in the drift table. Incidentally, the pass/fail decision process depicted in FIG. 16 is a concrete form of learning means.

[0113] The leak test procedure depicted in FIG. 13 is followed in substantially the same manner as in the procedure of FIG. 6, but differs therefrom in the two points described below. First, in step S3′, only when it is decided that the type of the current work under test is different from the type of the preceding work, the initial drift calculating process of step S8′ is executed following the flowchart of FIG. 14. Accordingly, on a first one of works under test which are different in type from a series of preceding works, the initial drift calculating process is executed in step S8′ as is the case with the first embodiment.

[0114] The other difference lies in that, in the test of a series of works of the same type, the initial pressure difference variation δp_smeasured on a work decided to be leak-free is estimated as and utilized as the initial drift at the current point in time, as depicted in FIG. 16 which is the pass/fail decision step S6′ in FIG. 13. In practice, however, the work decided to be leak-free may sometimes have a leak smaller than, but close to, the reference value L_th; therefore, it will impair the reliability of data to use the estimated initial drift intact as a correction value in the measurement on the next work. As shown in FIG. 16, according to the second embodiment, estimated amount of initial drifts (i.e. initial pressure difference variations) on works decided to be free of leaks are stored one after another in the drift table and an average (a moving average) of the plurality of previous initial drifts is used to correct the initial pressure difference variation measured on the next work. These processes will be described below.

[0115] The drift calculating procedure, which is followed when the type of the work under test is changed, differs from the procedure of FIG. 8 in that the initial drift p_dcalculated in step S88 is used to initialize the drift table in step G88. That is, when the initial drift on a work of the type different from the preceding works has been obtained, all drift data stored in the drift table on the type of the preceding works is erased. Alternatively, a drift table for the type of the current work under test is newly created (the process for which will hereinafter be called initialization of the drift table). The initial drift p_dis written as first data d₁in such a drift table. The initial drift p_dis derived from the difference between the initial and last pressure difference variations δp_sand δp_eas is the case with step S88 in FIG. 8.

[0116] In FIG. 15A, there is depicted a drift table 300, in which m pieces of drift data can be stored. The drift table 300 is provided, for example, in a predetermined area of the RAM 112 in FIG. 5. An appropriate value obtained empirically is used as the value m. The m pieces of drift data are stored in the drift table at addresses 1 to m, as depicted in FIG. 15A. On the first work after the change of the type of the work under test, the initial drift p_dis obtained following the procedure of FIG. 14, and is written as the data p_d1in the initialized drift table 300 at the address 1. On the succeeding works of the same type, the initial pressure difference variations δp_smeasured on the works decided as leak-free in the decision process of step S6′, respectively, are written as estimated initial drifts p_d2, p_d3, . . . in the drift table 300.

[0117] The decision procedure of FIG. 16 differs from that of FIG. 7 in the two points described below. A first difference lies in the inclusion of [step S61a] of calculating a moving average {tilde over (p)}_dof drift from the drift data stored in the drift table 300 and [step S61b] of calculating leakage p_L=δp_s−{tilde over (p)}_dfrom the moving average {tilde over (p)}_dand the initial pressure difference variation δp_sin the measurement period obtained in step S5 in FIG. 5. The moving average {tilde over (p)}_dmentioned herein is an average of the latest m pieces of data p_d1to p_dmstored in the drift table, and is given by the following equation. $\begin{array}{cc}{\tilde{p}}_d=\left(\sum _{i=1}^m{P}_{\mathrm{d1}}\right)/m& \left(6\right)\end{array}$

[0118] When the number of data is smaller than m, an average value of all data stored so far is calculated.

[0119] The other difference is the inclusion of [step S65] wherein, when the work under test is decided as leak-free (YES route of step S62), the pressure difference variation δp_smeasured on that work is estimated to be attributable only to drift and is then stored as the drift data p_din the drift table 300. In the event that m pieces of data have already written in the drift table 300, the oldest data is discarded and the latest data is written instead.

[0120] More specifically, in the leak test on the succeeding works of the same type as the work on which the initial drift p_dwas calculated in step S8′ in FIG. 13 after the change of the type of work and written as the data p_d1in the drift table 300, the pressure difference variation δp_son the work decided as leak-free for the first time is stored as the second data p_d2in the drift table 300 as depicted in FIG. 15A. Then, the pressure difference variation δp_son the work decided next as leak-free is stored as the third data p_d3in the drift table 300. Thereafter, pressure difference variations δp_son the next to mth-plus-one leak-free works are similarly stored as data p_dmin the drift table 300.

[0121] After this, the second to m-th data p_d2to p_dmin the drift table 300 are shifted from the addresses 1 through (m−1), respectively, and the pressure difference variation δp_son the next leak-free work is written as an mth-plus-one data p_dm+1in the drift table 300 at the address m as depicted in FIG. 15B.

[0122] In the second embodiment described above, the moving average {tilde over (p)}_dof the data stored in the drift table 300 at the addresses 1 to m is calculated, and then the leakage p_Lis calculated from the difference between the moving average {tilde over (p)}_dand the pressure difference variation δp_s{the third pressure difference variation}. Further, this pressure difference variation δp_sis additionally stored in the drift table 300 when the work under test is decided to be leak-free, and it is used to update the drift value (the moving average) which is used in the next work test. Since the pressure difference variation δp_sis susceptible to the influence of the ambient temperature change, the moving average {tilde over (p)}_dand the leakage p_Lalso changes, accordingly. Hence, even if ambient temperature changes, it is possible to make more accurate determination whether the work under test leaks or not. Before the number of data stored in the drift table 300 reaches m, the reliability of the average drift value can be enhanced by assigning larger weights to the first drift values Pd1 derived from the difference between the pressure difference variations δp_sand δp_eand then by averaging such weighted first drift values, rather than by obtaining simply averaged drift value. The reason for this is as follows: The estimated amounts of drift (i.e. the pressure difference variations δp_s) on the work decided as leak-free, which are held in the drift table 300, might contain leak components within the permissible range L_th. In contrast to this, the initial drift p_d1=δp_s−δp_e, which is calculated from the initial pressure variation δp_sand the last one δp_eobtained a relatively long period of time thereafter on the first work after the change of the work type, represents an amount of drift with no leak components; hence, this initial drift p_diis highly reliable.

[0123] In the second embodiment, when the work under test is free of leaks, the pressure difference variation δp_sis stored as an estimated drift value in the drift table 300 in step S65 in FIG. 16 so as to follow ambient temperature changes. In this instance, according to the pressure difference variation δp_sand the moving average {tilde over (p)}_don the current work, the pressure difference variation δp_sof an abnormal value might be written as an estimated drift value in the drift table 300. To avoid this, the range of the pressure difference variations for storage in the drift table 300 is defined in step S65. This range is one that is defined by the upper and lower limits p_maxand p_minwith reference to the initial drift p_das depicted in FIG. 17 (which range will hereinafter be called an “allowable range”).

[0124] The values p_maxand p_minare chosen such that (p_max−p_d) and (p_d−p_min) are smaller than the reference value L_th, and only pressure difference variations δp_swithin this allowable range are written as estimated drift values in the drift table 300. With this method, no pressure difference variations δp_sof abnormal values will be written in the drift table 300. Consequently, only pressure difference variations δp_swhich are highly reliable as drift values are stored in the drift table 300—this provides increased reliability in the determination as to whether the work under test leaks or not.

[0125] Provision may be made so that when the pressure difference variation δp_soutside the abovementioned allowable range is measured more than a predetermined number k of times, the process of step S8′ is executed even if the decision by step S3′ is YES. This is advantageous for the reason given below. That is, when the measurement operations are temporarily suspended, for example, at lunchtime or on holidays, environmental conditions often change sharply. In such an instance, previously calculated drifts p_dand data stored in the drift table 300 mostly become useless.

[0126] Such forced execution of step S8′ provides an initial drift p_dand data δp_sin the drift table 300 which match the current environmental conditions of measurement. Hence, even if the measurement operations are suspended, an accurate determination as to whether the pressurized gas leaks out of works under test can be made immediately after the resumption of the measurement operations.

[0127] In the second embodiment, the initial drift p_dis calculated for each type of work (steps S3′ and S8′ in FIG. 13). In the case where initial drifts p_dcorresponding to air temperatures have been obtained based on the past measurement data for each of the types of work W1, W2, . . . , a table of their correspondence may be prestored in a memory, for example, as shown in FIG. 18 so that the initial drift p_dcorresponding to the type of the current work and air temperature is read out of the table and set in step S8′ in FIG. 13.

[0128] This permits rapid setting of the initial drift p_dand hence hastens the beginning of the leak test on the work to be measured.

[0129] In the case of storing the pressure difference variations δp_sin the drift table 300 in step S65 in FIG. 16, the initial drift p_dand the pressure difference variation δp_smay be weighted according to the type of the work concerned. For example, letting weighting coefficients of works of the types W1 and W2 be represented by w₁and w₂, respectively, the pressure difference variations δp_son the works of types W1 and W2 are multiplied by the weighting coefficients w₁and w₂, respectively, and then stored in the drift table 300.

[0130] With this method, all works to be measured can be tested for leaks as usual without precalculating the initial drift p_dfor each type. Accordingly, works of a flexible manufacturing system can be tested for leaks in succession with high accuracy.

[0131] In the case where pressure difference values measured on works decided as non-defective are sequentially stored and their moving average is used as a drift correction value, a highly reliable leak test can be made without a hitch when the work and master temperatures gradually vary with ambient temperature.

[0132] However, when a work still hot is tested, for example, soon after a high-temperature cleaning or welding process, the work temperature changes more rapidly than room temperature (approaches the latter). In such a situation, the moving average for use as the drift correction value cannot quickly respond to the temperature change of the work, and consequently, the test results lack reliability. In particular, when a work whose temperature is appreciably higher than room temperature is connected by a jig to a conduit, the temperature difference between the work and the jig is so great that a large quantity of heat rapidly escapes through the jig, resulting in the work temperature quickly dropping and the jig temperature gradually rising. Under such test conditions, the amount of drift of the pressure difference is considered to rapidly change depending on the temperature difference between the work and the jig as well.

[0133] In the third embodiment, the amount of drift p_dis estimated not from the moving average but instead it is expressed as the following linear function of the difference between the work temperature T_Aand the jig temperature T_B:

[0134] P_D=K(T_A−T_B)+C  (7)

[0135] and the drift p_dis estimated by Eq. (7) from the temperature difference (T_A−T_B) detected on the work under test.

[0136] To obtain the constant C in Eq. (7), a drift calculating process is executed in advance under the condition that the work temperature T_Aand the jig temperature T_Bare equal. This process provides a constant P_D′=C=δps−δp_e=p_don the work when T_A=T_B. δp_sand δp_erepresent the initial pressure difference variation in the measurement period and the last pressure difference variation in the stabilization period (the period in which the drift can be regarded as having converged to zero) described in the first or second embodiment; hence, the latter represents the amount of leakage of the work. That is, when T_A=T_B, the constant P_D′ is the same as the initial drift p_dcalculated in the drift calculating process of the first and second embodiments.

[0137] Next, the work is mounted by a jig 25 (see FIG. 19) on the conduit after being heated so that the work's temperature T_Ais higher than the jig temperature T_B, and the initial pressure difference variation δp_Sis measured in the measurement period. This initial pressure difference variation δP_Scan be given by the following equation through the use of Eq. (7). $\begin{array}{cc}\begin{array}{c}\delta P_S=P_D+\delta p_s\\ =K\left(T_A-T_B\right)+C+\delta p_e\\ =K\left(T_A-T_B\right)+\delta p_s\end{array}& \left(8\right)\end{array}$

[0138] Substituting the initial pressure difference variation δp_smeasured with T_A=T_Band the initial pressure difference variation δP_Smeasured with T_A>T_Binto Eq. 88), the proportional constant K can be calculated as given below.

[0139] K=(δP_S−δp_s)/(T_A−T_B)  (9)

[0140] In this way, the constant C and the proportional constant K in Eq. 87 are determined. In the subsequent leak test mode, the detected work and jig temperatures T_Aand T_Bare used to calculate the estimated drift P_Don the heated work by Eq. (7), and the estimated value is subtracted from the initial pressure difference variation δp_sto estimate the leakage of the work.

[0141] FIG. 19 illustrates the functional configuration of a leak test apparatus according to the third embodiment of the present invention. This embodiment differs from the FIG. 4 embodiment in the provision of a temperature sensor 46A for detecting the temperature T_Aof the work 22W and a temperature sensor 46B for detecting the temperature T_Bof the jig 25. The detected temperatures are provided to the drift calculating part 31. The temperature sensors 46A and 46B may each be formed, for example, by a thermocouple. The temperature sensor 46A is disposed near the jig 25 so that the work 22W placed on the jig 25 inevitably contacts the temperature sensor 46A which measures the work temperature T_A. The temperature sensor 46B is mounted directly on the jig 25 to measure its temperature T_B. The detected signals from the temperature sensors 46A and 46B are provided to the control 100 and used to calculate the drift value. The control 100 comprises the valve control part 26, the drift calculating part 31, the leakage calculating part 32, the pass/fail decision part 33, and the sequence control part 34 as in the case of the FIG. 4 embodiment. In this case, however, the process in the drift calculating part 31 is a modification of the process depicted in FIG. 8 as described later on. The control 100 is formed by a computer as is the case with FIG. 5.

[0142] FIG. 20 is a flowchart of the leak test according to the third embodiment. This teak test procedure is basically similar to that depicted in FIG. 5 but differs therefrom in the points listed below.

[0143] (a) In step S80, as described later on with reference to FIG. 21, the constants C and K of the linear approximate expression of drift (7) are precalculated for the type of each work, and prestored in a memory.

[0144] (b) In step S30, the constants C and K of the drift linear approximate expression corresponding to the type of the work under test are read out of the memory.

[0145] (c) In the pass/fail decision process of step S60, as described later on with respect to FIG. 22, the drift P_Dis calculated, as the drift to be used for calculating the amount of leakage in step S61, from the already detected work and jig temperatures T_Aand T_Bby Eq. (7), and the thus calculated drift P_Dis subtracted from the initial pressure difference variation δP_Sto obtain leakage δP_E.

[0146] Turning next to FIG. 21, the drift linear approximation will be described. In the third embodiment, it is necessary to set up a linear approximate expression of drift on the type of each work in advance. In step S80s, a desired type of work held at room temperature is mounted on the jig 25. In steps S81 to S88, the initial pressure difference variation δp_sin the measurement period (step S83) and the last pressure difference variation δp_ein the stabilization period (step S87) are calculated on the work with the work and jig temperatures T_Aand T_Bheld equal by the same processes as those of steps S81 to S88 in FIG. 12, and the drift p_d=δp_s−δp_eunder the condition of T_A=T_Bis calculated (step S88). This drift pd is defined as the constant C in Eq. (7) (under the condition T_A=T_B). In the next step S89, the work is demounted from the jig 25, then heated up to a desired temperature, and mounted again on the jig, followed by introducing pressurized gas into the work and master up to a desired pressure. In steps S91 to S93 the initial pressure difference variation δp_sunder the condition T_A≠T_Bis calculated, and the work and jig temperatures T_Aand T_Bat that time are measured. In step S94 the constant K is determined by Eq. (7) from the initial pressure difference δp_sunder the condition T_A=T_B, the initial pressure variation δP_Sunder the condition T_A≠T_Band the measured temperatures T_Aand T_B, and the constants C and K are stored in a memory in correspondence with the type of the work. In step S95, a check is made to see if the linear approximation of drift has been completed on all types of works, and if not completed, then the procedure returns to step S80; if completed, the drift linear approximation procedure ends. The processes of steps S1, S2, S4, S5, and S7 in FIG. 20 are the same processes as in the first embodiment of FIG. 6. In the third embodiment, the linear approximate expression of drift (7) is set up in advance on every type of work to be tested for leaks and its constants C and K are prestored in a memory in correspondence with each type of work in step S80 as described above. To obtain the linear approximate expression of the drift of each work, the initial pressure difference variation δp_sin the measurement period and the last pressure difference variation δp_ethe stabilization period are calculated with the work and jig temperatures T_Aand T_Bheld nearly equal by the same processes as in FIG. 8,12, or 14 and the difference p_d=δp_s−δp_e=C is calculated. Next, the work is heated up to a desired temperature (T_A>T_B) and mounted on the jig, followed by measuring the initial pressure difference variation δP_Sin the measurement period and then calculating the constant K from Eq. (9) using the initial pressure difference variation δp_smeasured when T_A=T_B.

[0147] In the leak test mode (steps S1, S2, S30, S4, S5, S60, and S7) in FIG. 20, the type of the work mounted on the jig 25 is detected and the corresponding constants C and K are read out of the memory in step S30. This process may be performed in the pass/fail decision step S60. In the pass/fail decision process of step S60, as depicted in FIG. 22, the drift P_Dis calculated by Eq. (7) from the constants C and K read out in step S30 and the measured current work and jig temperatures T_Aand T_B. Then, the initial pressure difference variation δp_smeasured in step S5 is used to calculate the leakage δP_Eby the following equation (step S61).

[0148] δP_E=δp_s−P_D  (10)

[0149] Following this, the absolute value of the leakage δP_Eis compared with the reference value L_th(step S62). If the former is smaller than the latter, the work under test is decided to be free of leaks (step S63), but if the former is larger than the latter, the work is decided to be leaky (step S64).

[0150] While the first, second and third embodiments have been described to introduce high-pressure gas into the work 22W and the master 22M to test the work for leaks, it is evident that the present invention is also applicable to a leak test in which the work 22W and the master 22M are evacuated of gas by means of a pressure reducing pump to test for leaks into the work 22W from outside.

[0151] As described above, according to the first embodiment of the present invention, the pressure difference variation is measured at least twice in the drift calculation process. Each measurement is carried out after a pressure equilibrium is established between the work and master prior to the closing of the valve 20. This makes smaller the step size of variation of the pressure difference to be measured than in the prior art, and hence allows the use of a high-sensitivity differential pressure sensor. Consequently, the value of drift that is obtained by subtracting the second pressure difference variation from the first one becomes accurate accordingly. The calculation of the drift does not require the use of a leak-free, non-defective work but can be made using a given work. In the second embodiment of the present invention, the initial pressure difference variation δp_smeasured on the work decided to be leak-free is stored as an estimated drift value in the drift table, and a moving average of the data stored in the drift table is used as the drift value of the next work. This provides a drift value that follows a temperature change, and hence makes it possible to more accurately determine whether the work leaks or not, irrespective of a temperature change or the like.

[0152] In the third embodiment of the present invention, the drift is regarded as a linear function of the difference between the work and jig temperatures and its linear approximate expression is obtained in advance on every type of work. By calculating the drift of a heated or cooled work from the linear approximate expression and using the drift for calculation of the leakage, the work of a given temperature can be tested for leaks. In the case of setting up this linear approximate expression of drift, the initial drift is measured under the condition that the work and jig temperatures are equal, as is the case with the first embodiment. That is, the measurement of either of the first and second pressure difference variations is immediately preceded by opening the valve 20 to establish a pressure equilibrium between the work and the master. This provides a highly accurate value of drift.