means of two screw plugs (fig. 416). Very serious defects have occasionally been exposed by this means.

Having obtained a good material, several precautions have to be taken to ensure a good weld. The heating should be carried to the right point. This knowledge can only be gained by practice. Steel, unlike iron, should not be heated till sparks make their appearance.

Both sides of the plate should be exposed to the fire. If this is impossible the joint must be left wide open, to allow the flame to pass to the other side, where a cap of firebrick is placed. Having completed a short length of weld, and while heating the next, care should be taken to let the escaping flame pass over the recently welded part so as to keep it hot, otherwise the resultant irregular contraction will produce cracks.

The anvil on which the welding takes place should be as solid as possible. Any appreciable amount of spring affects the quality of the

seam.

FIG. 417

Fluxes are used for iron, but seldom for steel. The edges of the plates, whether of steel or iron, are slightly tapered, as shown (fig. 417), care being taken to keep the surfaces convex, so that the centres touch

first and the slag thereby gets driven out. Or the edges are formed into V's (fig. 418). They are first pressed firmly together, and then hammered. A very convenient method of welding steel plates together is to insert a separately-heated good weldable iron bar into the seam. Where practicable this piece is shaped like a double-headed rail (fig. 419).

Fires for Welding. The heating of the plates is usually done over coke fires, using a blast. Much time is saved if these furnaces are filled and replenished with red-hot coke taken out of an adjoining furnace. The air blast for welding ought not to pass through more than about 18 ins. of coke, otherwise the combustion is not perfect, and the flame not hot enough. (See W. van Folten, Stahl und Eisen,' 1893, p. 26.)

Fig. 420 shows a movable furnace delivering its flame in a horizontal direction. It can only be used with plates placed vertically, and then it is usual to place one furnace on each side.

Longitudinal seams in furnaces are very conveniently heated as shown in fig. 421. As the flame strikes only one side of the plate, it

is necessary to prevent radiation by the little firebrick cover C. It is also necessary to keep the seam wide open at this point, by bending

back its edges, so that the flame can pass through it; otherwise only one side of the plate will weld. When ready, the furnace plate is lifted and turned round by means of the pulley P. This can be dispensed with by placing the furnace inside of the flue, as shown in fig. 422. In this case, when the seam has been sufficiently heated it is placed over the projecting anvil A, and welded.

Where gas is used for heating purposes, the arrangement shown in fig. 423 is a very convenient one, the ignited mixture of gas and air

passing down the tube T. A is the anvil, and H the steam hammer. It is dangerous to use fans for supplying the air to the gas, as their action is not so reliable as that of any of the positive blowers, and causes explosions.

When the tube plate and furnace are to be welded together, this is usually done at the corner (see fig. 424). No flanging is then

necessary.

But another plan may be mentioned, according to which the tube plate is first flanged so as to form a saddle, and this part is then welded to the furnace (fig. 425). This method is particularly convenient for use with patent flues, which, with one exception, are weakest near the combustion chamber end. With the above plan the thick metal of the tube plate supports the flue.

Although not carried out to any great extent, all the seams of the furnaces and combustion chambers can be welded, and the finished article will then have the appearance shown in fig. 426.

Electric welding is said to give excellent results, but has not yet been used for boiler work. Welded seams are tested with oil to see whether they are perfect.

Hydraulic Test.-Although not forming part of the construction, the final hydraulic test is the concluding operation before the boiler is put into the ship. It is usual to have a preliminary test the day before, when any defective caulking can be made good, but with good work

FIG. 426

this should not be necessary. The generally adopted plan is to raise the pressure step by step, and at once caulk any defect which shows itself. If the full pressure is put on at once, the leakages may be so excessive that they cannot all be put right. During this preliminary test the stay nuts are not screwed on, so that the stays may if necessary be recaulked.

Much has been said and written against testing boilers to double the working pressure, but, in spite of assertions to the contrary, riveted seams, and even welds, which were found to be perfectly tight with the cold test, commenced to leak at half that pressure when hot. This is probably due to the difference in the conditions of testing, and not, as is often stated, to the previous excessive proof stress.

It is also argued that an hydraulic test not exceeding the working pressure will detect defective material and flaws. This is not borne out by experience (see p. 227).

In one sense, the hydraulic test can therefore be looked upon as a guarantee of good material, although its chief object is to detect bad workmanship. This will show itself by leakages.

The amount of water which has to be pressed into a boiler amounts to about per cent. of its gross volume, unless all the air has not been removed, or unless the boiler is weak, in which case more water has to be pumped into it.

Boiler Deformations.-Observations as to the deformations are also made, but it cannot be said that they have been of much assistance in detecting local weaknesses; but they undoubtedly offer a means of studying the actual stresses in boilers, and a few remarks on the subject will not be out of place. That they are very much larger than most engineers expected was shown during the discussion on J. T. Milton's paper (N. A.,' 1893, vol. xxxiv. p. 157) on this subject.

The stretching of the boiler circumference can be roughly measured by coiling a wire 1 times round it (fig. 427) and weighting the two

ends, and then marking the two wires at the top of the boiler before and during the test. The stretch ought to be about 10 of the circumference, or, say, 3 in. for an 11-ft. boiler, and in. for a 13-ft. boiler. It will be found that the reading is less, and also, on relieving the pressure, that the original marks do not coincide. This is not permanent set, but is due to the friction between the wire and the shell. More accurate results can be obtained if one end of the wire is bolted to the boiler and small rocking frames interposed between the shell and the wire at other points (see fig. 428). The wire is secured by means of a small bolt to the rivet-head A, and led from one rocking frame to another till it reaches R, and the weight W is then hung to the loose end, while a scale is secured partly to the bolt-head A and